Professional Documents
Culture Documents
Dynamics of Offshore Structures PDF
Dynamics of Offshore Structures PDF
Offshore
Structures
James F. Wilson, Editor
Copyright © 2003 by John Wiley & Sons, Inc. All rights reserved.
Limit of Liability/Disclaimer of Warranty: While the publisher and author have used
their best efforts in preparing this book, they make no representations or warranties
with respect to the accuracy or completeness of the contents of this book and specifi-
cally disclaim any implied warranties of merchantability or fitness for a particular
purpose. No warranty may be created or extended by sales representatives or written
sales materials. The advice and strategies contained herein may not be suitable for
your situation. You should consult with a professional where appropriate. Neither the
publisher nor author shall be liable for any loss of profit or any other commercial
damages, including but not limited to special, incidental, consequential, or other dam-
ages.
For general information on our other products and services or for technical sup-
port, please contact our Customer Care Department within the United States at
(800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002.
Wiley also publishes its books in a variety of electronic formats. Some content that
appears in print may not be available in electronic books.
TC1665.W55 2002
627⬘.98—dc21 2002028858
10 9 8 7 6 5 4 3 2 1
Contents
Preface Xl
Contributors Xlll
Acknowledgments XV
V
VI CONTENTS
Cable Restraints, 43
Soil Foundation Restraints, 52
Problems, 54
References, 56
Jumps, 126
Subharmonics, 129
5.6 ='Jonlinear Responses for a SALM Buoy, 131
Physical Description and Dynamic Model, 132
='Jumerical Results and Conclusions, 136
Problems, 138
References, 142
JAMES F. WILSON
Chapel Hill, North Carolina
XI
Contributors
Bruce J. Muga, Professor Emeritus of Civil and Environmental Engineer-
ing at Duke University, received his B.S. in Civil Engineering from the University
of Texas, and his M.S. and Ph.D. degrees in Civil Engineering (Hydrodynamics)
from the University of Illinois.
From 1961 to 1967 he was employed as a Project Engineer in the Port and
Harbor Division of the U.S. Naval Civil Engineering Laboratory, Port Buen-
eme, California. In 1964, he was assigned as Consultant to the U.S. Military
Assistance Command, Vietnam, to advise on coastal and harbor engineering
projects.
In 1967, Dr. Muga accepted a position in teaching and research at Duke
University and was Chairman of the Department of Civil Engineering in 1974.
He has served as a consultant to many international corporations engaged in off-
shore and deep ocean engineering activities. He has written numerous technical
papers and for seventeen years served on the North Carolina Marine Sciences
Council. Prior to retirement, Dr. Muga was a Registered Professional Engineer
in California, a member of the American Society of Civil Engineers and the
Marine Technology Society. He is a life member of the Permanent International
Association of Navigational Congresses.
Xlll
xiv CONTRIBUTORS
JAMES F. \V1LSON
Chapel Hill, North Carolina
XV
1
Structures in the
Offshore Environment
James F. Wilson
2 STRUCTURES IN THE OFFSHORE ENVIRONMENT
In perspective, offshore structures include a great deal more than the towers
and platforms. They include moored or mobile ships whose positions may be
precisely controlled. They include the guy lines for compliant towers, the cables
for buoys and for tension-leg platforms, and the associated pipelines without
which the platforms and submerged oil production systems would be useless.
Detailed descriptions of such installations may be found in the references at the
end of this chapter. Of particular note is the review article on compliant offshore
structures by Adrezin et al.(1996), with its 130 dtations to the world literature
on the subject up to the mid-1990s. For descriptions of current practice in all
types of offshore installations, the reader is referred to the yearly conference
proceedings such as found in the References at the end of this chapter.
This chapter begins with a short history of offshore structures, describes typ-
ical state-of-the-art installations, and concludes with a discussion of engineering
challenges for future designs. Subsequent chapters address in sorne detail both
the mathematical modeling and the environmental loading of offshore struc-
tures, together with ways to predict their dynamic responses and structural
integrity, from both the deterministic and the statistical viewpoints.
The earliest offshore structure for oil drilling was built about 1887 off the coast
of southern California near Santa Barbara. This was simply a wooden wharf
outfitted with a rig for drilling vertical wells into the sea floor. More elaborate
platforms supported by timber piers were then built for oil drilling, including
installations for the mile-deep well in Caddo Lake, Louisiana (1911) and the plat-
form in Lake Maracaibo, Venezuela (1927). Soon after these early pier systems
were built, it became apparent that the lifetime of timber structures erected
in lakes or oceans is severely limited because of attacks by marine organisms.
For this reason, reinforced concrete replaced timber as the supporting structure
for many offshore platforms up to the late 1940s. Over the next 50 years about
12,000 platform structures were built offshore, usually of steel but more recently
of precast concrete. The chief features of these structures, together with their
supporting components such as mooring systems and pipelines, are discussed in
this chapter. See also Gerwick (1999) and Will (1982).
Offshore mooring systems have a variety of configurations. All have anchors
or groups of piles in the seabed with flexible lines (cables, ropes, chains) leading
from them to buoys, ships, or platform structures. The function of a mooring
system is to keep the buoy, ship, or platform structure at a relatively fixed
location during engineering operations. Engineering efforts in mooring systems
have focused in recent years on the development of new anchor configurations
with higher pullout loads, larger capacity and lower cost of installation for deeper
water applications.
When pipelines were first laid offshore, no extraordinary analyses or deploy-
ment techniques were needed since they were in shallow water and were of small
diameter, somewhat flexible, and made of relatively ductile steel. As platforms
were built in deeper and deeper water with multiple well slots, larger diameter
HISTORICAL PERSPECTIVE 3
pipelines of higher strength were required. During the 1960s, engineers met
this challenge with new designs and with refined methods of analysis and de-
ployment. Pipeline systems evolved into two main types: sea fioor and vertical
configurations. Both are used to transport gas and oil, but the vertical sys-
tems also include risers to carry drilling tools, electric power lines, dredge pipes
for deep sea mining, and cold water pipes (CWP) for ocean thermal energy
conversion (OTEC).
Throughout the world there are at present about 80,000 km of marine
pipelines. Since 1986, the rate of building new marine pipelines has been about
1000 km per year. Individual pipelines on the sea fioor vary in length from 1 to
1000 km and in diameter from 7 to 152 cm. For instance, a Norwegian project
features a 1000 km line extending from the Troll field to Belgium, which was
completed in 1992. At present, Kuwait has the loading line of largest diameter,
152 cm. The pipelines of smaller diameter are used to transport oil and gas
from wellheads, and those of larger diameter are used to load and unload oil
from tankers moored at offshore terminals. The deepest sea fioor pipelines at
present are the 46 cm diameter gas lines in the Gulf of Mexico, for which the
e 5oo
~
1!l1ooo
a::
1:!
~ 1500
2000
(f)
Figure 1.1 Six offshore platforms at their maximum depths: (a) jackup rig; (b)
gravity platform; (e) jacket structure; (d) compliant tower; (e) tension leg platform;
(f) semisubmersible.
4 STRUCTURES IN THE OFFSHORE ENVIRONMENT
maximum depth is 1400 m. Sea floor pipelines are often anchored to the seabed
or buried in trenches for protection from erosion and the undermining effects of
currents. Sorne seabed pipelines have a coating of concrete to add protection
and to reduce buoyancy.
1.2 PLATFORMS
Six general types of offshore platforms are depicted in Figure l. l. The first three
are designed for depths up to about 500 m, and the last three are for depths to
2000 m. Not shown are subsea production platforms, which are presently rated
for 3000 m depths.
Fixed-Bottom Platforms
A mobile structure often used for exploratory oil-drilling operations is the
self-elevating platform commonly called a jackup or mat-supported rig. A con-
structed version of this platform, depicted schematically in Figure l.la, is shown
in Figure 1.2. Typically, such a platform is supported by three to six legs that
are attached to a steel mat resting on the sea floor. In soft soils, the legs pass
through the mat and may penetrate the soil to depths of up to 70 m. To the
bottom of each leg is attached a steel saucer or spud can to help stabilize the
PLATFORMS 5
structure and to minimize leg penetration into the soil. The height of the
platform above the seafloor, up to 100 m, may be adjusted by using motor
drives attached to each leg.
With ballast consisting of sand in the bottom of the tanks and seawater in
the legs, these structures depend on self-weight alone to maintain an upright
position when subjected to the highest waves that are expected to occur in a
100 year time period. A realistic 100 year wave that may occur in the northern
North Sea is 27.8 m. At present, the largest concrete gravity platform is the
Troll structure, and one of moderate size is the Statfjord-A Condeep structure,
both located in the North Sea. The latter structure is 250 m high and has
three legs. Located off the coast of Norway, the Statfjord-A Condeep unit has
slots for 42 oil wells that reach to depths of 2800 m. When in operation, it
accommodates a crew of 200 people who live and work on this structure.
COMPLIANT ~\ JACKET
TOWER : : TEMPLATE
i,'\~
11\i
~ ,¡
i \
l ~
1 '\
1
t
'~~\-- ... ~
J '·······-~~=-.-:-.:.,._...
o 0.1 0.2
WAVEFREQUENCY, Hz
Figure 1.4 Two storm wave height spectra in the Gulf of Mexico, showing two offshore
structures with natural frequencies beyond the frequencies of the highest energy waves.
Compliant Platforms
An alternative class of offshore structures meant for depths from 300 to 800
m is the compliant tower such as that shown in Figure 1.1d. Such a tower
may or may not have mooring lines. It is a pile-supported steel truss structure
designed to comply or flex with the waves and has considerably less structural
material per unit height when compared with a common jacket template tower.
The first compliant tower was the Lena, which was installed in the early
1980s in the Gulf of Mexico. Including its three-level drilling and production
deck and its drilling rigs, this tower reaches a total height of 400 m. Each
of the 20 stabilizing cables, attached 25 m below the water line and arranged
symmetrically about the structure, extends a horizontal distance of about 1000
m to a line of clumped weights that rest on the sea floor, to an anchor cable
and an anchor pile. Under normal weather or small storm conditions, the cables
act as hard springs, but with severe storms or hurricanes, the cable restraints
become softer or compliant. That is, the amplitude of tower rotation increases
at a rate greater than that of the loading, since the clumped weights lift off the
sea floor to accommodate the increased storm loads on the tower. When storms
or hurricane conditions are anticipated, operations on compliant towers cease
and the crew is evacuated.
Installation of the Lena cables was more difficult and costly than anticipated.
Subsequently, compliant towers without cables have been designed by Exxon,
and two such designs were installed in 1999 in the Gulf of Mexico. U:nlike
the jacket-template structures, the compliant towers have natural frequencies
in bending or sway near 0.03 Hz, or well below the 0.05 Hz frequency of the
highest energy sea waves in the Gulf of Mexico during storm conditions. Thus,
an important feature of such structures is that they are designed to have natural
sway frequendes well removed from the frequency range of the highest energy
waves for normal seas (0.1 to 0.15 Hz) and for stonn seas (0.05 to 0.1 Hz). This
frequency spread is necessary to avoid platform resonance, which can lead to
failure. The sway frequencies of two platforms in comparison to the frequency
range for the spectrum of the highest energy storm waves in the Gulf of Mexico
are depicted graphically in Figure 1.4. The measurement and meaning of this
wave height spectra, which is highly site-dependent, will be discussed in detail
in subsequent chapters.
Buoyant Platforms
The tension leg platform (TLP) can be economically competitive with com-
pliant towers for water depths between 300 m and 1200 m. The schematic design
of the TLP is depicted in Figure 1.1e. In such designs, the total buoyant force
of the submerged pontoons exceeds the structure's total gravity or deadweight
loading. Taut, vertical tethers extendi:ng from the columns and moored to the
foundation templates on the ocean floor keep the structure in position during
all weather conditions. The heave, pitch, and roll motion are well restrained by
the tethers; but the motions in the horizontal plane, or surge, sway, and yaw,
are quite compliant with the motion of the waves. The first production TLP
8 STRUCTURES IN THE OFFSHORE ENVIRONMENT
was built 150 km off the coast of Scotland in the mid-1980s. Conoco installed
the Julliet in 1989, and Saga Petroleum installed the Snorre near Norway in
1991. The tethers for the Snorre are 137 cm in diameter. By the late 1990s,
a total of eleven TLPs were installed, three in the North Sea and eight in the
Gulf of Mexico.
For water depths of about 1500 m, a subsea production system provides an
excellent alternative to a fixed surface facility. Much of a subsea system rests
on the ocean floor, and its production of oil and gas is controlled by computer
from a ship or other buoyant structure above the subsea unit. The buoyant
structure and the subsea unit are often connected by a marine riser, which will
be discussed presently.
1.3 MOORINGS
Temporary Anchor Moorings
A classical example of ternporary offshore mooring is the spread mooring
configuration for a ship in relatively shallow water. Six to eight cables of wire
rope or chain are unreeled from onboard winches symmetrically placed around
the perimeter of the ship. Tug boats aid in spread mooring installations. In
place, each cable hangs as a catenary curve and is attached either directly to
a drag embedment anchor in the seabed or to a buoy that is anchored. An
example of a spread mooring configuration is shmvn in Figure 1.6. Particular
mooring configurations were reported by Baar et al. (2000) and O'Brian and
Muga (1964).
SWAY
PLAN
VIEW
SIDE
VIEW z ~
: :,:...:m
Figure 1.6 A spread mooring configuration.
nJIINTABLE
BUOY
BUOY_,.,F!!!!!!!:¡;:-·
1 1 1 1
1 1 \
(a) (b) 1 f 1 l MOOIUNG CHAlNS
1 1 \ \ (60R8)
¡ 11-....---+TWIN HOSBS
1
PIPELINE MANIFOLD
Figure 1.8 Two single-point mooring systems: (a) a single anchor leg mooring
(SALM); (b) a catenary anchor leg mooring (CALM).
Successful SPMs are found at the world's largest terminals. One of these is
Saudi Arabia's Ju'aymah exporting terminal, which has two SAL~f buoys and
four CALM systems. For this SPM system, crude oil and fuel oil are loaded
simultaneously to a moored tanker through the swivel assembly on the seabed.
A second example is the CALM system for service vessels associated with the
Cognac platform in the Gulf of Mexico. This SPM buoy has 12 catenary lines,
each anchored to O. 76 m diameter piles. The water depth here is 275 m.
1.4 PIPELINES
Sea Floor Pipelines
Most of the 80,000 km of offshore pipelines have been installed by one of
the following three methods. In these methods, the deployed pipeline forros an
S-shape between the vessel and the sea floor.
TENSIONER
SEAFLOOR
Figure 1.9 The laybarge method of pipeline construction offshore.
12 STRUCTURES IN THE OFFSHORE ENVIRONMENT
l. Laybarge method. Pipe sections, which sometimes have been coated pre-
viously with concrete for protection, are welded together on the deck of a barge
and deployed on rollers over the stern. Near the stern, the pipeline passes
over pontoons called stingers that relieve excess bending in the pipeline as it is
deployed to the seabed. See Figure 1.9.
2. Reel barge method. Small to medium diameter pipe sections (up to 41 cm
in diameter) are prewelded and coiled onto a reel mounted to the deployment
vessel. As the pipeline is unreeled at sea, it passes through straightening rollers
and then deployed as in the laybarge method.
3. Bottom pull method. Pipe sections are assembled on shore, and the pipe
string is towed into the sea by a barge. During the launch to its place on
the seabed, pontoons are often used under the string to avoid excess pipeline
bending.
The following two methods of pipeline deployment have been studied exten-
sively but have yet to be used.
l. J-lay methorL A dynamically positioned vessel such as a drill ship, a
converted pipelaying vessel, ora semisubmersible may be used in this operation.
On board the vessel a derrick is used to hold the pipeline vertically as it is
lowered, and the pipeline forms a J-shape between the vessel and the seaftoor.
An efficient single station pipe welding procedure has to be developed before
this method can be adapted to common practice.
2. Floating string method. According to this concept, the pipeline is ftoated
on pontoons at the water surface. Then the pontoons are released successively
so that the pipe string gradually sinks to the sea ftoor.
Vertical Pipelines
One type of vertical offshore pipeline is the marine riser, which is shown
for several of the structures in Figure 1.1. Although marine risers make up
a fraction of the 80,000 km network, they are nonetheless key components of
offshore structures and serve a variety of functions. For example, a riser may
contain a bundle of smaller pipelines connecting a wellhead to its platform, or
it may transport oil directly from its platform to a sea ftoor pipeline. A riser
may act as a drilling sleeve, or it may contain electric power lines for operating
seafioor mining vehicles or other subsea facilities. A typical drilling riser, with a
hall joint at the bottom and a telescoping joint at the top, is maintained under
tension to ensure stability and is kept to within 8 degrees of the vertical by
computer-controlled positioning of its parent drillíng vessel. In the mid-1980s,
the longest riser \vas operating off the east coast of the United States, where
depths of 2100 m were reached. At present, risers for depths approaching 3000
m are becoming a reality.
Vertical pipelines are used in deep-sea dredging operations. A most chal-
lenging engineering problem, which began to receive serious attention in the
late 1960s, was that of designing dredge pipes to suck up and transport man-
ganese nodules from depths of 3000 to 5500 m to the ship. See Lecourt and
Williams (1971).These nodules, about the size of a man's fist and found in a
PIPELINES 13
monolayer on the sea floor in many parts of the world, contain other minerals
as well, including cobalt, copper, and nickel with traces of molybdenum, vana-
dium, and titanium. By the mid-1980s, remote-controlled mining of the nodules
was achieved with a 5500 m long dredge pipe. In this design, air bubbles are
pumped into the string at various points along its length to aid in pumping
the nodules to the surface. The dredge pipe's position is computer-controlled,
synchronized to the movement of both the self-propelled mining scoop at one
end and the ship at the other end. A schematic diagram of this system is shown
in Figure l.lüa.
(a)
MAN<lANESE
NODULES
Figure 1.10 Two vertical pipeline systems: (a) a dredge pipe for mining manganese
nodules; (b) a cold water pipe for ocean thermal energy conversion.
Vertical cold water pipe system..<; for ocean thermal energy conversion (OTEC)
have been studied extensively. See \:Vilson et al. (1982). Calculations show that
practica! full-scale pipe designs are 10-20 m in diameter, are about 1000 m long,
and may be used to raise the cooler water in the deep ocean to the warmer
surface water. \Vith the differential temperature between the ends of the pipe
(10-20°C), it is possible to produce net power through heat exchange. Such a
system is depicted in Figure l.lüb. In 1979, a small-scale OTEC power plant
operated successfully on a barge 2.3 km offshore of Kona, Hawaii. This pipe
was 61 cm in diameter, 660 m long, and made of polyurethane. The CWP was
supported by a surface buoy at the barge stern and was tension-moored with
14 STRUCTURES IN THE OFFSHORE ENVIRONMENT
1.5 CHALLENGES
To meet the need for new sources of energy and minerals, marine engineers must
work at the frontiers of known technology. Their main challenge..'l are to design,
deploy, and operate facilities and equipment in environments where none have
before existed--in deeper and deeper water, on the slopes of the continental
shelves, and in the hostile Arctic seas. To meet the challenges, marine engineers
continue their research and development efforts in the following four broad and
necessarily overlapping topics of concern.
Environmental Forces
Every offshore facílity is subjected to severa! types of environmentalloads
during its lifetime. Thus, site-dependent databases are being developed to char-
acterize the time-varying fluid-induced loads of winds, currents, and waves.
Such loads occur both on a short time scale of seconds and minutes, induced
by periodic vortex shedding, wind gnsts, and wave slamming; and over longer
periods of hours, days, and even years, where the loads are induced by steady
waves, tides, and hurricanes. At sorne sites, data are also needed on subsea
earthquake intensity, on the scouring of sea floor foundations by currents, and
on the total and differential settlements of the sea floor due to the withdrawal
of hydrocarbons during the lifetime of the structure. At present, there is the
particular challenge of minimizing differential settlements for gravity platforms.
For the Ekofisk structural complex in the North Sea, however, this challenge
appears to have been met by employing water injection. In Arctic regions, data
are needed on the rates of ice accretion and on the velocity, yield strength,
and mass of floating ice. Reliable deterministic and statistical methods are be-
ing developed to measure and interpret time-dependent field data suitable for
predicting structural loadings.
Structural Materials
An offshore structure should have a high ratio of strength to self-weight.
For instance, for each added unit of deck weight for a tension leg platform,
an additional 1.3 units of hull weight are required for buoyancy support, and
an additional 0.65 unit of mooring pretension force is needed. In this case,
high-strength steels with high fracture toughness are being investigated for the
purpose of reducing hull weight. Hollow cylindrical steellink chains or synthetic
mooring line materials, such as Kevlar with an abrasion-resistant polyethylene
cover, are being developed to increase mooring capacity. To determine the
suitability of new, high-strength steels and composite materials in the offshore
environment, test data are being generated. These data involve measures of cor-
rosion fatigue, fracture toughness, stress corrosion cracking, and weldability for
REFERENCES 15
REFERENCES
Adrezin, R., Bar-A vi, P. and Benaroya, H., Dynamic Response of Compliant Offshore
Structures-Review, Journal of Aerospace Engineering, October, 114-131, 1996.
Baar, J. J. M., Heyl, C. K., and Rodenbusch, G., Extreme Responses of Turret
Moored Tankers, Proceedings of the Offshore Technology Conference, OTC-
12147, 2000.
16 STRUCTURES IN THE OFFSHORE ENVIRONMENT
Gerwick, B. C., Construction of Marine and Offshore Structures, second ed., CRC
Press, Boca Raton, FL, 1999.
Jin, Z., and Bea, R., Enhancements of TOPCAT-3 Dimensional Loadings, Relia-
bility, and Deck Structure Capabilities, Proceedings of the Offshore Technology
Conference, OTC-11939, 2000.
Lecourt, E. J., and Williams, D. W., Deep Ocean Mining-New Applications for Oil
and Field Marine Equipment, Proceedings of the Offshore Technology Confer-
ence, OTC-1412, 1971.
Mulcahy, M., Fixed Offshore Structure Design Responds to Tougher Operating Re-
quirements, Sea Technology, 10-12, June 1979.
O'Brian, J. T., and Muga, B. J., Sea Tests on a Spread-Moored Landing Craft,
Proceedings of the Eighth Conference on Coastal Engineering, Lisbon, Portugal,
1964.
Will, S. A., Conventional and Deep Water Offshore Platforms, Civil Engineering 58,
February 1982.
Wilson, J. F., Muga, B. J., and Pandey, P., Dynamics During Deployment ofPipes for
Ocean Thermal Energy Conversion, Proceedings of the First Offshore Mechan-
ics/Arctic Engíneering/ Deepsea Systems Symposium, ASME, New Orleans, LA,
53-58, March 1982.
2
Structure-Environmental
Force Interactions
James F. Wilson
18 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
(2.1)
G m
Figure 2.1 Free body sketch for the spread moored ship of Figure 1.6.
Example Problem 2.1. Consider the surge motion for the spread moored
ship shown in Figure 1.6. This ship is asumed to be a rigid structure for which
the motion is described solely by its horizontal displacement coordinate v = v(t)
of its mass center at G. Sway, pitch, heave, and yaw motions are neglected. The
free body sketch of the ship in the vertical plane, shown in Figure 2.1, depicts the
four types of externally applied loads: self-weight or ship displacement W, which
is balanced by its buoyant force equal in magnitude to W; the equivalent of all v-
directed, time-dependent environmental forces p 1 (t); the equivalent v-directed
mooring line restraint force of the general form q( v ); and the velocity-dependent
SINGLE DEGREE OF FREEDOM STRUCTURES 19
damping force of the general form f(v). When equation (2.1) is applied in the
v direction to this free body sketch, the resulting equation for surge motion
becomes
In the first term of equation (2.2), the virtual mass m for surge motion is about
15 percent higher than the actual ship's mass. Quantitative values for the
remaining terms of this equation are discussed later in this chapter.
(2.3)
e
in which is the absolute angular acceleration of the rigid body and Jo is the
virtual mass moment of inertia of this body with respect to the reference axis
through point O and perpendicular to the plane of motion. The value of J0 is
defined as
(2.4)
where r is the distance from that reference axis to the virtual mass element
dm, and the integration is over the whole rigid body. In applications it is often
convenient to express Jo in terms of Jc, or the value of Jo when the point O
coincides with the mass center G. The connection is through the parallel axis
theorem, or
DECK
CAISSONS
JoB+ f(Ó) + q(O)- (moghc- mbghb) sinO= -F(t) ho- Mpc (2.6)
In equation (2.6), Jo is based on the virtual mass of the submerged portian of the
structure. For instance, for a submerged cylinder, virtual mass is approximately
the cylinder mass plus the mass of the water displaced by the cylinder. Quan-
titative calculations for Jo and the other terms of equation (2.6) are illustrated
in Example Problem 5. 2.
In the last example problem, the rigid body assumption may not always be
a realistic one. The rigid body model would be entirely useless, for instance,
if the dynamic flexura! stress were needed in the legs of the gravity platform.
FLUID-INDUCED STRUCTURAL FORCES 21
(2.7)
where p is the fluid density and D is the cylinder diameter. Shown in Figure
2.3 is this case of unseparated, unsteady, ideal flow, together with values of CM
for severa! ratios of cylínder length to diameter, e¡ D. These results, reported
by Wendel (1956), are based on theoretical values of another nondimensional
parameter, the added mass coeflicient CA, defined by
(2.8)
It is observed from the theoretical data in Figure 2.3 that as the cylínder length
becomes much larger than its diameter, the value of CM approaches the limit
of 2, for which CA approaches unity by equation (2.8).
22 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
.l/I¿~
1.2 1.62
2.5 1.78
5.0 1.90
9.0 1.96
00 2
Consider now the case of Figure 2.4, a rigid cylinder with a mass per unit
length mo, immersed in a fluid. This cylinder has an absolute translatíonal
displacement v v(t) in the fluid medium that would normally be at rest,
except for the presence of the cylinder. Based on Newton's second law of motion
and classical hydrodynamic theory, the force per unit length required to achieve
an acceleration v for the cylinder is
mo + CAp1r~
2
Equation (2.9) defines the cylínder's virtual mass per unit length, m. Thus, m
is the sum of mo in vacuo and the added or apparent mass per unit length,
C A(Ylf D 2 j4, resulting from those fluid particles that are pulled along by the
intruding cylinder.
(2.10)
The use of the absolute value sign on one of the velocity terms guarantees that
qv will always oppose the direction of u, as shown in Figure 2.5. For this
e
flow case, the experimental relationships of D to two nondimensional parame-
ters, cylinder roughness, and the Reynolds number, are well known (Schlichting,
1968). Here the Reynolds number is defined by
puD
Re (2.11)
f..L
where f..L is the absolute viscosity of the fluid. For a smooth cylinder subjected
to this constant, uniform, free stream ftow, the value of ev is approximately
unity for Re in the range of about 1000 to 200,000.
If the cylinder were rotating about its longitudinal axis or if it were not
circular, or if other solíd elements or rigid boundaries were nearby, one would
need an additionalloading parameter, the lift coefficient CL. In such cases, the
lift force per unit length, which is perpendicular to u and ijv, has the same form
24 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
S= !sD (2.12)
u
pattem times: t = 1!f;, 2!f;, ... n!J; pattem times: t = 1.5!f;, 2.5!f;, ... (n+0.5)1J;
(n = 1, 2, .. .) (n = 1, 2, .. .)
Generally, S correlates well with the Reynolds number. For instance, corre-
lations of S with Re showing the apparent effects on vortex-shedding frequency
of a cylinder in proximity toa twin cylinder anda ground planeare reported by
Wilson and Caldwell (1971). In sorne instances, vortex-induced pressure forces
on a cylindrical structure could be large enough to destroy the structure. For
instance, periodic vortices behind the cylindrical piles supporting an offshore
pier lead to a complete destruction of that pier in a tidal current of two knots.
Since vortex-induced loading is unpredictable, one generally makes provisions to
avoid periodicity of the vortices. Although many methods have been proposed
to suppress vortices (Hafen et al., 1976), a particularly practica! one is the ad-
dition of helical strakes around the cylinder, as shown in Figure 2.7. Optimal
strake geometries for the least vortex-induced loads on circular cylinders were
studied by Wilson and Tinsley (1989).
FLUID-INDUCED STRUCTURAL FORCES 25
uoT
Kc (2.13)
D
where u 0 is the amplitude ofthe wave velocity (Keulegan and Carpenter, 1958).
Conservation of Linear Momentum and Flow Superposition
For a stationary cylinder in a plane fiow field v.rith the free stream velocity
u = u(t), the total time-varying load per unit length on the cylinder may be
expressed by superimposing the two fiow models described by Figures 2.3 and
2.5. Thus, by adding their respective drag and inertialloadings expressed by
equations (2.10) and (2.7), the result becomes
(2.14)
Equation (2.14) was first proposed by Morison and his colleagues (1950) as an
empírica! result, and that equation now bears his name.
------:
r-n"""1
DISC
RIGID
cvu;-mER
8-I
fcoNTROL
J
¡,.,..__ q flZ
¡
VOLUME
There is a theoretical basis for Morison's equation, the general form for
which can be deduced by applying the principie of conservation of linear fluid
momentum. This conservation principie is based on Kewton's second law as
applied to the fluid occupying a fixed control volume V at any instant of time.
In this case, V is chosen as the imaginary rectangular box surrounding a disc
element of a submerged circular cylinder, as shown in Figure 2.8. Here, V =
D 2 ó.z, where D is the disc diameter and ó.z is the disc height. For water of
density p and with a horizontal velocity u along x, the net horizontal shear load
L,Fx on the disc is given by
(2.17)
large compared to T, then data show that CD and CM are functions of three
parameters, expressed as
The main advantage of Sarpkaya's forms is that u 0 , the free stream amplitude of
the periodic velocity, then appears only once in each function of equations (2.20),
instead of twice in each function of equations (2.18). for períodic flows, this
alternative gives effi.cient correlations of measured cylinder forces with CD and
e M. Among the phenomena neglected in both of these functional relationshlps
are cavitation, fluid compressibility, three-dimensional flow, proximity effects,
and all movement of the cylinder. In the following example problem, Morison's
equation is modified to account for one of these neglected factors: the lateral
vibrations of the cylinder.
_.,...V, V
ij
--év Tcv
u, ü FREE BODY
SKETCH
The free stream horizontal flow velocity u = u(t) is in line with the cylinder's
translational motion v = v(t). As shown on its free body sketch, the restraint
a:nd dampi:ng forces per unit length that oppose the cylinder motio:n are kv a:nd
cv, respectively. The virtual mass per u:nit le:ngth, m, is deduced from equation
(2.9). The fluid loading per unit length is given by equation (2.14), modified so
that the drag force is based on the relative velocity (u v) between the fluid
and the cylinder. When Kewton's second law, equation (2.1), is applied to this
cylinder, the equation of motion becomes
(2.21)
The term involving CM results from fluid motion only, such as wave action,
where ü is the absolute acceleratio:n of the fluid.
Equation (2.23) clearly shows that the damping of a moving cylinder is increased
due to the fluid drag force, a force that in general overwhelms the internal
structural damping c. The conclusion holds true for its nonlinear counterpart
also, equation (2.21).
Thus, to solve equation (2.21) or (2.23) for the structural displacement
v = v(t), one needs to know four structural parameters: m 0 , D, e, and k;
the fluid density p; the free stream flow field u= u(t); and the three empirical
constants e A' eD and e M. In subsequent examples, k and e will be estimated for
particular cases, and the dependency of the latter three empírica! constants on
offshore waves will be discussed in greater detail. Other e:nvironrnental factors
affecting the motion of offshore structures are now quantified.
Buoyancy and Gravity
It is well known that a salid object can be lifted much more easily when
in water than in air. This is because the water pressure exerts an upward or
buoyant force on the submerged solid. The Greek mathematician Archimedes
(287-212 B.C.) stated this principie in precise terms:
1
.~
G-$:
---
~
Figure 2.10 Cross section of a ship showing buoyant and gravitational forces.
For example, consider the ship shown in Figure 2.10 which has an actual
weight (in air at sea level) of mog. The volume of water that the ship displaces
is V and the weight density of the water is denoted by "( w. The result of applying
Archimedes principie to this system is
(2.24)
This buoyant force 'Yw V acts at poínt B on the ship, in a vertical direction
opposite to that of the resultant gravitational force mog. The latter force acts
at the ship's center of gravity. Thus, a ship's actual weight is generally referred
to as its displacement.
In general, point B is not coincident with point G of the solid body. In
precise terms:
This statement can be proved in general for solid body of arbitrary shape (Batch-
elor, 2000). Its validity can be simply illustrated for the following special case
of a solid, prismatic block partially submerged in water. Refer now to the se-
quence of illustrations (a) through (f) in Figure 2.11. Suppose that the solid
block (a) is removed from the water and then replaced by pressure forces around
the void of volume V so that the water remains undisturbed (b). The pressure
distribution, shown as "gage" pressure, is constant on any horizontal plane and
varíes linearly with depth. The pressure distribution on the solid block (e) is
identical to that on the void. In fact this identical pressure distribution would
also occur on a water block (d) of volume V whích would just fill the same void.
30 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
YwV
mog
J X. V
m,g m.g
(e) STATICALLY EQUIVALENT (t) STATICALLY EQUIVALENT
SOLIDBLOCK WATER BLOCK
WATER
r..
(a)
The three buoyant forces on this platform are: (1) a buoyant force due to
the mud of 'Ym Vm, which is located at a distance of hm from the base and acts
upward at the centroid of the displaced mud volume Vm; (2) a buoyant force
for the portion of the caisson in water of l'w Ve , which is located at he from the
base and acts upward at the centroid of its displaced water volume Ve; and (3)
a buoyant force on the submerged portian of each leg of l'w Ve, which is located
at he from the base and acts at the centroid of its displaced water volume of
Ve. If the structure has N identical legs, the total buoyant force mbg and its
statically equivalent location hb on the structure are, respectively
(2.25)
(2.26)
Figure 2.13 Static stability of three gravity platforms supported by a liquefied soil
foundation.
The exponent n depends on many factors. For ínstance, n = 3 fits data for
rough coastal areas; n 7 to 8 for sustained winds over an unobstructed sea;
and n 12 to 13 for gusts. At heíghts of 100 ft or more above the surfaces,
the vertical gust-velocity becomes about the same as its horizontal value. Wind
data ü(h) and further discussions of equatíon (2.27) are given by Muga and
Wilson (1970), Sherlock (1953), Simíu and Scanlan (1978), and Vellozzí and
Cohen (1968).
On a member of the superstructure which is nine or more diameters removed
from neíghboring structural elements, ü = ü(z) calculated from equation (2.27)
can be used with equation (2.10) to estimate wínd drag force, where Cv is based
both on Reynolds number and on the cross-sectional shape of the member.
Values of Cv for common shapes are readíly available (Hoerner, 1965, and
Pattison et al., 1977).
For a truss structure in wind, the sum of the drag forces on each individual
member may give a low estímate of the total drag force. This effect, called
34 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
design of floating and fixed structures. Offshore tidal currents, or the horizontal
water flow due to the vertical rise and fall of tides, often attain a maximum
velocity of 1 to 2 knots and may even reach 10 knots in sorne locations. Tidal
currents are higher in the spring than in any other time of the year. Data on
tidal currents for many offshore locations on the coasts of the United States and
Asia are published annually by the National Oceanic and Atmospheric Admin-
istration, U.S. Department of Commerce. Data for the tidal current Ut(z) as a
function of water depth z generally follows a power law similar to that for wind,
equation (2.27). That is
z)l/7
Ut(z) = ( 1 + d Ut(O) (2.30)
where d is the total water depth, z is the coordinate of the tidal velocity (a
negative number measured downward from the sea surface), Ut(O) is the tidal
velocity at the surface z =O, and Ut( -d) =O at the seafloor.
The other important current is Uw (z), the current generated by a sustained
wind blowing over the sea surface. The velocity profile of this wind-stress current
is approximated as linear with depth, with a maximum value uw(O) at the sea
surface, where uw( -d) =O at the seafloor. That is
z)l/7 Ut(O)
U= ( 1 + d + ( 1 + dz) Uw(O) + Uwave (2.32)
With the design values for Ut(O), uw(O), and Uwave, and with a knowledge ofthe
geometry for all the structure's tubular members whose longitudinal axes are
perpendicular to u, the total horizontal drag force from those tubular members
of the structure is obtained by using equation (2.32) with equation (2.10) and
integrating the result over those tubular members.
Currents do affect structural integrity in other ways. Current-induced scour-
ing, for instance, can undermine pile-supported jacket template platforms and
gravity platforms by erroding surrounding sand and soil. Currents carry ice that
can impact and damage structures. High currents accelerate the corrosion rate
of submerged metal structures. Currents also modify waves and wave loading
of structures (Thng, 1974). Except for the impact of ice, which is discussed
briefly in the next section, these environmental hazards to offshore structures
are subjects that are beyond the scope of this book.
36 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
0.4g .-----..-----....-----.....-----,----..,-----..
0.2g
0.2g
0.4g '-----'------''------'------'-----'----~
O lO 20 30
TIME, t seconds
Figure 2.14 Accelerogram for the El Centro earthquake, south east component,
May 18, 1940.
EARTHQUAKES, ICE IMPACT, AND WAVE SLAMMING 37
tEa,:,7JJ
STRUCTURAL MODEL
~----,
1 m 1
,, ., ,,~ ~---...J
C¡V
'k
FREEBODY
STRUCTURE SKETCH
Figure 2.15 Offshore structural model with horizontal sea fioor motion (an
earthquake).
Vt =V+ Vg (2.33)
(2.34)
The restoring forces due to structural stiffness and damping depend only on
the relative displacement v and the relative velocity v. Thus, when Newton's
second law of motion is applied to the spring-mass model, the equation of motion
becomes
(2.35)
38 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
1 2
Fs = 2C 8 pDfu (2.38)
Here, the coefficient Cs is in the range of 3.5 to 3.6; D and f are the cylinder
diameter and length, and p and u are the water mass density and the peak
horizontal water particle velocity, respectively. If the cylinder is not rigid but
a flexible, elastic body (a tubular brace of a jacket-template platform, for in-
stance), then Sarpkaya and Isaacson (1981) recommend the following procedure
for computing the cylinder load: let Cs = 3.2 and then multiply the resulting
force calculated from equation (2.38) by the force-impact magnification factor
40 STRUCTURE-ENVIRONI\:IENTAL FORCE INTERACTIONS
P1(t) = ü (2.39)
The dominant mode of motion 7/J(x) is shown by the broken lines of Figure
2.16b. The midspan coordinate is v v(t), which locates the lumped, virtual
mass m of the massless cross brace of bending stiffness El, as shown in Figure
2.16c. The virtual mass is deduced from equation (2.23), or
mo+CAprr~ )
2
m ( f1e (2.40)
in which mo is the actual mass per unit length. Note that h = 1 gives an upper
bound for m, but this is a bad choice since it is obvious that all of the mass
along the length does not have the same displacement as that of the lumped
mass. In Chapter 5, Example Problem 5. 3, it is shown that fi 0.370 for this
STRUCTURAL MASS, DAMPING, AND RESTRAINT 41
structural element. The bending stiffness of the brace is derived from elementary
beam theory and is
El
kl = 192-
p3 (2.41)
Here, k1 can be interpreted as the lateral force that, when applied to the
midspan, will produce a static defiection of v = 1 at midspan. The structure-
fiuid damping is assumed to be linear-viscous which, referring to equation (2.23),
has the following form:
(2.42)
-----i<'-X
_,, l /D
~y·
'1+-----t-----&.' ~
(a) TUBULAR CROSS BRACE CROSS SECTION
(SIDE VIEW)
m
l/f(x)
t ..,"' " _v ] - ...
',
~ "'-: ; > ',_~, ax
' .......... __ ..... "" ,. ....
(b) DEFLECTION ENVELOPE (e) LUMPED MASS MODEL
(TOPVIEW) (TOPVIEW)
forces on m are shov.rn on the free body sketch in Figure 2.16e, from which the
equation of motion is deduced as
(2.43)
Here, since p 1 ( t) is alllumped at midspan instead of being uniformly dístríbuted,
the solution v v(t) to equation (2.43) will be on the high side. This particular
mathematical model is thus a conservative one. )Jeeded for p 1 (t) are the explicit
forms for the flow field u u(t) and its associated constants CD and CM, topics
that are deferred to Chapters 3 and 4.
P,(t} m k,v
~
(e) FREE BODY
SKETCH
CURRENT
LOAD
3El
1+------210 - - - - ' - - 1
(a) TYPICAL STRUCTI:RE (b) MATHEMATICAL
(DIMENSIONS: ft) MODEL
small by comparison to v. Now lump a fraction f¡ of the virtual mass of all three
flexible legs with the rigid deck mass, ma. From the coefficient of ii in equation
(2.23), the equivalent virtual mass for this structural system is deduced as
(2.44)
in which iño is the actual mass per unit length of a single leg, d is the water
depth, and (e - d) is the length of a leg between the still water line and the
bottom of the deck. (Example Problem 5.4 will show that f¡ = 0.375.) Assume
that structural damping is mainly produced by the submerged portions of the
legs. From the coefficient of v ín equation (2.23), the damping coefficient is
deduced as
(2.45)
which is the magnitude of the horizontal force at deck level that produces a unit
deflection (v = 1) at that point. (For a single leg, k 1 l2EI / C3 ). The value
of k1 given by equation (2.46) is an upper bound value for two reasons. First,
this stiffness is decreased as the full fudty conditions on the legs are relaxed
at the mud line or at the deck. Second, this leg stiffness is also reduced as the
magnitude of the deck load approaches the Euler buckling load for this structure,
an effect that will be explored ín a subsequent example problem.
In summary, the jackup drilling rig of Figure 2.17a and Figure 2.17b is
modeled as a single degree of freedom system whose free body sketch is shown
in Figure 2.17c. When Newton's second law is applied to the equivalent virtual
mass in this latter sketch, the governing equation (2.43) is obtained. With the
respective values of m, c1 , and k 1 given by equations (2.44), (2.45), and (2.46),
the explicit form of equation (2.43) becomes
(2.47)
Cable Restraints
Cables or guy lines are employed to restrain the motion for severa! types of
offshore structures. The simplest cable configuration is the vertical one used
to restrain floating, tension leg platforms such as in Figure l.le. In addition,
44 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
there are two classes of flexible mooring systems in which stationary guy lines
hang as catenary curves from the structures to the sea floor. The first is the
single line that constrains a vessel or buoy. The second is the multi-line system
that constrains vessels, semisubmersible platfonns, and compliant towers. The
lines in present use are ropes of metallic wire or of synthetic fiber such as
nylon, Dacron, or Kevlar; and steel chains with solid or hollow links. Practica!
aspects of guy line designare discussed in the U.S. Navy publication NAVFAC
DM-26 (1968), in the four papers by Childers (1973-1975), and in the work of
Niedzwecki and Casarella (1975). These references include analyses of multi-
line systems with inextensible or nonstretching lines. Refined analyses which
include clumped weights and additional anchors along the cables, as well as the
effects of elastic cable stretching, are presented by Adrezin et al. (1996), Ansari
(1980), and Wilson and Orgill (1984).
For a taut cable with negligible sag, the longitudinal extension /5 depends
on both the applied longitudinal force Fe and the material properties of the
cable. For instance, Wilson (1959) used the following power law to correlate the
load-extension behavior of both steel wire and synthetic fiber line employed in
mooring ships:
(2.48)
Here, Co and n are constants depending on the material, its length, and its
cross-sectional area. If the deflections are sufficiently small, n 1 and the
force-deflection relationship based on elementary theory is given by
(2.49)
(2.51)
where Fx is the horizontal component of the tension force. Since the cable's
bending stiffness is neglected, the resultant end tensions Fo and F are in a
direction tangent to the catenary curve. For static equilibrium, then, the hori-
zontal component of tension remains unchanged, or
(2.54)
A closed form solution to equation (2.51), or the (x, y) coordinates of the cate-
nary curve, is
x = Fx .
w smh- 1 (wfF"'x + tan0 )
0 (2.55a)
z = Fx [wx .
w cosh Fx + smh- 1 (tan0 0 ) ] Fx cosh [sinh- 1 (tan0o)] (2.55b)
w
Equations (2.55) can be used to describe the static shapes of guy lines formed
of multiple uniform segments placed end-to-end in which the segments have dif-
ferent geometric and material properties. In such configurations, compatibility
of both slope and tension force at the junction of each adjacent segment needs
to be maintained (Ansari, 1980). The following example problem illustrates
the use of the catenary solutions to solve for the restraining stiffness q( v) of a
fioating platform stayed with identical guy lines.
SIDE VIEW
TOPVIEW
REFERENCE __~~-~~~--~0--
ANCHOR
POSITION
DISPLACED
T
POSITION Zo = const.
--~----.--.~~.-~-------7~~~~~-------ns~~IM«
Consider the stiffness of the single cable defined in Figure 2.20. For the static
equilibrium state (v = O), the origin of the cable coordinates is at O, for which
the flat length, the suspended length, and the horizontal projected lengths are
Lr, lnand Xr, respectively; and the tension force is Fr at angle Or at the top
suspension point. At the bottom of the cable, the slope is zero at both Ür and
at the shifted origin O for v =f. O, at which points the condition F0 Fx is
always true. The vertical projected length remains constant, or z = z 0 . Since
the cable is assumed to be inextensible, the reference lengths subscripted r can
be expressed in terms of their corresponding unsubscripted values for v =f. O as
constant (2.56)
Also, the distance between the anchor and the origin of the platform displace-
ment coordinate v remains constant, or
Lr + Xr = L +x V (2.57)
vVhen (L Lr) is eliminated between equations {2.56) and (2.57), the platform
displacement is
(2.58)
48 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
Two additional equations, deduced from equations (2.55) for Bo Oand z = zo,
are respectively
Fx
zo= w cosh
(wx) Fx
F:x - w (2.60)
From the overall equilibrium conditions, equations (2.51) and (2.52), it follows
that F:x = F cos () and wt F sin e. The ratio of these latter equations leads to
For a fixed value of Zo and w andan initially fixed value of e Br, then the three
initial values Fx = Fxr, X = Xr, and e = e,. can be calculated from equations
(2.59), (2.60), and (2.61). For values of () ::/=- Bn the corresponding values of
Fx, x, R., and v are calculated from equations (2.58)-(2.61). The restoring force
of a single cable is simply
One can now deduce that the restoring force for the pair of identical cables (the
single cable of Figure 2.20 and its mirror image across the plane v O) is given
by the superposition of the results just derived for the single cable, or
-e--·
--
e, Cr
er
(2.64b)
zo zo
X x,. V
x Xr v= (2.64c)
zo zo zo
With equations (2.64), equations (2.58)-(2.61) become, after sorne rearrange-
ment,
(2.65)
STRUCTURAL NfASS, DAMPING, AND RESTRAINT 49
(2.66)
f F, tanB (2.67)
(2.68)
For any fixed value of B (including (q, the values on the left sides of equations
(2.65)-(2.68) can be calculated explicitly, in sequence.
!Or---~----~----~----~----r---~
(jr
5
1
1
'k. /
r..á
u
~
o ;!.!,___
.....
....J o ---,
/
- 60°
-<
¡.... ,- /
z
o /
/
N /
C2 /
o /
::r: /
1
1
-5 1
1
1
--SINGLE CABLE
1
1 --- -TWO OPPOSING
1 CABLES
-1 o "------l'------L----'-----1---.....1..--~
-0.15 -0.1 -0.05 o 0.05 0.1 0.15
HORIZONTAL DISPLACEMENT, v
Figure 2.21 Static behavior for a single and two opposing cables of a cable stayed
fioating platform.
50 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
Numerical results for Br 45 and 60 deg are shown in Figure 2.21. For
the single cable, the nondimensional horizontal restoring force Fx is small for
negative nondimensional displacements v and increases at a growing rate as
v becomes positive. For the broken-line curves which correspond to opposing
pairs of cables 180 deg apart, F'x is antisymmetric about v = O and behaves
as a "hardening" spring for both positive and negative displacements v. For
these opposing cables, if the actual displacement v is less than about 5 percent
of the attachment height zo, the horizontal force-displacement relationship is
linear, for practica! purposes. In general for such diametrically opposed cable
systems, the horizontal restoring force q( v) can be approximated by an odd
order polynomial in v, or
(2.69)
Calculations of q (v) for other Br valu es and for muliple pairs of identical, sym-
metric arrays of cables, with their ultimate approximation in the form of equa-
tion (2.69), are left for the reader. In making such calculations, it is recalled
that this analysís is valid only for cables that are relatively inextensible. Cal-
culations by Irvine (1981) imply that for steel cables over 13 cm in diameter,
where 45 ~ Br ~ 85 deg, the cable extension strain is less than 0.1 percent. In
such cases the assumption of cable inextensibility is a reasonable one and the
results obtained herein are valid.
80
1
::9 1
-:E:
M
o /
/
40 /
/
~
/
¡.¡.f
u
~
o
~ o
o ....
.... ....
~
ot-< /
/
/
/
rJ':J
-40 /
ga 1
/
1 SURGE
1 --- SWAY
-80
-4 -2 o 2 4
SHIP DISPLACEMENT, v ft
Figure 2.22 Static behavior of a moored LST (O'Brien and Muga, 1964).
STRUCTURAL MASS, DAMPING, AND RESTRAINT 51
Example Problem 2.1 O. Consider the explicit forms for the cable restraint
functions q( v) of the spread moored ship depicted in Figure 1.6 and previously
introduced in Example Problem 2.1. As in the original experimental study and
analysis by O'Brien and Muga (1964), traditional English units are used to
express the numerical results. The subject ship was an LST (Landing Ship,
Tank) moored in 45ft of water in the Gulf of Mexico. The calculations for the
separate restraint functions q( v) for surge and sway were based on the catenary
theory just discussed and the actual mooring geometry of all seven chains for this
experimental study. The results are shown in Figure 2.22. These results typify a
hard spring nonlinear restraint system and are similar to the two-cable example
shmvn in Figure 2.21. When each curve of Figure 2.22 is fit to the odd order
cubic polynomial of equation (2.69), the restraining force for surge (longitudinal)
displacement and for sway (lateral) displacement become, respectively
Example Problem 2.11. The purposes of this example are to compute the
virtual masses for the moored ship described in Example Problem 2.10, and to
set up the uncoupled equations of motion in surge and sway. Traditional English
units are employed for the purpose of clarifying the unit of mass. As previously,
the ship is assumed to be a rigid body with an actual mass mo. Ba.sed on its
given displacement (weight) of 4400 long tons, the ship's actual ma.<>s is
1 lb
mo = 32 .1 ft / sec2 x 4400 long tons x 2240 3.06 x 105 slug
long ton
in which the mass unit of lb-sec2 /ft is designated as slug. Experimental evidence
shows that for surge motion only, the virtual mass is approximately 15 percent
greater than m 0 , or m 1.15mo = 3.52 x 105 slug; and for sway motion only,
the virtual mass for this ship (unstreamlined for sway) is about twice m 0 , or
m = 2mo = 6.12 x 105 slug. For either motion, the governing equations are
of the form of equation (2.2). Assume negligible damping, f(v) O, and calm
seas, p(t) =O. Gnder these conditions, with q(v) given by equations (2.70) and
(2.71), and with the virtual masses just calculated, the respective equations of
motion for surge and sway are as follows:
In these equations of motion, iJ has units of ftjsec 2 and v has units of ft.
52 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
(2.74)
in which
(2.75)
For submerged sandy soil, the horizontal subgrade reaction constant No in the
last equation has a range of 4 tons/ft 3 to 34 tons/ft 3 for relatively loose to
dense sand, respectively (Terzaghi, 1955). With the value of fe from equations
(2.74) and (2.75), k 1 for the pile can calculated from classical beam theory for
a cantilevered beam oflength fe as: k1 3EI/f~. Kocsis (1976) also presented
STRUCTURAL MASS, DAMPING, AND RESTRAINT 53
equations similar to (2.74) and (2.75) for the calculation of €e for apile in clayey
soil foundations.
Consider now the measured effects soil behavior to the motion of a contacting
structure. Veletsos and Wei (1971) performed extensive laboratory experiments
on soils in contact with a disc of radius r 0 , as depicted in Figure 2.24a. In a
typical experiment, this disc was subjected to a harmonic frequency w, first in
the direction of sliding, v, and then in pure rotation e. Nataraja and Kirk (1977)
correlated these data for applications to gravity platform dynamics. For disc
sliding motion only, the respective constants for soil stiffnesss and damping are
k1 and c1 ; and for rotational motion only, these respective constants are ke and
co. These constants, shown in the disc model of Figure 2.24b, are as follows:
(2.76)
80
VCJ: rg =
8
ke = s (1 - 0.215wro [i¡Gp ) ao - bow (2.78)
3(1-v) 8
(2.79)
Here 0 8 , v, and Ps are the shear modulus, Poisson's ratio, and mass density,
respectively, of the soil. The parameter w (radjsec) is the frequency of the disc.
The use of this dynamic soil model is illustrated in Example Problem 5.2.
The validity of employing equations (2. 76)-(2. 79) for full-scale design of grav-
ity platforms still needs to be shown through full-scale testing. Efforts to employ
the generalizations of continuum mechanics to characterize the dynamic, me-
chanical properties of soil, including the formulation of constitutive equations
from carefully designed experiments, are discussed by Zienkiewicz et al. (1978,
Chapters 10-16) and Prevost et al.(1981).
54 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
PROBLEMS
Yw
--\
~
2.3 The cylindrical tank of Figure 2.25 has diameter D, height h, andan
average weight density in air of "Yt when filled with salid, radioactive wastes.
The tank penetrates to a depth d into the mud of weight density "Ym at the
bottom of the sea. The mud layer behaves as a liquid.
(a) Derive an expression for the tank dimension ratio djh in terms of the
system densities. If "Yw = 64 lbjft 3 , "Ym = 95 lbjft 3 , and "Yt = 85 lb/ft 3 , what
is this ratio?
(b) Locate the center of buoyancy for the cylinder in its upright position,
where D = 8 ft and h = 10 ft. Discuss the stability of this cylinder for small
angles of tilt, e.
(e) For D= 8 ft and the densities given in part (a), calculate the maximum
height h for which the cylinder will remain stable for small e.
2.4 The concrete monotower pictured in Figure 2.26 is in 160 m of water.
The caisson is filled with oil, and the leg is filled with seawater up to the still
waterline. The pertinent data are:
PROBLEMS 55
!-l-
¡
1
1
1
... 1 1.¿::
160 o.s- :- 1
1
j DIA
1
1
G
t 60
ha 1
~ lo
r
l
~80DIA ~ DIMENSIONS: meters
2.5 The mean wind velocity is used to calculate the wind-induced load on
an offshore structure. Discuss briefly a method for converting experimental
measures of the fastest mile of wind speed to a mean wind velocity. State your
assumptions and cite pertinent references.
2.6 Using equation (2.27), plot the normalized horizontal wind velocity
u( z) fu( h) as a function of the height parameter z /h. Show three typical curves:
one for a rough coastal area, one for sustained wind over an unobstructed sea,
and one for a gusty wind. Summarize your observations about these particular
wind velocity profiles.
2. 7 The addition of helical strakes to a cylindrical structure sueh as shown
in Figure 2.7 is not the only way to negate trailing periodic vortices. Describe
other practica!, alternative devices that may be added to the structure to elim-
inate such vortices. Use the references cited in the text as a starting point.
2.8 On the same graph, plot two flow velocity profiles as a function of water
depth: a tidal current and a wind-stress current. For this graph, assume that
56 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
both flows are in the same direction. If the wind velocity is offset by an angle
e to the tidal current, what is the equation for the total velocity in terms of e.
2.9 A relatively isolated, vertical cylindrical pile is fully submerged and
subjected to a steady tidal current of magnitude 4 knots. The pile has diameter
D = 0.3 m and height h 5 m. Compute the total current-induced drag force
on this pile. State all assumptions you make in arriving at your answer.
2.10 Based on studies reported in the most recent literature (for instance,
The Proceedíngs of the Offshore Technology Conference ), discuss briefl.y both
the advantages and dangers of using onshore seismic data for ground motion in
the design of fixed-bottom offshore structures for earthquake resistance.
2.11 The coefficient Ci of equation (2.37), which is used to predict the
horizontal impact of ice on structures, varies by more than a factor of two.
Investigate the physical reasons for this variability.
2.12 After consulting the recent literature on the subject, write a summary
report comparing the particular design features that allow the offshore platforms
near Alaska to resist ice hazards. How do these platforms differ from the jacket-
template structures in the Gulf of Mexico?
2.13 For the case of two opposing cables shown in Figure 2.19, where the
equilibrium angle Br is 45 deg, calculate k 1 and k3 for a best fit to equation
(2.69). If w = 39 lb/ft and zo 1000 ft, what is q( v )?
2.14 Assume that the fl.oating structure of Figure 2.19 is supported by four
identical, symmetrically arranged cables, 90 deg apart. Assume that the only
motion is horizontal and along the Une of two opposing cables. Show that the
restoring force as a function of defl.ection as given by the curves of Figure 2.21 is
still quite accurate because the cables perpendicular to the direction of motion
contribute very little resistance to this motion. For what range of initial angles
Br is this approximation not a good one?
2.15 Assume that N pairs of identical cables are arranged at equal angles
(3, as shown in Figure 2.19. Derive a general expression for the restoring force
along the line of one of these pairs. State all your assumptions. Then plot the
antisymmetric curve in the nondimensional form of Figure 2.21 for N = 6, Br
= 45 deg, and (3 = 30 deg.
2.16 Based on the analysis of Kocsis (1976), calculate fe, the equivalent
length of a solid concrete pile submerged in medium stiff sand. Use the following
data: e 30 ft, D = 4 ft, E = 3000 kips/in2 , and No = 14 tons/ft 3 • (Hint:
First convert to a consistent set of units.)
REFERENCES
Adrezin, R., Bar-A vi, P., and Benaroya, H., Dynamic Response of Compliant Offshore
Sructures~Review, Journal of Aerospace Engineeríng, 1996.
REFERENCES 57
Bazant, Z. P., and Kim, J. J. H., Size Effect in Penetration of Sea Ice Plate with
Part-Through Crack.s. I: Theory, II: Results, ASCE Journal of Engineering
Mechanics 124 (12), December 1998.
Bea, R. G., Audibert, J. M. E., and Akky, M. R., Earthquake Response of Offshore
Platforms, Jo1trnal of the Structural Division, ASCE 105 (ST2), 1979.
Bea, R. G., Xu, T., Stear, J., and Ramos, R., Wave Forces on Decks of Offshore
Platforms, ASCE Journal of Waterway, Port, Coastal, and Ocean Engíneering
125 (3), 1999.
Bercha, F. G., and Stenning, D. G., Arctic Offshore Deepwater Ice-Structure Interac-
tions, Proceedings of the Offshore Technology Conference, OTC-3632, Houston,
TX, 1979.
Berge, B., and Penzien, J., Three-Dimensional Stochastic Response of Offshore Tow-
ers to Wave Forces, Proceedings of the Offshore Technology Conference, OTC-
2050, Houston, TX, 1974.
Blevins, R. D., Flow-Induced Vibrations, Van Nostrand Reinhold, New York, 1977.
Childers, M. A., Mooring Systems for Hostile Waters, Petroleum Engíneer, 1973.
Childers, M. A., Deep Water Mooring-Part II, The Ultradeep Water Spread Moor-
ing System, Petroleum Engíneer 2, 1974.
Childers, M. A., Deep Water Mooring-Part III, Equipment for Handling the Ultra-
deep Water Spread Mooring System, Petmleum Engineer 5, 1975.
DeFranco, S. J., and Dempsey, J. P., Crack Propagation and Fracture Resistance in
Saline Ice, Journal of Glaciology 40, 1994.
Gaythwaite, J., The Marine Environment and Structural Design, Van Nostrand Rein-
hold, New York, 1981.
58 STRUCTURE-ENVIRONMENTAL FORCE INTERACTIONS
Gould, P. L., and Abu-Sitta, S. H., Dynamic Response of Structures to Wind and
Earthquake Loadíng, Wiley, New York, 1980.
Hafen, B.E., Meggitt, D. J., and Liu, F. C., Strumming Suppression-An Annotated
Bibliography, N-1456, U.S. Civil Engineering Laboratory, Port Hueneme, CA,
1976.
Hoerner, S. F., Fluid Dynamic Drag; Published by author, New York, 1965.
lrvine, H. M., Cable Structures, The MIT Press, Cambridge, MA, 1981.
Keulegan, G. H., and Carpenter, L. H., Forces on Cylinders and Plates in an Oscil-
lating Fluid, Journal of Research of the National Bureau of Standards 60 (5),
1958.
Kocsis, P., The Equivalent Length of aPile or Caisson in Soil, Civil Engineering 63,
1976.
Marshall, P. W., Fixed Pile-Supported Steel Offshore Platforrns, Journal of the Struc-
tural Division, ASCE 107 (ST6), 1981.
Moríson, J. R., O'Bríen, M. P., Johnson, J. vV., and Schaaf, S. A., The Forces Exerted
by Surface Waves on Files, Petroleum Transactions, AIME 189, 1950.
Muga, B. J. and Wilson, J. F., Dynamic Analysís of Ocean Structures, Plenum, New
York, 1970.
Munson, B. R., Young, D. F., and Okushi, T. H., Fundamentals of Fluid Mechanics,
Wiley, New York, 1998.
Nataraja, R. and Kirk, C. L., Dynamic Response of a Gravity Platform under Ran-
dom Wave Forces, OTC-2904, Proceedings of the Ojjshore Technology Confer-
ence, 1977.
REFERENCES 59
NAVFAC DM-26, Design Manual-Harbar and Coastal Facilities, U.S. Navy Publica-
tions and Forms Center, Philadelphia, PA 19120, 1968.
Newman, J. N., Marine Hydrodynamics, The MIT Press, Cambridge, MA, 1977.
Neumann, G., and Pierson, W. J., Jr., Prínciples of Physical Oceanography, Prentice-
Hall, Englewood Cliffs, NJ, 1966.
Niedzwecki, J. M., and Casarella, M. J., On the Design of Mooring Lines for Deep
Water Applications, ASME paper No. 75-WA/OCE-1, 1975.
O'Brien, J. T., and Muga, B. J., Sea Tests on a Spread-Moored Landing Craft,
Proceedings of the Eighth Conference on Coastal Engineering, Lisbon, Portugal,
1964.
Pagon, W. W., Drag Coefficients for Structures Studied in Wind Tunnel Model Stud-
Engineeríng News-Record 113 (15), McGraw-Hill, New York, 1934.
Pattison, J. H., Rispin, P. P., and Tsai, N., Handbook of Hydrodynamic Characterís-
tics of Moored Array Components, NSRDC Report No. SPD-745-01, 1977.
Peyton, H. R., Ice and Marine Structures, Ocean Industry, Houston, TX, 1968.
Prevost, J. H., Cuny, B., Hughes, T. J. R. and Scott, R. F., Offshore Gravity Struc-
tures: Analysis, Journal of the Geotechnical Engineeríng Dívision, ASCE 107
(GT2), 1981.
Sarpkaya, T., Vortex Shedding and Resistance in Harmonic Flow About Smooth and
Rough Cylinders at High Reynolds Numbers, NPS-59SL76021, Naval Postgrad-
uate School, Monterey, CA, 1976.
Sarpkaya, T., and Isaacson, M., Mechanics of Wave Forces on Offshore Structures,
Van Nostrand Reinhold, New York, 1981.
Schlichting, H., Boundary-Layer Theory, sixth ed., McGraw-Hill, New York, 1968.
Sherlock, R. H., Variations of Wind Velocity and Gusts with Height, Transactions
ASCE 118, 1953.
Simiu, Equivalent Static Wind Loads for Tall Building Design, Proceedings of the
Structural Division, ASCE 102 (ST4), 1976.
Simiu, E., and Scanlan, R. H., Wind Effects on Structures, Wiley, New York, 1978.
Veletsos, A. S. and Wei, Y. T., Lateral and Rocking Vibration of Footings, Journal
of the Soil Mechanics and Foundations Division, ASCE 95 (SM9), 1971.
Vellozzi, J., and Cohen, E., Gust Response Factors, Journal of the Structural Dívi-
sion, ASCE 94 (ST6), 1968.
Wiggins, J. H., Hasselman, T. K., and Chrostowski, J. D., Seismic Risk Analysis for
Offshore Platfonns in the Gulf of Alaska, OTC-2669, Proceedings of the Offshore
Technology Conference, Houston, TX, 1976.
Wilson, B. W., Case of Critica! Surging of a Moored Ship, Paper No. 2318, ASCE
Proceedings, 1959.
Wilson, J. F., and Orgill, G., Optimal Cable Configurations for Passive Dynamic
Control of Compliant Towers, ASME Journal of Dynamic Systems, Measure-
ment and Control 106, 1984.
Wilson, J. F., and Tinsley, J. C., Vortex Load Reduction Experiments in Optimal He-
lical Strake Geometry for Rigid Cylinders, ASME Journal of Energy Resources
Technology 111, 1989.
Zienkiewicz, O. C., Lewis, R. W., and Stagg, K. G., eds., Numerícal Methods in
Offshore Engineeríng, Wiley-Interscience, New York, 1978.
3
Deterministic Descriptions
of Offshore Waves
Bruce J. Muga
62 DETERMINISTIC DESCRJPTIONS OF OFFSHORE WAVES
DETERMINISTIC PROBABILISTIC
T
1 1
1
ANALYTIC NUMERIC
1
1 1
LINEAR NONLINEAR WAVE
THEORIES THEORIES SPECTRA
Before discussing the most important classical wave theories, we first introduce
sorne fundamental definitions and concepts. The natural occurrence of waves
is usually complicated by the simultaneous superposition of waves of many dif-
ferent shapes and energies. Thus we define a single wave or a train of waves of
pure forro that always behaves in exactly the same way on passing any given
point, and that propagates without change of shape when the surrounding con-
ditions remain the same. Further, we consider only plane or long-crested waves
of permanent forro, waves that are independent of the coordinate normal to the
wave propagation direction. Therefore their crests or troughs can be represented
by straight lines in a horizontal plane. By permanent forro, we mean that the
field of motion, pressure distribution, and surface configuration are maintained
a..'l one follows the wave at a speed e, the designation of the phase velocity or
celerity. (In a strict sense all real waves have finite length crests, a factor that
is ignored herein.) Consider a wave forro that is simple harmonic so that at
any time t the wave has a sinusoidal shape with reference to the still water line
(SWL), or the x-axis as shown in Figure 3.2. At t =O, the instantaneous surface
elevation r¡ is then
Z,1J
In equation (3.1), A defines the amplitude of the wave and k is the wave
number. In radian measure, kx = 21rx j >.; or x j >. is the multiple of the angle 21r,
where >. is the particular value of x for which cos kx = 1, yielding the next crest
of the wave in advance of the first at x O. The wave length >. is expressed in
terms of wave number by
(3.2)
(3.5)
where T is the particular value of t that makes the cosine term unity, giving the
next crest of the wave in succession to the first at t = O. The quantity T is the
period of the wave and
21r
kc = T =w rad/sec (3.6)
where w is the angular frequency. From equations (3.2) and (3.5) it follows that
.X=cT (3.7)
Equation (3.7), which could have been inferred directly, is a fundamental rela-
tionshlp in wave theory and has general application regardless of wave form.
Equation (3.6) enables us to rewrite equation (3.3) as
rJ(x, t) A cos(kx- wt) (3.8)
If we now compare two similar waves of identical form that pass the same place
at different times, we have for one at t = O
rJ~=O = A cos kx (3.9a)
and for the other
·7J~=O = A cos( kx + e) (3.9b)
where e is the phase displacement, as illustrated in Figure 3.4. This is positive
if ·7J" lags the first wave, but negative if rJ 11 leads rJ'. The interpretation of sign
is the same as for wt in rJ = Acos(kx ± wt): negative for a forward wave, and
positive for a rearward wave, where forward means motion in the direction of x
positive.
1]
Figure 3.4 Illustration of phase displacement for two waves of identical form.
r¡
The foregoing discussion includes the fundamental concepts and most of the
definitions needed to describe classical wave theories. It is noted that the symbol
H, often used for wave height, has yet to be defined. The reason for this will
soon become apparent.
The theory of linear waves is alternatively known as Airy's theory, small am-
plitude theory, and first-order theory. Developed primarily by Airy (1845) and
by Laplace (1816), it is the most important of the classical theories because it
is both easy use and it forms the basis for the probabilistic spectral descrip-
tion of waves. For a full development and discussion of the theory, the reader
should consult the intellectually entertaining work by Kinsman (1965). We now
summarize the assumptions, the governing equations, and the solutions for the
wave velocity and pressure profiles useful for predicting wave-induced forces on
offshore structures.
Suppose that the simple harmonic plane wave defined in Figure 3.6 is prop-
agating in the positive x direction in water of density p. The vertical coordinate
is z, directed positive upward; and the origin is located at the still water line
(SWL) or the mean surface level. The nine assumptions inherent in línear theory
are as follows:
l. The amplitude A of the surface disturbance is very small relative to the
wave length .X and the water depth d.
2. The velocity head (u 2 + w 2 )j2g is small compared with the hydrostatic
pressure head pgz. Here u and w are the horizontal and vertical water particle
velocities, respectively.
66 DETERJ\iJINISTIC DESCRIPTIONS OF OFFSHORE WAVES
Figure 3.6 Schematic representation of a simple harmonic wave. The vertical scale
of the surface profile is exaggerated for clarity.
When the nine assumptions are invoked, the differential equations and bound-
ary restraints for the water particle velocity and pressure reduce to the following
forros:
{)u aw
=0 (3.lla)
8z
{)u+Bw=O (3.llb)
8x 8z
Bu 1 8p
(3.llc)
8t p ax
{)w {)p
g (3.1ld)
8t 8z
LINEAR PLANE WAVES 67
{)r¡
w = &t at z o (3.12)
w =O at z = -d (3.13)
P = Pa at z O (3.14)
Ptotal Pa + Ps + P (3.15)
w 2 = gk tanh kd (3.16)
From equation (3.6), the wave phase velocity or celerity is given by e wjk.
With this and equation (3.16), the celerity becomes
(3.17)
This equation for celerity reduces to rather simple limiting forms for cases in
which k is either very large (short wave lengths) or very srnall (long wave
lengths). Such examples are included as problems at the end of this chapter.
68 DETERMINISTIC DESCRIPTIONS OF OFFSHORE WAVES
Consider a few comments about the results for linear wave theory that are
summarized in Table 3.1. First, note that the water particle velocities, the
wave celerity, and the wave group velocity (derived by Sarpkaya and Isaacson,
1981) are all different in form and have different physical meanings. The water
partide velocities and particle accelerations are those used in Morison's equation
to compute the drag and inertial forces of these waves on offshore structures.
Second, the origin x O of the wave is arbitrary, which implies that a constant,
arbitrary phase angle can be added to the term ( kx wt) in the formulas of
Table 3.1. To the casual reader, this may be a source of confusion since many
references locate the origin at the trough of the wave rather than at its crest.
Third, sorne classical references define the coordinate z as positive downward,
which has the effect of reversing signs for those terms containing z. Further,
the vertical coordinate is sometimes labeled y in place of the more common z.
Trochoidal Theory
Gerstner (1802) and Rankine (1863) developed the trochoidal wave theory
independently. The three distinguishíng features of the trochoidal theory are:
circular particle orbits; a rotational fluid; and a trochoidal wave surface profile.
Historícally, this nonlinear theory found favor with naval architects, who focused
not so much on the fluid kinematics with its rotational characteristic, but on
the trochoidal surface profile of these finite amplitude waves. With the advent
of offshore structures, more realistic wave kinematics were needed, and thus
trochoidal theory is not usually used by engineers in offshore structural design.
The importance of trochoidal wave theory is that it serves as a link from linear
theory to the finite amplitude oscillatory wave theory as developed by Stokes
(1845), Levi-Civita (1925), Struik (1926), and Havelock (1914).
Cnoidal Theory
A finite amplitude wave theory appropriate for shallow water is the cnoidal
wave theory, first studied by Korteweg and de Vries (1895) and more recently
by Masch and Wiegel (1961). As suggested by Sarpkaya and Isaacson (1981),
the cnoidal wave parameters are formulated in terms of elliptic cosine functions,
from which the term "cnoidal" arises. Tables and charts published in the latter
two citations aid in application of this theory.
The cnoidal wave theory was developed from the governing equations for long
waves using the assumption that the square of the slope of the water surface, or
wave steepness, is small relative to unity. One important feature is that cnoidal
waves are periodic. For small values of H / d, where H is the crest to trough
wave dimension and d is the water depth, the cnoidal wave profile is sinusoidal.
Another limiting case is for very long wave lengths, which yields a solitary wave
profile, as discussed below. The use of the cnoidal wave theory is limited to the
following range: 0.01 ::; H / d :::; O. 78 and >../ d < 8. Within this range, cnoidal
theory describes the progression of the periodic waves more accurately than
does the theory for Stokes waves. Cnoidal theory bridges the gap between the
periodic and the solitary wave theories.
Stokes Theory
The basic assumption in the development of the finite amplitude wave theory
is that the fluid motion is irrotational. This assumption can be justified phys-
ically if the fluid viscosity is vanishingly small. The governing equations are
then formulated in a manner parallel to that for linear wave theory, equations
(3.11). Those equations are as follows:
(3.18)
(3.19)
70 DETERMINISTIC DESCRJPTIONS OF OFFSHORE WAVES
au au au 1 ap
+u-+w- (3.20)
ax f)z p ax
aw aw aw 1 ap
-+u-+w-= g (3.21)
f)t ax az paz
Equations (3.18) and (3.19) express the zero vorticity and continuity conditions,
respectively, and equations (3.20) and (3.21) express the conservation of linear
momentum. Once the velocity field can be computed, then the pressure field
can be determined. The boundary conditions are that the pressure on the free
surface of the wave is everywhere a constant, or
ap ap ap
- + u - +w- =0 (3.23)
at ax f)z
Thus the free-surface boundary condition is nonlinear with respect to the un-
known variables u, w, and p.
Stokes (1847) and others solved equations (3.18)-(3.23) by a successive ap-
proximation procedure in which the solutions were formulated in terms of a
series of ascending order terms. Solutions to the second and third order are
widely available in the open literature. See, for instance, Kinsman (1965), Ip-
pen (1966), and Sarpkaya and Isaacson (1981).
Sorne frequently used results of the finite amplitude theory to the second
order are presented in Table 3.2. \Vnen the solutions for the wave surface
profiles, the partíele velocities, the particle accelerations, and the pressures given
in Table 3.2 are compared respecively to those in Table 3.1, it is noted that
each fust order term in Table 3.2 corresponds to its counterpart given by the
linear theory. The remaining terms are the second-order corrections due to
the nonlinear convective inertia terms appearing in the governing equations
(u au / ax, etc). Higher-order expressions of the Stokes theory are simply those in
which the approximations for corrective effects are carried to the corresponding
power term. In principie, if Stokes theory is carried toa sufficiently high order, it
would be adequate for describing water waves in any depth of water. In practice,
this is only possible for waves in deep water. In shallow water the convective
terms becorne relatively large, the series convergence is slow and erratic, and
a large number of terms is required to achieve a uniform degree of accuracy.
Other classical formulations such as the solitary and cnoidal theories require
fewer terms to achieve the desired accuracy.
NONLINEAR WAVES 71
acceleration wt)]
Dynamic pressure
112
Wave celerity e= (¡tanh kd)
(3.24)
where r¡ 0 is the maximum elevation above the still water level. When the Ursell
parameter is very small, linear wave theory is valid. However, one should note
that, although the Ursell parameter is a useful guide, it is not the sole measure of
determining the relative importance of the nonlinear terms. In shallow water, for
instance, the relative amplitude H/d becomes the more important parameter.
72 DETERMINISTIC DESCRIPTIONS OF OFFSHORE WAVES
---..;¡;-;;_::---- f -
r¡
S W L - - - - - - - - __ j _ - -
H--
i- --
d
Vl\\VI\WT\WT\WI~
Figure 3. 7 Profile of the solitary wave.
Solitary Theory
The solitary wave profile is illustrated in Figure 3. 7. This wave has no trough
since its profile never extends below the still water level. Solitary wave theory
describes such a wave of infinite length that propagates in water of uniform
depth. Technically, the solitary wave is the limiting case of a periodic shallow
water wave of finite height in which the wave length approaches infinity while the
relative height, H/d, is maintained constant. The usual method of generating a
solitary wave in the laboratory is by adding a finite volume of water at one end
of a closed tank.
J
Russell (1845) derived the celerity for a solitary wave as e = g(H + d).
Most analyses of the solitary wave superpose the wave celerity -e into the
field of fluid motion, thereby reducing the moving solitary wave to a stationary
wave for which the origin of the coordinate axes translates with a velocity -c.
The conditions to be satisfied are those of continuity, equation (3.llb), and
zero vorticity, equation (3.lla). The analysis begins with the definition of the
velocity potential 4;, or
aq; aq;
U=-· w (3.25)
fJx' fJz
It is seen that equations (3.25) satisfy the continuity condition, equation (3.1lb).
When equations {3.25) are combined with the zero vorticity condition, equation
(3.11a), the result is Laplace's equation, or
[J2q; &q;
fJx2 + fJz2 O (3.26)
(3.29)
If the moving horizontal axis is denoted by~' where ~ x~ct, then the potential
function cjJ, the surface water elevation r¡, and the vertical water particle velocity
w must vanish for large ~. For these conditions, and provided that H / d is less
than approximately 0.7, the solutions for cjJ and r¡ are as follows (Lamb, 1945):
(3.31)
(3.32)
H
(3.33)
When H and d are specified, lv! and N are computed by trial and correction
from the last two equations, and the particle velocity components are computed
directly from equation (3.30) using equation (3.25). It is noted that for sinu-
soidal waves r¡ is periodic and changes sign, whereas for solitary waves r¡ is
always positive. This is consistent with the solitary wave form shown in Figure
3.7.
Further analysis shows that the maximum value of H j d is equal to O. 78,
which occurs when u c. Also the expressions for particle velocities and surface
elevation are compatible with linear theory when Hjd approaches zero. Munk
(1949) surnmarized the work on solitary waves and proposed a modification
applicable to periodic waves.
Numerical Theory
vVith the advent of high speed, high capacity computers, and with the de-
velopment of efficient programming and numerical techniques, numerical wave
theories have become increasingly popular. Such theories, which are more ac-
curately described as procedures, are all based on deterministic solutions of the
flow field equations, with statistical features sometimes incorporated in the so-
lution procedures. The distinguishing features are the different treatments of
the boundary conditions and the alternative criteria defining accuracy. One
74 DETERMJNISTIC DESCRIPTIONS OF OFFSHORE WAVES
The velocities in terms of '1/J and the velocity potential rjJ are
Z=- d d
~ w-o ~
{
\WJ\'A\VI\\tJ\\V/\w/\\v~A\W\WJ\\\
Figure 3.8 Wave boundary conditíons used ín stream functíon theory (Dean, 1965).
The boundary conditions to be satisfied are shown in Figure 3.8 and are sum-
marized:
l. At the sea fioor where z = -d, the boundary is fiat, horizontal, and
impermeable. Thus
8'1/J 8q;
w = 8x = 8z o (3.36)
2. At the free surface where z = r¡, the particles remain on that surface, or
(3.37)
A question that frequently arises concerns the selection of a wave theory for
a given situation. Unfortunately, there are severa! bases for evaluating these
various theories. No consensus has yet emerged as to a cornrnon basis. Dean
(1974) and LeMehaute (1976) have studied the problem and provided Figures
3.9 and 3.10, respectively, to aid in selecting an appropriate theory. Here H is
the wave height (twice the ampitude A of Figure 3.2); Hs is the value of H
when the wave break:s; and d is the water depth.
76 DETERMINISTIC DESCRIPTIONS OF OFFSHORE WAVES
Dean's study of 1974 is a quantitative one, based on the closeness of fit of the
kinematic dynamic free surface boundary conditions. It is limited to waves lying
between the upper breaking limit curve and the lower curve labeled H = HB/4
in Figure 3.9. We see that within this range there are three possibilities: the
cnoidal first order, the linear, and the free stream function fifth order theory. If
it is desired to use only analytic theories, then the range of application of the
cnoidal theory may be extended to cover all of that region lying below the lower
dashed curve; the Stokes fifth order may be employed for that region lying below
the upper dashed curve; and the linear theory may be extended for application
to the regían lying between the two dashed curves.
Figure 3.9 Periodic wave theories providing best fit to dynamic free surface
boundary conditions. The region of validity for the free stream function fifth order
theory is encompassed by the bold boundary lines (Dean, 1974).
there is sorne overlapping '".rith the higher-order Stokes theory domains and the
cnoidal theory domain. Le Mehaute's graph is particular!y useful during pre-
liminary engineering calculations since it indicates the possibility of employing
simple theories easily salvable on hand-held calculators.
The following numerical examples íllustrate the utility of these graphs by
Dean and Le Mehaute.
0.1
STOKES 2nd ORDER
N
o
!U
~
N
¡.....
--::e
AIRY (LINEAR THEORY)
10
Figure 3.10 Limits of validity for selected wave theories. The region of validity for
the free stream function fifth order theory is encompassed by the bold boundary lines
(Le Mehaute, 1969).
Example Problem 3.1. Find the maximum horizontal and vertical water
particle velocity at an elevation of 10 ft below the still waterline, in a wave that
has a period of 8 sec, a height peak to trough of 2 ft, and is propagating in a
constant water depth of 200 ft.
l. Deduce the appropriate theory by first computing the following fiow
parameters:
dfT 2 200/82 = 3.125 ft/ sec2
78 DETERMINISTIC DESCRIPTIONS OF OFFSHORE WAVES
From Figure 3.10, we can see that this wave can be described by linear wave
theory, and that it is a deepwater wave.
2. Compute the wave length ). and wave number k by combining equations
(3.2), (3.7), and (3.17). In the last of these equations, use the approximation
for deep water: tanh kd e:::- 1:
32 2 2
JL T 2 = · 8 328 ft
27r 27r
27r
k 0.0192 ft- 1
328
3. Calculate u and w from linear wave theory using Table 3.1:
This value is a maximum when the cosíne term ís + l. Note that for z = -··10
ft, d + z 190 ft. Substitute the numerical values to obtain the maximum
horizontal water particle velocity, or
1r(2) cosh [0.0192(190)]
Umax 0.649 ftjsec
8 sinh [0.0192(200)]
From Table 3.1, we observe that the maximum value of w, the vertical water
particle velocity, occurs when the sine term is + l. Thus
1rH sinh [k(d + z)) 1r(2) sinh [0.0192(190)]
'W -
T sinh kd 0.648 ft/sec
max - -8- sinh [0.0192(200)]
We observe that the maximum particle velocitíes are very nearly identical,
where the difference is due to roundoff errors associated with the hyperbolic
functions. Further, these maximums do not occur at the same time. Profiles of
Umax and Wmax can be generated by varying z from O to -200 ft; and the entire
history of the wave fiow can be mapped by varying x and t in the trigonometric
functions for u and w.
Example Problem 3.2. Determine the surface profile and the pressure vari-
ation at an elevation of 30 ft above the seabed in a wave that has a period of 6
sec, a height of 3 ft, and is propagating in a constant water depth of 50 ft.
l. Deduce the appropriate theory by first computing the following two fiow
parameters:
From Figure 3.10, we see that this is an intermediate depth wave and that Stokes
second-order theory is appropriate.
2. Compute the wave length >., the wave number k, and the circular fre--
quency w. For this intermediate depth wave, use the dispersion relation, equation
(3.17), to compute the wave length. Using k 21rj>. from equation (3.2) and
e= A./T from equation (3.7), the dispersion relation can be expressed implicitly
in terms of ). and the given wave parameters, or
This equation, when solved for >.. by tríal and correction, leads to the following
results:
3. Use the appropriate expressions from Table 3.2, together the given values:
H = 3 ft, and d 50 ft. With the values of >.., k, and w just computed, the
surface elevation r¡ is thus:
These results for the surface elevation show that, due to the second term on the
right, there is approximately a 3 percent correction to the linear theory due to
second order effects. The results for the pressure p at the depth z 30 - 50 =
- 20 ft in water of mass density p = 1.94 slug/ft 3 are found by substituting the
foregoing numerical results into the full expression for p in Table 3.2. In the
numerical results below, the four terms on the right side of the equation, from
left to right, represent, respectively: the hydrostatic pressure of linear theory,
or -pgz -1.94(32.2)( -20) = 1249.36 lb/ft 2 ; the dynamic pressure of linear
theory; the second order dynamic correction to the linear theory; and the second
order hydrostatic correction to linear theory:
p = 1249.36 + 49.52 cos (kx wt) + 0.07 cos 2(kx wt) - 0.47lb/ft2
Again we observe that the second order corrections to linear theory theory are
very small, less than 1 percent in this example problem.
Example Problem 3.3. Determine the celerity e for a wave with a period of
15 sec and a wave height of 2 ft, propagating in water 20 ft deep.
l. Deduce the appropriate wave theory by computing the following param-
eters:
By referring to Figures 3.9 and 3.10, we see that both the cnoidal and
free stream function theories are appropriate for describing this wave, for which
computer-aided solutions are needed.
2. Rather than generate our own numerical results, refer to results of Masch
and Wiegel (1961). That is,
Compute: T ¡¡¡¡;i = 15J32.2/20 19.03
PROBLEMS
3.1 For simple wave swells with periods T = 2n-jw = >.je, 1 < T < 30 sec,
where the depth to wave length ratio satisfies dj A > 0.5, the waves are defined
as deepwater waves. In this case show that the wave celerity and period have
the following approximate forms:
e= (~~r/2 i T = 27re
g
If T = 10 sec and >. 512 ft, calculate the wave celerity and the mínimum
depth for which these relationships are valid.
3.2 A simple, shallow water wave is defined when the ratio of water depth
to wavelength satisfies dj >. < 0.05. In this case, show that the wave celerity is
given approximately by e = ygd. If >. = 512 ft, what is the maximum water
depth for which this relationship holds?
3.3 For simple linear waves of length >. at intermediate water depths d,
then 0.05 < dj >. < 0.5. Deduce that the square of the wave celerity is given by
the following formula.
e= g..\ tanh 21rd
27r
REFERENCES 81
Plot three curves corresponding to ,\ 200ft, 400ft, and 600ft, which show
the wave celerity as a function of water depth in this intermediate range.
3.4 A deepwater wave having a length of 512 ft and a height of 20 ft
propagates from deep to shallow water over the sea bottom which has a slope
of 1:30. Determine the wave length and celerity as a function of depth.
3.5 In order to collect site-specific oceanographic data, a tower is installed
in a water depth of 80 ft. One gage located 50 ft above the bottom senses an
average maximum dynamic pressure of 150 lbjft 2 at a period of 9 sec. Using
these data, compute the height and length of the corresponding wave. Is your
computed wave height the maximum that is present at the site? Explain.
3.6 We know that surface tension forces become more important as the
wave lengths decrease. In particular, as wave lengths approach 1 in., the surface
tension forces become important relative to the gravity forces. We wish to make
a model study of a semisubmersible platform (200 ft x 200 ft) as it is excited
by waves with periods ranging from 4 to 15 sec. What is the mínimum scale
ratio that is acceptable? What factors other than surface tension effects should
be considered in determining this scale ratio?
3. 7 Suppose that a deepwater wave having a period of 10 sec and a height
of 10 ft is propagating normally toward shore over a gently sloping beach. De-
termine the depth of water and the wave length when the wave breaks. (Hint:
Remember that as the wave length changes, the height must also change.)
REFERENCES
Dean, R. C., Relative Validities of Water \Vave Theories, ASCE Proceedings: First
Conference on Civil Engineering in the Oceans, San Francisco, CA, 1967.
Dean, R. C., Evaluation and Development of Water Wave Theories for Engineering
Application: Presentation of Research Results, Special Report No. 1, U.S.
Army Corps of Engineers, Coastal Engineering Research Center, Ft. Belvoir,
VA, 1974.
Gerstner, F., Theorie der Wellen, Abhandlungen der koniglichen bohmi.schen Gesellschaft
der Wissenschaften, Prague, 1802.
Hirt, C. W., Nichols, B. D., and Romero, N. C., SOLA-A Numerical Solution Algo-
rithm for Transient Fluid Flows, Report LA-5852, Los Alamos Scientific Labo-
ratory, 1975.
lppen, A.T., ed., Estuary and Coastline Hydrodynamics, McGraw-Hill, ;'1/ew York,
1966.
Kinsman, B., Wind Waves, Their Generation and Propagation on the Ocean Surface,
Prentice-Hall, Englewood Cliffs, NJ, 1965.
Korteweg, D. J., and de Vries, G., On the Change of Form of Long Waves Advanc-
ing in a Rectangular Canal, and On a New Type of Long Stationary Waves,
Philolophical Magazine, fifth series 39, 1895.
Laplace, P. S., de, Sur la vitesse du son dans fair et dans l'eau, Annales de chimie et
de physique 2 (3), 1816.
Masch, F. D., and Wiegel, R. L., Cnoidal Waves: Tables of Functions, Council on
Wave Research, The Engineering Foundation, University of California, Rich-
mond, CA, 1961.
Munk, W. H., The Solitary Wave and Its Application to Surf Problems, Annals of
New York Academy of Science 51 (3), 1949.
Rankine, W. J. M., On the Exact Form of Waves near the Surface of Deep Water,
Philolophical Magazine, London, 1863.
Russell, J. S., Report of the Committee on Waves, Meeting of the British Association
for Advancement of Science, 1845.
Sarpkaya, T., and Isaacson, M., Mechanics of Wave Forces on Offshore Structures,
Van Nostrand Reinhold, New York, 1981.
Stokes, G. G., On the Theories of the Interna! Friction of Fluids in Motion, and
of Motion of Elastic Solids, 'Jlransactions, Cambridge Philosophical Society 8
(287), 1845.
REFERENCES 83
Ursell, F., The Long Wave Paradox in the Theory of Gravity Waves, Proceedings,
Cambridge Philosophical Society 49 (4), 1953.
Young, l. R., Wind Generated Ocean Waves, Elsevier Science Ltd., Oxford, UK,
1999.
4
Wave Forces
on Structures
James F. Wilson
WAVE LOADING OF FLEXIBLE eYLINDERS 85
accounts for both cylinder flexibility and the effects of the water wave length >.
can be written as follows:
eMl¡P u+ e M2¡P
7r D 2 ' 7r D 2 (.
U vi (4.1)
(4.2a)
(4.2b)
1rD
1.54 for T > 0.5 (4.2d)
d
Tl
The total horizontal force PI (t) is given by integrating equation (4.1) over
the length of the cylinder. For Cn =O, this is
7r ¡(i'-d) 7r ¡(i'-d)
CM1- pD
2
üdz + CM2- pD 2 (ü v)dz (4.3)
4 -d 4 -d
The stiffness of the cylinder, equal in magnitude to the horizontal static force
applied to the top which will produce a unit horizontal deflection, is easily
shown to be k 1 3Eijf. 3 . The equivalent lumped mass, computed in chapter
n, 1s m 0 0.227m0 t'. For a conservative (or high) estímate of v, lump p¡(t)
at the top of the cylinder, or the coordinate point, z t' d. For a material
damping constant of c 1 , the equation of motion becomes
moii + e¡ v + k1 v = PI (t) (4.4)
It is further assumed that ü is much larger than ii along the cylinder and that ii
is independent of the coordinate z, consistent with this single degree of freedom
model. Using equation (4.3) with the mass and stiffness results just obtained,
equation (4.4) becomes
(4.5)
Using ü from Table 3.1 (with x = 0), the integral in equation (4.5) is easily
evaluated. It is seen from this last result that C,H2 is actually the added mass
coefficient CA. The virtual mass, or the coefficient of ii in equation (4.5), thus
arises in a natural way in this formulation.
Example Problem 4.2. Solve Example Problem 4.1 shown in Figure 4.1, but
this time include the fluid drag term. Then linearize the resulting equation of
motion following the method discussed by Berge and Penzien (1974).
The steps in the solution are summarized as follows. The equation of motion
is
moii + c1 v + k 1 v ¡ -d
(i'-d)
qdz = p¡(t) (4.6)
v) a (4.7a)
(4.7b)
CLASSIFICATION OF FLUID LOAD REGIMES 87
Here, T0 is the time required for several oscillations of the cylinder. When
equations (4.6) and (4.7a) are combined, the result is
(4.8)
The results ofthe two dynamic models expressed by equations (4.5) and (4.8)
lead to two important conclusions regarding the fluid-structural interactions.
First, the added mass term is the same, whether or not velocity-dependent fluid
drag is present. Second, system damping is increased with the addition of fluid
drag, as is seen by comparing the coefficients of v on the left sides of equations
(4.5) and (4.8).
TOPVIEW
CAISSON
SIDEVIEW
For this example, the total horizontal force, vertical force, and overturning mo-
ment on the cylinder are
(4.9)
(4.10)
The terms Fkx and Fkz, defined as the Froude-Krylov forces, are the net pressure-
induced forces on the vertical sides and on the top horizontal surface, respec-
tively. Those forces, located at the respective centers of pressure z and x, are
time-dependent. The basic assumption in the calculation of the Froude-Krylov
forces is that the wave pressure field is completely undisturbed by the presence
of the structure. Hogben and Standing (1975) recommended the following flow
coefficients for a submerged cylinder of diameter D and height h such as shown
in Figure 4.2. The water depth is d and the length of the incident wave is ,\.
(4.11)
7rD)
1+0.74 (""T"
2
h
D' for 1fh
-,\
<1 (4.12a)
1rD 1fh
Cv = 1+ 2I, for ->1
,\
(4.12b)
(4.14)
h
0.3 < D < 2.3, for Ch, Cv only (4.15)
h
0.6 < D < 2.3, for Co only (4.16)
With the complementary results for Fkx, Fkz, and lv/k derived in closed form by
Sarpkaya and Isaacson (1981), the total diffraction loading for vertical cylinders
can be calculated from equations (4.9) and (4.10).
Economical calculations for diffraction foroes can be made on noncylindrical
shapes such as the submerged tank cluster of the monotower shown in Figure
90 WAVE FORCES ON STRUCTURES
Example Problem 4.3. Calculate Fkx and Mk, the Froude-Krylov hori-
zontal force and overturning moment on a rectangular box caisson resting on
the ocean floor. The box has cross section dimensions 2a x 2a and height h,
as shown in Figure 4.2. Assume that there is a single incident wave and that
linear wave theory is appropriate. Relative to the coordinate system of Figure
4.2, the dynamic pressure given in Table 3.1 becomes
Hcosh
p(x, z, t) = pg- h cos (kx wt) (4.17)
2 cos
in which k 2nj>. and w is given by equation (3.16). The value of Fkx is
computed by integrating the pressure difference over the vertical sides of the
box normal toa right-traveling wave, or
Fkx = 2a ¡-d
(h-d)
[p( -a, z, t) p(a,z,t)]dz (4.18)
With equation (4.17), the last equation is integrated and use is made of trigo-
metric identities, which leads to
D Hsinhkx . k .
... kx - 2pga k cosh kd sm ·a sm wt (4.19)
The overturing moment is the sum of two integrals. The first is lvhh, the moment
about O due to the side pressure forces. The second is Ji,fkv, the moment about
O due to the pressure forces on the top of the box. Thus
Mkh = 2a ¡
(h-d)
-d
[p( -a, z, t) p(a, z, t)](z + d)dz (4.20)
where
(4.22)
The evaluation of this latter moment in closed form using equations (4.17),
(4.20), and (4.21) is a straightforward exercise. With these results, the total
horizontal diffraction force p 1x(t) and the overturning moment Mo(t) are then
given by equations (4.9) and (4.10), using the respective flow coefficients of
equations (4.11) and (4.13). The calculation of Fkz and its corresponding total
vertical load P1 z(t) is straightforward. See Problems 4.13 and 4.14.
FLOW REGIMES FOR OFFSHORE STRUCTURES 91
A perspective on the size of components (caissons, legs, bracings, etc.) for typi-
cal offshore structures is shown in Figure 4.3. The gravity and tethered buoyant
platforms, the most massive of the offshore structures, will experience diffraction
forces under most wave conditions. At the other extreme, the jackup platforms
with their relatively thin legs and small diameter bracing will experience mainly
drag forces.
r¡
-
1
1 1
GRAVITY AND CAISSONS
TEATHERED BUOYANT
PLATFORMS COLUMNS
HULLS
- -
SEMISUBMERSIBLE
STRUCTURES
COLUMNS
BRACING
JACKET
LEGS
-
STRUCTURES
BRACING
JACKUP
RIGS
-
BRACING
LEGS
1 1 1 1 11 d 1 1 1 11 1
0.1 0.2 0.5 2 5 10 20 50 100 200
CYLINDER DIAMETER, meters
A further insight into the influence of wave height and cylinder diameter
on the occurrence of the three flow regimes was presented by Hogben (1976).
Such a mapping, shown here only for cylinders at the water surface (z 0), is
illustrated in Figure 4.4. The following assumptions were used in deriving these
curves.
l. Linear wave theory, Table 3.1.
2. Deep water approximation, where w2 e::: gk.
3. A wave length to wave height ratio of )..j H 15.
4. C111 = 2.0 for all flow regimes. Cv 0.6 for postcritical drag where the
drag force changes abruptly at Re= 5 x 105 . Cv 1.2 for subcritical drag.
5. Re based on the amplitude of the wave's horizontal particle velocity.
92 WAVEFORCESONSTRUCTURES
lOOn---~~~---T~----~~
50
10
5
DIFFRACTION
DOMINA TES
0.1'--"-'-'-"---__._.___ __.___,
0.1 0.5 1 5 10 50
CYLINDER DIAMETER, meters
The five main observations of this study are summarized. First, gravity-type
platforms experience diffraction at wave heights up to 30 m, the upper extreme
for ,\ 450 m. Second, as the depth of the cylinder below the surface is in-
creased, the boundary of the diffraction regime remains fixed, but the boundary
of the inertia regime of Figure 4.4 expands upward, a result shown in other
studies for which z >O (Hogben, 1976). Third, towers supporting the deck of a
platform generally líe in the inertia flow regime in the extreme wave conditions
usually assumed for design purposes. Fourth, loads on very small diameter
components such as those of conductor tubes, legs of jackup platforms, and
jacket-template structures are generally drag-dominated, but in deeper waters
such as in the North Sea (150m or more in depth) the main structurallegs will
be large enough to incur both inertial and drag loads. Fifth, the ratio iJ.n/iir
in Figure 4.4 shows the relative effect of fluid drag forces to fluid inertía forces.
When this ratio is 0.9 (the upper dashed curve), the drag and inertia forces are
of comparable magnitude; but when this ratio is 0.1 (the lower dashed curve),
the inertia forces domínate.
Because of the inherent assumptions in Hogben's 1976 study, Figure 4.4 for
z = O and bis companion results for z > O should not be used directly to design
offshore structures. However, calculated results presented in a similar form, but
based on flow coefficients and wave parameters appropriate to the site of the
offshore structure, can be quite useful in engineering design.
SUMMARY OF FLOW COEFFICIENTS 93
A survey of measured values for CD and CM made by the British Ship Research
Association (1976) is depicted in Figure 4.5. This chart is a composite of histor-
ical data for smooth, vertical cylinders. This chart shows the wide variability
of these flow coefficients and their dependency on the Reynolds number, the
Keulegan-Carpenter number, and the wave height to cylinder diameter ratio,
H/ D. For the curves showing the detailed behavior of CD and CM in the sub-
critica! regime of Reynolds number, the reader is referred to the original report
of Keulegan and Carpenter (1958) and the summaries by Sarpkaya and Isaacson
(1981). For this lower range of Re, both CD and CM depend strongly on Kc.
1.2- CD-0.5
CM=1.5
Inertia forces are small
Inertia forces are so errors in CM are not
small so errors in C:W significant. Possible
are not significant; errors: 20% for total wave
possible error in wave forces, but may exceed 50%
forces is 100% for local wave forces.
1.2- CD-0.6
CM= 1.5
Figure 4.5 Summary of CD and CM for smooth, vertical cylinders (British Ship
Research Association, 1976).
94 WAVE FORCES ON STRUCTURES
The most important aspects of the data in Figure 4.5 are summarized.
L The Cn and CM ·values apply only to smooth cylinders in deepwater for
which dj(gT 2 ) > 0.003, where d is the water depth, T is the wave period, and
g is the acceleration due to gravity.
2. No allowances were made for wave slamming or fluid interactions with
other structural members in proximity to the smooth cylinder.
3. No allowances were made for current-wave interactions.
4. Re and Kc were not always well defined in this data survey. This may
account somewhat for the wide scatter in the reported results for these coeffi-
cients.
5. The e D and e M values listed in this table should be used with a wave
theory appropriate to environmental conditions, where the wave theory chosen
is used as a basis for calculating Re and Kc. Approximate values of these latter
parameters are generally adequate for estimating the flow coefficients.
The basis for the formulation of the transfer f1mctions in this section is linear
theory for a single water wave, often called a simple wave. Recall that for linear
theory, Table 3.1, the wave characteristics are defined by its height H = 2A, its
period T 2trjw, and its wave number k 2tr/A. where).. is its wave length.
Recall also that for deep water waves where the water depth d > 0.5).., then
k= w 2 1g. See Problem 3.1.
The transfer function G(w) is defined as the function that relates H of the
incident wave to the load it imparts to the structural component. In general,
transfer functions are defined in the following harmonic form:
loading function
real{G(w)} (4.24)
wave height
The transfer function for a loading ij is calculated, for instance, by fust finding
ij1H, and then casting that result in the forro of equation (4.23). The following
identities are useful in such a procedure:
Example Problem 4.4. Calculate the transfer function for the flexible,
cantilevered cylinder of Figure 4.1. Assume that the fluid-induced forces are
inertia dominated and that the motion of the cylinder is much smaller than the
motion of the water particles. Thus
(4.27)
. H coshk(z+d).
u= --w2 s1n wt (4.28)
2 sinh kd
Using equation (4.28) with (4.27), it follows that
ij _-~e D2 2 coshk(z + d) .
H - 8 M pw sinh kd sm wt (4.29)
With equation (4.25a), the latter ratio yields the transfer function as
The square of the modulus, calculated from equations (4.26) and (4.30), is
IG(w )l =[~e
2
2 D 2 2 cosh k(z+d)] (4.31)
8 M pw sinh kd
It is noted that the expression for G(w) just derived is for ij and is a function
of the location z on the cantilevered beam. The transfer function required for
the lumped mass model, equation (4.4), must be based on the total load p 1 (t).
This latter transfer function can be derived by simply integrating equation (4.30)
over the range of z = -d to z = (C- d). The resulting transfer function and its
corresponding modulus will then be independent of z.
96 WAVEFORCESONSTRUCTURES
Example Problem 4-5. Calculate the transfer function for the total hori-
zontal load on the box-type caisson shown in Figure 4.2 and described in Exam-
ple Problem 4.3. Assume that the fiuid-induced forces are diffraction-dominated.
It follows that the total horizontal load p 1 x(t) is the product of Fkx given by
equation (4.19) and the diffraction coefficient eh given by equation (4.11). Thus
the totalload-wave height ratio is
~w
-¡¡- = ~2pgaeh k~~ .
cosh kh sm ka sm wt
. (4.32
)
The required transfer function is this ratio expressed in complex notation, or
These examples illustrate that there is a different transfer function for each
fiow regime and for each structural component of an offshore structure. In
practice, transfer functions for all components are calculated and assembled
for the structure, all based on linear wave theory and a single, simple wave.
Such results can then superimposed to accouut for the loading effects of many
simple waves selected to simulate the design sea state at the si te of the offshore
structure. This simulation involving the selection of simple waves over particular
ranges of H, w, and wave phase e is discussed in Chapter 6. The corresponding
structural responses are then discussed in Chapter 7 for single degree of freedom
systems and in Chapter 9 for multi-degree of freedom structures.
PROBLEMS
4.1 The total force PI(t) on a submerged, flexible pile in the inertia fiow
regime is given by the right side of equation (4.3). For a single, simple wave
evaluate p 1 ( t) by carrying out the necessary integration.
4.2 The diameter of the submerged pile shown in Figure 4.1 is sufficiently
small so that the drag forces are comparable in magnitude to the inertia forces.
Evaluate p 1 (t) in this case by carrying out the integrations of equation (4.8).
Figure 4.6 Monopod structure for Problems 4.3, 4.4, and 4.5.
PROBLEMS 97
M
e
1:--...--fño, El
(4legs)
SECT. A-A
4. 7 The unbraced platform shown in Figure 4. 7 has four legs, each of di-
e
ameter D and length 0 , separated by distance E. The legs are subjected toa
single, simple plane wave for which >. ~ D. Neglect the fluid drag forces and
also neglect the motion (v, v) of the structure compared to the respective hori-
zontal motíon parameters (u, ü) of the wave particles. Derive the expression for
wave loading in each leg, accounting for the relative magnitude of the separation
length Ccompared to the wavelength >.. Then discuss the net wave-induced force
98 WAVE FORCES ON STRUCTURES
on the whole structure for the three cases: f/ >.. «: 1, f/ >.. = 0.5, and f/ >.. = l. If
the net structuralload is zero for any of these cases, explain why the individual
wave loads on the legs can still be significant. Sketch possible leg deformation
patterns for each of these three cases.
4.8 Search the recent literature, such as the annual Proceedings of the Off-
shore Technology Conference, to determine whether diffraction coefficients such
as Ch, Cv, and C 0 have been measured for fixed objects other than vertical, right
circular cylinders.
4.9 For a deep water wave, show that equation (3.16) reduces to w2 = gk.
Then, for a vertical cylinder of diameter D and height h, express the diffraction
coefficients of equations (4.11) through (4.13) in terms of g, D, h, and w only.
4.10 Determine an efficient analytical or numerical procedure to calculate
wave diffraction loads on a rigid, right circular conical monopod leg of an ice-
resistant offshore platform. Consult the classical references of Hogben (1976)
and also recent literature on the subject.
4.11 The legs of a jackup platform are cylinders of diameter 0.5m. Each
leg is subjected to a single, simple wave, with each impact occurring separately
and at a different time. For each ratio >..j H = 10, 15, and 20, determine the
range of H for which the drag, inertia, and diffraction dominates.
4.12 Based on the data of Sarpkaya and Isaacson (1981), deduce the values
CM and CD for cylinders in the flow regime of subcritical Reynolds numbers,
where 10 < Kc < 40. That is, add typical flow coefficients to the left column of
Figure 4.5.
4.13 Calculate the explicit expressions for the Froude-Krylov force Fkz,
and the overturning moment Mk for the solid box caisson of Figure 4.2.
4.14 With the results of Example Problems 4.3 and 4.5, together with
Problem 4.13, complete the calculations for the three most important wave load
transfer functions on the box caisson: those for the total horizontal load, the
vertical load, and overturning moment. Use complex notation to describe G(w)
2
in each case; then calculate each value of IG(w)l .
REFERENCES
Berge, B., and Penzien, J., Three-Dimensional Stochastic Response of Offshore Tow-
ers to Wave Forces, OTC-2050, Proceedings of the Offshore Technology Confer-
ence, 1974.
Hogben, N., and Standing, R. G., Experience in Computing Wave Loads on Large
Bodies, OTC-2189, Proceedings of the Offshore Technology Conference, 1975.
REFERENCES 99
Keulegan, G. H., and Carpenter, L. H., Forces on Cylinders and Plates in an Oscil-
lating Fluid, Journal of Research of the National Bureau of Standards 60 (5),
1958.
Moan, T., Haver, S., and Vinje, T., Stochastic Dynamic Response Analysis of Off-
shore Platforms, with Particular Reference to Gravity Platforms, OTC-2407,
Proceedings of the Offshore Technology Conference, 1975.
Nataraja, R., and Kirk, C. L., Dynamic Response of a Gravity Platform under
Random Waves, OTC-2904, Proceedings of the Offshore Technology Conference,
1977.
Sarpkaya, T., and Isaacson, M., Mechanics of Wave Forces on Offshore Structures,
Van Nostrand Reinhold, New York, 1981.
5
Deterministic Responses for
Single Degree of Freedom
Structures
James F. Wilson
NATURAL FREQUENCIES OF LINEAR SYSTEMS 101
In this section, two methods for calculating the natural frequency of linear single
degree of freedom structural models are presented: the Direct Method, which is
based on the equation of motion; and the Rayleigh Method, which is based on
energy principles. Several example problems are presented that illustrate the
relationships between these two methods.
Direct Method
The direct method of obtaining the natural frequency of a single degree
of freedom structure is to use its equation of motion. The structure's natural
frequency wo is defined as the frequency compatible with an undamped structure
of constant mass, with a restraint force that varies linearly with the displacement
coordinate, and with no externa! excitation force. Under these conditions, the
governing equation of motion, written in terms of the displacement coordinate
v, is
mv + klv =o (5.1)
which is a special case of equation (2.2). When this linear structure is given an
arbitrarily small displacement amplitude vo and then released, then v exhibits
free harmonic oscillations of the form
When equation (5.2) and its second derivative are substituted into equation
(5.1), the result is
In the latter equation, the term sin w 0 t is not zero for all time t and thus the
term in brackets must be zero. This leads the following equation for the natural
frequency of the structure:
(k;
wo =y-;;¡; (5.4)
mg+oW
(a) EQUILIBRIUM STATE (b) FREE BODY SKETCH
Shown in Figure 5.1b is the free body sketch of the buoy with a vertical
displacement v. In the static state, v =O and according to Archimedes principie,
the weight of the buoy floating at its equator is given by the weight of the water
displaced by half ofthe sphere's volume, or mg = 21rR3 "Yw/3. Also by Achimedes
principie, the incremental upward force 8W for a downward buoy displacement
v corresponding to an approximate incremental water volume displacement of
1rR2 v, is 8W = 1rR2 "'fwV. When Newton's second law is applied to the buoy of
Figure 5.1 b, the equation of motion for the buoy becomes
After substituting the loads and rearranging, the above equation becomes
(5.6)
When equation (5.2) is used with the above result, the natural frequency of the
buoy is computed as
side sinO e,f(O) =O, and q(B) = koB (the soil foundation reaction moment).
Thus, equation (2.6) reduces to
(5.8)
(5.9)
where Oo is a constant. When ()ande from equation (5.10) are substituted into
equation (5.9), the result is a quadratic equation in wo, which can be solved for
the positive root wo using the quadratic formula. The result is
wo
2
1 2
Jo {[bo + 4Jo(ao- m 0 gha + m 0gh0 )] 11~ bo} (5.11)
t
30 40
t
80
771TI=+------4----L,.,....,;,r--DIMENSIONS:
meters
Table 5.1 System Parameters and Results for Example Problem 5.2
Soil properties Gs = 10 MPa to 50 MPa
Ps = 2000 kg/m3 ; l/ 8 = 0.33
Soil parameters a 0 = 3.63 x 10 12 N·m for Gs = 10 MPa
bo = 4.97 x 1011 N·m·s for Gs= 10 MPa
Structural mass, actual mo = 3.56 x 108 kg
8
Structural mass, buoyant mb = 2.59 x 10 kg
Center of actual mass ha=30.7m
Center of buoyant mass hb=31.7m
Structural inertia, virtual Jo= 1.46 x 10 12 N·m·s
Mínimum rocking frequency wo 1.41 rad/s
Maximum rocking frequency wo = 3.17 rad/s
which satisfies the geometric constraints imposed at the ends: for displacement,
1.¡'1(0) = 1.¡'1(!!) O; and for slope, 1.¡'1'(0) = 1.¡')'(f) O. Here (')denotes the operator
(djdx).
For the jackup platform of Figure 2.17, a judicious choice of mode shape is
which satisfies the geometric constraints: 1.¡'1(0) = 1.¡')' (O) 1.¡'} 1 (f) O and 1.¡'1( e)
2. The last constraint is nonzero, which is consistent with a nonzero amplitude
106 SINGLE DEGREE OF FREEDOM STRUCTURES
for the deck. If this jackup platform were modified so that the deck was instead
hinged instead of clamped to each leg, then an approximate mode shape would
be
1rX
'1/J(x) 1- cos- (5.14)
2€
which satisfies the geometric constraints of full fixity at the sea floor: w(O) =
'!¡0'(0) O; and of nonzero displacement and slope at the deck level: w(C) = 1
and 1/;'(C) = rr/(2€).
For the Rayleigh Method, the actual magnitudes of the nonzero end slopes
and deflections are unimportant, provided that the chosen mode shape w(x)
satisfies the beam's geometric constraints. Then this method leads to an upper
bound value for wo. A straightforward proof of this upper bound property was
provided by Den Hartog (1947).
The steps of the Rayleigh Method used to calculate w 0 for beam-type struc-
tures are summarized as follows:
l. Choose a simple mode shape w(x) that satisfies the geometric boundary
conditions of deflection and slope. Express the lateral displacement of the beam
at location x along its length as
v v(x, t) vo ?f'>(x) sin wot (5.15)
where vo is an arbitrary constant.
2. From the latter equation, calculate V U+ Vg, the total potential energy
for the beam. Here U is the ela..<;tic strain energy for bending of the beams (legs
and/or cross braces) and Vg is the loss in potential energy dueto a downward
displacement of a mass, as for instance, the decrease in the gravitational energy
of a deck mass md due to the lateral displacement of its vertical supporting legs.
3. Based on equation (5.15), calculate the total kinetic energy K: that for
the beam plus that for its end mass, if any.
4. Calculate wo by equating the expressions for the maximum kinetic energy
(when V O) to the maximum potential energy (when K 0), or
(5.16)
Befare the Rayleigh method is illustrated, consider sorne general forms of
system energy. From elementary beam theory, the strain energy for bending of
a beam element of length dx at position x is
dU = EI (fPv)2 dx (5.17)
2 8x 2
where v v(x, t) is given by equation (5.15). The total potential energy of
N identical beams each of length e, and each with the identical displacement
v(x, t) is thus
(5.18)
NATURAL FREQUENCIES OF LINEAR SYSTEMS 107
(5.19)
where !:lh is the vertical drop in height of the deck dueto curvature of the elastic
beams (legs). The beam element of are length ds shown in Figure 2.17b can be
approximated using a binomial expansion, or
from which the total drop of md is calculated by integrating over the length of
the beam, or
fe
!:lh =lo (ds- dx)
1 {'-
2 lo
(8v)2
ax dx (5.22)
\Vith the latter result and equation (5.19), the decrease in the potential energy
of the deck mass due to the lateral leg motion is thus
T7
~g-
- (5.23)
Consider the kinetic energy for the beam and deck elements. For a sub-
merged beam element of length dx with a virtual mass per unit length of m, the
kinetic energy is m(8vj8t) 2 dxj2. (If the element is not submerged, replace m by
iño, its actual mass per unit length.) The kinetic energy for a fully submerged
beam such as in Figure 2.16a is thus
(5.24)
For N beams or legs only partially submerged, such as in Figure 2.17, the result
is
K -N
2
1d (av)
0
m
at -
2
N
dx+-
2d
1'- (av)
. mo
at -
2
dx (5.25)
where d is the water depth. For this jackup platform there is an additional
kinetic energy term due to the movement of the deck mass. If the rotational
108 SINGLE DEGREE OF FREEDOM STRUCTURES
(5.26)
Example Problem 5.3. Use the Rayleigh Method to determine wo for the
fully submerged cross beam clamped at each end, as shown in Figure 2.16a.
From this frequency, deduce the lumped mass for the single degree of freedom
model equivalent to the first mode of vibration.
l. Since 1/J(x) as given by equation (5.12) satisfies the geometric constraints,
the harmonic, lateral displacement from equation (5.15) is
2
v(x, t) =va ( 1- cos ~x) sin wot (5.27)
{R 2 27rx
Vma.x = Uma.x = 21 (271")4 2
--¡ v0 EI Jo cos -1!-dx =
47r4 2
f)f"EI v0 (5.28)
(5.29)
4. When the energies of equations (5.28) and (5.29) are equated, then
519.5EI
wo = (5.30)
f!4m
Now compare this frequency with that derived for this same problem mod-
eled previously in Example Problem 2. 7 as a lumped mass, single degree of
freedom system. There, the equation for undamped motion without external
excitation was
.. 192EI _O
mv + ---¡:¡-v- (5.31)
which follows from equations (2.41) and (2.43). The total virtual mass of the
lumped system is m and is a fraction fi of the total virtual mass m€, or
(5.32)
NATURAL FREQUENCIES OF LINEAR SYSTEMS 109
With the last two equations, the natural frequency from equation (5.5) becomes
192EJ
wo = (5.33)
C4 hm
By equating the frequencies of equations (5.33) and (5.30), then h = 0.370.
Note that this value of JI is tempered by the choice of 'ljJ(x). However, Den
Hartog (1947) points out that the exact frequency for this problem is only 1.3
percent lower than that of equation (5.30). Thus, lumping 37 percent of the
beam's virtual mass at midspan is a good approximation in this case.
Table 5.2 System Parameters and Results for Example Problem 5.4
Height of leg e= 3180 in.
Depth of water d = 2880 in.
Leg, Young's modulus (steel) E = 30 x 106 psi
Leg outside diameter D = 144 in.
Leg inside diameter Di = 140.25 in.
Leg pipe weight density (steel) 'Yp = 0.283 lb/in. 3
Sea water weight density 'Yw = 0.0375 lb/in. 3
Deck weight mdg = 1.02 x 107 lb
Second area moment, one leg I = 1r(D 4 - D[)/4 = 2.114 x 106 in. 4
Added mass coefficient CA= 1
Virtual weight per unit length, mg = 1r"fp(D 2 - D?)/4
one leg (submerged) +1· "fw1rD 2 /4 = 1417lb/in.
Actual weight per unit length, m0 g = 7r'Yp(D 2 - D;)/4 = 237 lb/in.
one leg (above water)
Acceleration due to gravity g = 386 in./sec 2
Length parameter, equation (5.37) C' = 897 in.
Length parameter, equation (5.38) C" = 296 in.
Equatl.on (5.36).· -
Wo-
[3.88x10-4(1.86x10s-l.02x107J)l/2 -136
9.88xl03+0.545xl03+26.4xlO - · ra
d/
sec
leads to
max
dx-
1
2mdg
le (08 )
O
~ 2
max
dx (5.34)
3. Using the same function v(x, t), calculate the maximum kinetic energy
based on the sum of equations (5.25) and (5.26). The result is
Kmax =- 31d
2o
(av)
iñ -
&tmax
2
dx + 31e
-
2o
(av)ziño -
atmax
dx
(5.35)
(5.36)
This same problem but without the dead weight effect of the deck was pre-
viously modeled in Example Problem 2.8 by equation (2.47), from whlch the
equation for undamped, free vibrations is deduced as
When equation (5.2) is used with the last equation, the resulting frequency is
(5.40)
In the last two equations, iñ is the virtual mass per unit length of each submerged
leg, which is the sum of its actual mass per unit length m0 and its added mass
per unit length, or
(5.41)
FREQUENCIES FOR NONLINEAR STRUCTURES 111
Several important features about the two system models can be deduced
from their respective frequencies as given by equations (5.36) and (5.40). Those
features are summarized as follows:
e
l. The Euler buckling load for this three-legged structure is PE = 37r2 EI j 2 .
The Rayleigh model frequency of equation (5.36) shows that increasing the deck
load reduces the effective bending stiffness of the legs. In fact, the structure
buckles if the deck load is sufficiently high, a critica! condition for which wo O
and mdg PE.
2. If PE ::P mdg, then the bending stiffnesses for the two models are nearly
the same, a result deduced by comparing the numerators of equations (5.36)
and (5.40). That is, the coefficients of EI j C3 in these respective equations are
37r4 /8 36.5 and 36.
3. If PE :;p mdg and the deck mass is much greater than that of the legs,
then the system mass is simply md for both models. However, such a design is
probably unrealistic.
4. If the deck mass is of comparable magnitude to that of the legs, then the
Rayleigh model shows that the virtual mass of the legs has a signi:ficant infiuence
on w 0 • For instance, in the special case where C :::::: d, then the parameters of
equations (5.37) and (5.38) become: e' 3Cj8, C" =O, and the virtual mass of
the legs becomes m= 3mftC. When the denominators of equations (5.36) and
(5.40) are compared, then 3iñC' 9iñCj8 = 3mftC, or J1 0.375. This justifies
the assumption in Example Problem 2.8 in which 37.5 percent of the virtual
mass for the submerged legs, together with 37.5 percent of the actualleg mass
above water, was lumped with the total mass of the deck to formulate the single
degree of freedom model. The recommended frequency expression for this case
is thus
e d (5.42)
in which the virtual mass for each of the three legs is given by equation (5.41).
Realistic numerical data for a jackup platform with three steel pipes for legs
are listed in Table 5.2. Listed also is the system frequency computed from equa-
tion (5.36): wo 1.36 radjsec or Jo= 0.217 Hz. From the numerical values in
the numerator of w0 , it is observed that the deck weight has a small effect on
the stiffness bending stiffness. However, numerical values in the denominator
indicate that the leg mass is a significant portion of the system mass. In con-
clusion, it is noted that the fundamental period of oscillation of this platform is
To = 1/ Jo= 4.61 sec, which is well below the 12 to 15 sec periods of the highest
energy offshore waves in a typical sea state.
In the last section it was demonstrated that for linear structural models with
small motion, the free vibration frequency was independent of the amplitude
112 SINGLE DEGREE OF FREEDOM STRUCTURES
of motion. Using first order perturbation theory, it will now be shown that
for structures with nonlinear restoring forces, the free vibration frequency will
be dependent on the amplitude of motion. In both the linear and nonlinear
systems considered here, this amplitude of free vibration is imposed as an initial
condition for the motion.
Consider the free, undamped motion of a virtual mass m with the indepen-
dent coordinate v and with a nonlinear restoring force given by equation (2.69).
The corresponding equation of motion is
(5.43)
Recall that the restoring force term in the last equation was used to model the
line stiffness of a spread moored ship. See Example Problems 2.10 and 2.11.
Now rewrite this last nonlinear equation as
v + w5v + ev 3 = O (5.44)
Now assume that the solution to equation (5.44) and its corresponding frequency
ware
v = vo(t) + ev1 (t) (5.46)
w2 = w5 + ef(A) (5.47)
(5.48)
The coefficients of e 0 and e 1 are each equated to zero and terms of higher in
e, or o(e 2 ), are assumed to be small enough to be neglected. The result is the
following two linear differential equations:
(5.49)
FREQUENCIES FOR NONLINEAR STRUCTURES 113
..
VI + W-2 VI = f(A) VQ - Vo
3 (5.50)
vo = Acos wt (5.51)
This latter solution is then inserted into the right side of equation (5.50) and
the following identity is used:
where E 1 and E 2 are constants. In this solution, the last term on the right
is called the secular term, and this term is observed to become unbounded as
t becomes large. Unboundedness is not physically possible for this system so
restrained, and therefore the coefficient of t in the secular term must vanish.
Excluding the trivial case for A =O, this reasoning leads to
f(A) = ~A
4
2 (5.54)
This result clearly shows that the frequency of the nonlinear system increases
dueto the addition of the cubic restoring force term, assuming that k3 > O, and
that this increase depends on the initial system displacement A. However, if
k3 < O, then w < wo.
Next v 1 is calculated from equation (5.53) by imposing the zero initial con-
ditions v 1 (O) = v1 (O) = O, with which the constants are evaluated as E1 =
-A/(32w 2 ) and E 2 = O. When this result and the solution for v 0 , equation
(5.51), are used with equation (5.46), the first-order perturbation solution be-
comes
v - - k3A3
= A cos wt -t -
mW 2 (cos w cos 3-t)
w (5.56)
32
This result shows that the free oscillation amplitude for the counterpart linear
system (k 3 =O) is distorted by a frequency component 3W when k3 >O.
Example Problem 5.5. Calculate w 0 and w for both surge and sway for
the spread-moored ship described in Example Problem 2.1 O. Assume that the
114 SINGLE DEGREE OF FREEDOM STRUCTURES
20,300
= 0.240 radjsec
3 .5 2 X 105
3 400 32
w= O 2402 + ( ) = 0.255 radjsec
. 4(3.52 X 105)
12,700
wo = = 0.144rad/ sec
6.1 2 X 105
2 2
w= 0.144 + 4 (63(950)3
_ x 105 )
12
-
- 0.177radjsec
These results show that, although the surge frequency is increased by only
about 6 percent by the nonlinear restoring force term, the sway frequency is
increased by a significant amount, 22.9 percent. The calculation of the displace-
ment responses using equation (5.56) is straightforward.
A= ~ k1(w
3 k3 w6
2
_ 1) (5.57)
The latter equation shows that for a hard restraint with k 3 jk 1 = 0.1 and 0.02,
then A is a real number only for wj w 0 ~ l. These results are shown in the right
RESPONSE FUNCTIONS FOR LINEAR STRUCTURES 115
half of Figure 5.3. For a sojt restraint with k3jk 1 = -0.1 and -0.02, then A
is a real number only for wj wo ~ l. These results are shown in the left half of
Figure 5.3.
There are several general conclusions for this study. First, for the same
absolute value of k3 jk 1 , and the same percent change in wjw 0 from unity, the
results show that the soft restraint allows for smaller amplitudes than the hard
restraint. Second, as w/ w 0 extends beyond the ranges shown, the results become
increasingly inaccurate because the perturbation solution begins to break clown.
Strictly speaking, these solutions are valid only for w arbitrarily close to wo.
Third, the vertical line at wj wo = 1 in Figure 5.3 is the condition reached as
1k3 / k 1 1approaches zero. Here the ampli tu de of oscillation becomes independent
Sections 5.1 and 5.2 illustrated methods for calculating the fundamental, free
vibration frequency of several one degree of freedom structural models, char-
acterized by either linear or nonlinear restoring forces. The remainder of this
chapter makes use of these frequencies to compute the solutions v = v(t) of
116 SINGLE DEGREE OF FREEDOM STRUCTURES
wo = lfi; (5.64)
(5.65)
RESPONSE FUNCTIONS FOR LINEAR STRUCTURES 117
'§'
:§_
~
¡::::
u
s 6
'~"'"
en
2
en
~
<=>:::
u 0.1
z
o 4
~ QUASI-
..: STATIC INERTIA
:r: REGIME REGIME
1 2
FREQUENCY RATIO, ro/ro 0
Figure 5.4 Harmonic response function for a linear system with small damping.
The utility of this form will become apparent later in the analysis of random
motion.
Consider the four regimes of response as deduced from equation (5.65). Note
that the behavior of IH(w)l is the same as that for lv(t)l of equation (5.60),
since the latter quantities differ only by the multiple p 0 jk 1 . For the first regime,
O ::; w / wo ::; 0.1, the case in which p 1 ( t) is applied very slowly. Then the system
response is essentially static, H(w) ---+ 1, and
-k IH ()l
Po Po = peak v(t)
w ---+- (5.66)
1 kl
118 SINGLE DEGREE OF FREEDOM STRUCTURES
For the second regime, 0.1 :::; w/ wo :::; 0.5 and the response is quasi-static, or v( t)
is amplified only a little above its static value. fbr the third regime, wfwo--> 1,
the response is in resonance and the peak response v(t) is large compared to
its static value, especially for ( near zero. In the fourth regime, w/ wo --> oo
(> 10 for practica! purposes) and the peak system response approaches zero for
all damping ratios (. The latter three response regimes for 1H (w) 1 as a function
of w/wo for several values of ( are shown in Figure 5.4. The usual range for
structural damping is 0.02 < ( < 0.10.
For the case of structural excitation by a simple harmonic water wave, how-
ever, the response amplitude curves of Figure 5.4 become distorted and thus
should not be used directly. This distortion occurs because the wave loading
amplitude Po is not independent of the wave frequency w. Recall from equation
(4.32) that Po= H · Go where H is the given wave height and Go is the transfer
functíon which is dependent on w.
Impulse Response Functíon
The impulse response function h(t) is the solution to equation (5.58) for an
impulsive load such as a swift kick ora hammer blow. Define the impulsive load
as
where C =1lb· sec or 1 N·s. The Dirac delta function b(t) is defined as zero for
all time except at t = O; and in addition has the integral property
o+
fo_ 8(t)dt =1 (5.68)
Using this loading, the impulse response function is derived as follows. Since
h(t) is a solution to equation (5.58) for p 1 (t) b(t), then
(5.69)
For t >O, then 8(t) =O by definition, and the general solution to the homoge-
neous form of equation (5.69) is
(5.70)
where C3 and C4 are arbitrary constants. The damped frequency in terms of
wo and ( of equations (5.64) is
(5.71)
One can verify by substitution using equations (5.64) and (5. 71) that equa-
tion (5. 70) is a solution to equation (5.69). The system is not displaced at
t =O, so h(O) =O. With equation (5.70), then C4 O. Next h(O) is found by
differentiating equation (5.70) and setting t =O. The result is
(5.72)
RESPONSE FUNCTIONS FOR LINEAR STRUCTURES 119
This value of h(O) must be compatible with equations (5.69) and (5.68). This
is accomplished by integrating equation (5.69) term by term over the impulse
time E, the interval from t = o- to t = 0+, Or
o+ o+
mh(E) + c1 h(E) + k 1 fa_ h(t)dt =fa_ 6(t)dt (5.73)
h(t)
Figure 5.5 Impulse response function for a linear, lightly damped system.
Shown in Figure 5.5 is a sketch of h(t) for light damping orO< (~O. l. The
corresponding period of free vibration is given by T 0 = 27f /wd. For ( :S 0.1, the
frequency ratio lies in the range 0.994 ~ wd/w 0 ~ 1.0, indicating that wo ~ Wd
is a good approximation for lightly damped linear systems.
Convolution Integral
Consider the response or solution to equation (5.58) for an arbitrary loading
Pl(t). Ata particular time t = T, apply an impulsive load of magnitude Pl(T)
over the time interval dT. This impulsive load is depicted as a shaded portion of
the function p 1 ( t) in Figure 5.6. The response dv observed at time t is simply
a multiple of the unit impulse response function over dT, which is
For the last equation, note that the time between the application of the unit
impulse and observation of its response is (t - T) instead of t as in equation
(5.74), where the impulse was applied at T =O. Define the arbitrary loading
Pl (t) as zero for all times less than zero. Thus the total response due to all
impulses applied in the interval O ~ T ~ t is given by integrating equation
(5.75) over this time interval. Using equation (5.74), the result is
v(t) = -1-
mwd o
1t PI(T)e-(wo(t-r) sin wd(t- T)dT (5.76)
pt (t)
This is a particular solution to equation (5.58) and is the form used to eval-
uate the responses of offshore structures in the present text. It is one form
of the convolution integral, sometimes called the Duhamel integral. Equation
(5. 76) does not include the two independent solutions to the homogeneous form
of equation (5.58) for two reasons. First, their inclusion would necessitate the
use of initial conditions v(O) and v(O) to obtain the complete response solution,
and those initial conditions are rarely if ever known for an offshore structure.
Second, those homogeneous solutions die out rather quickly due to structural
and externa! damping. It is appropriate to employ the Duhamel integral as
given by equation (5. 76) in computing steady state dynamic responses of off-
shore structures to the highly variable environmental loading conditions. The
Duhamel integral in the above form will be used in the next section for earth-
quake motion analysis and in later chapters for random motion analysis.
Consider the motion of the fixed-legged platform shown in Figure 2.15. The
equation of motion for this platform with horizontal base or sea floor earthquake
excitation was derived in Chapter 2 as equation (2.36), which is repeated here
with the negative sign on the right side omitted:
(2.36)
(5.77)
V
(s, mo) fixed
Figure 5.7 Typical time history of structural displacement resulting from horizontal
ground motion (an earthquake).
(5.78)
(5.79)
(5.80)
~
"'
<U
.5
0
:;..."
-
E-<
u
o.....:¡ 0.05
0.1
~
o>
0.2
o
:::::>
~
(/.)
¡l.,
Example Problem 5. 1. For the jackup rig shown in Figure 2.17, estimate the
critica! responses to the horizontal motion of the 1940 El Centro earthquake for
which ii 9 is given in Figure 2.14. The structural data needed for this calculation
are given in Example Problems 2.8 and 5.3 as: ho = 3180 in., wo = 1.36 rad/sec,
and k 1 = 6.82 x 104 lb/in. The structural period is To = 4.61 sec.
From Figure 5.8, the values of Sv for (=O and 0.1 are about 12 and 8 in.jsec,
respectively. From equations (5.78)-(5.80), the corresponding upper and lower
RESPONSE OF NONLINEAR STRUCTURES 123
lirnits for rnaxirnurn displacernent, horizontal shear load, and overturning rno-
rnent are thus
It is noted that these nurnerical results are quite sensitive to srnall changes in
the structural period. That is, since T 0 líes in a trough, a 10 percent increase
or decrease in To would increase each of these calculated responses by about 50
percent. To avoid the rnany peaks and troughs that typify the Sv vs. To curves
calculated directly frorn earthquake data, such curves are often smoothe.4 in a
statistical sense before being used for design purposes.
w =
2
w5 + eg(a) (5.84)
Here v1 (t) is the response correction function and g(a) is the amplitude function,
both arising from the presence of k 3 • The proced ure for calculating v1 ( t) and
g(a) follows that of Section 5.2, starting with equation (5.46). Also, the two
initial conditions here are those of equations (5.45). Note that w replaces the
symbol wof that previous formulation.
The procedure is summarized. When equations (5.83) and (5.84) are substi-
tuted into equation (5.82), the coefficients of e 0 and e 1 that result after sorne
algebra are each equated to zero. When the trigonometric identity of equation
(5.52) is applied and the higher order terms involving e2 , e3 , ... are ignored,
the result is two linear differential equations in v 0 (t) and v1 (t). A steady-state
solution to each is found, they are superimposed according to equation (5.83),
and the initial conditions v(O) =a and v(O) =O are applied. The secular term
in this solution, which is one of the terms of v 1 (t), is eliminated by equating
to zero the coefficient multiplying the variable t. This procedure is physically
appropriate since it preserves the boundedness of the solution as time increases.
The result is
- 3 -2 1 (5.85)
g(a) = ¡k3a - -aPo
With e= 1/m and this last equation combined with equation (5.84), the sought-
after relationship among the system parameters is derved as
(5.86)
This last result can be compared to the result for free oscillations. For free
oscillations, a = A, Po = O and w = w, for which equation (5.86) is identical
to equation (5.55), as it should be. Also it is not difficult to show that the
forced vibration response v(t) is given by equation (5.55), provided that w is
substituted for wo. Of course in the present case the amplitude a depends on
the magnitude of the excitation p 0 , as well as on the system's frequency and
stiffness. As for the previous case of free oscillations, the forced oscillation
response is distorted by a frequency component 3w.
Equation ( 5.86), the key result needed to study the behavior of this nonlinear
system, is recast in the following two forms using = kl/m: w6
a3 _ ~ k1(w
3 k3 w6
2
_ 1) a_~ k1 Po = 0
3 k3 k1
(5.87)
K A3 - (0 2 - 1)A- 1 =O (5.88)
kl
A=-a· (5.89)
Po '
RESPONSE OF NONLINEAR STRUCTURES 125
Example Problem 5.8. In the San Diego harbor a ship moored with multiple
lines was observed to oscillate or gallop in an erratic manner, even though the
harbor waves were reasonably regular and were of normal height. The initially
rather taut mooring lines were then tightened further, and after that this ship
ceased to gallop and oscillated in the same regular manner as the ships moored
nearby. Can the initially erratic ship motion and its subsequent correction be
explained using a simple, nonlinear dynamic model?
Consider the simplified nonlinear dynamic model of the ship-wave system
described by equation (5.81) in which a single incident harbor wave produced
a hull force of amplitude p 0 ata excitation frequency w. The connection of the
ship's response amplitude a to w and the restraint constants is equation (5.86).
The task is to investigate the nature of the ship's response amplitude to the
system parameters.
The mathematical question is: for what combination of system parameters
does a of equation (5.86) have one or more real roots? If there is only one
real root for a, then the motion v(t) is given by equation (5.56) with w =
w, and is quite regular, without gallops. If more than one real root exists,
then the motion could pass from one amplitude to another, giving rise to the
observed erratic behavior of the moored ship. Fortunately, the nature of the
roots for a third-order reduced cubic polynomial in a, such as equation (5.87),
was thoroughly studied early in the sixteenth century, and the results presented
below are available in standard algebra texts (Rosenbach et al.,1958). Rewrite
the latter equation in the form
a3 + aa + f3 =O (5.90)
where a and f3 are real, found by comparing equations (5.87) and (5.90). Define
1 2 2 3 81 2
-(w -wo) >--Po
3
(5.92)
k3 16m
126 SINGLE DEGREE OF FREEDOM STRUCTURES
1 2 81 2
-(w2 wo) = 16m3Po (5.93)
k3
There is only one real root if
2)3 81 2
(5.94)
Wo < 16m3Po
To explain the ship's behavior, two reasonable assumptions are made. First,
the wave excitation parameters p 0 and w remained constant, or nearly so, from
befon~ to after the mooring lines were adjusted. Since the ship's mass m re-
mained constant, then the right sides of equations (5.92)-(5.94) also remained
constant. Second, the inequality given by equation (5.92) was true before the
line adjustment since multiple real roots a lead to erratic motion, which is ex-
plained in more detail in the next section. This inequality (5.92) implies that
w > wo. The equality condition of equation (5.93) can be eliminated because
such exactness is rarely possible in the physical world.
After tightening the mooring lines, k 1 definitely increased; and k3 decreased
somewhat because the static restraint stiffness curve q( v) vs. v would tend to
straighten out as the initial slope of this curve increased. For a small increase in
k1, the wij term increased, causing the term in brackets on the left of equation
(5.92) to decrease. Since this latter term is to the power three, its decrease
more than offset the increase of 1/k3 • The inequality of equation (5.92) with
its multiple amplitudes a thus reverted to the inequality of equation (5.94), the
condition of a single amplitude a with regular oscillations.
Jumps
The erratic changes or jumps in the response amplitudes for the nonlinear
model described by equation (5.81) can be predicted from the IAI vs. n behavior
of equation (5.88). This behavior is shown in Figure 5.9 for two values of K.
Several observations can be made about these curves. First, a..s K is decreased at
constant n, as from the solid curves to the dashed curves, the response behavior
approaches that of the undamped, linear system response IH(w)l of Figure 5.4.
Second, as K is increased at constant n, the response curves lean more to the
right. Third, multiple amplitudes exist for values of Q > flm where flm is
the value of Q at the knee of each curve. Fourth, if linear damping had been
included in the nonlinear model, equation (5.81), then that effect would have
shown up as the amplitude-limiting, small dashed line shown as the upper knee
for K 0.001, around the point labeled 2' (Stoker, 1963).
Consider an example based on Figure 5.9 in which a particular set of con-
ditions can lead to a jump in response amplitude. For a wave load of constant
magnitude Po and for K 0.001, suppose that the wave excitation frequency
initially given by w = 1.2 wo is decreased very slowly. The response amplitude
follows the path from point 1 to point 2 on the lower knee of the solid curve,
at which point the amplitude must almost double in magnitude by jumping to
RESPONSE OF NONLINEAR STRUCTURES 127
30
11
25
11
~o
K 0.0002
11
K= 0.001
,.,¿-
~ 20
11
Light dampíng
/ 1
3..
ul'
1 1
Cl
¡:: 15
1 1
:::3
¡:¡...
::E
1
~
¡.,:¡
[/j
1
z
o
tl..
[/j
10 1
¡.,:¡
e:: 1
1
/
5 ,¿
0~----~~----~------~------_J------~~----~.
0.7 0.8 0.9 1.1 1.2 1.3
FORCING FREQUENCY, Q=rotro 0
Figure 5.9 Response amplitudes and jurnps for excursions in excitation frequency at
constant Po.
Again refer to Figure 5.9 and consider a second example for which Po is
constant and K = 0.001. Assume light damping with an initial excitation
frequency w 0.8wo. Then w is increased very slowly from point 1' along this
solid curve to point 2' at the upper knee of the damped amplitude curve. If
there is a subsequent increase in w, this must be accompanied by a decrease in
amplitude of about a factor of ten, or a downward jump from point 2' to 3'.
Then smooth amplitude behavior persists along the same curve to point 4'.
Because of system inertia it takes a finite time for a jump in amplitude to
take place. If the total excursion time for the excitation frequency is much
larger than 2r.1wo, then the time required for a jump is of the order of 2rr1wo.
This is demonstrated in the problem of the oscillating buoy, which is discussed
in Section 5.6.
128 SINGLE DEGREE OF FREEDOM STRUCTURES
8~----~------~------~------~-----.------·
2
Po 1k, = 0.6 ft
2
- - - p, /k,=0.4ft
o
/
/
__.... /
-
0.8 0.9
ii>O
OL-----~~----~------~-------L-------L------~
0.7
ii<O
1.1 1.2
--
1.3
FORCING FREQUENCY, !.1 = ro/ro 0
real roots of a for 0.7:::; wjw 0 :::; 1.3. As in Figure 5.9, the response amplitude is
negative or opposite in sign to p 0 to the right of the Po = O line and is positive to
the left of this line. The two branches of each Po/ k1 curve were connected by the
broken lines to show the approximate effects of light damping, or ( = 0.05. The
estimated peak amplitude of each was chosen as ten times the respective static
response, the same amplification calculated for its linear counterpart, Figure
5.4. Damping is discussed in detail by Stoker (1963).
With the damping shown and with a value of wjw 0 = 1.15 for the lower
load ratio of p 0 jk 1 = 0.4 ft, the system oscillates initially with an amplitude
of a = 1.3 ft at point l. For this lower load ratio and with system damping,
the response amplitude is single-valued at this frequency. As the load ratio is
increased to p 0 jk 1 = 0.6 ft, the response amplitude increases toa = 2.85 at the
knee of the solid curve, or point 2. However, a response amplitude of a = 5.4
ft could also exist on the upper branch of the solid curve. It follows that there
can be an upward jump in amplitude response between points 2 and 3 as the
excitation load ratio is increased. Furthermore, there can be a downward jump
from points 3 to 2 and then a smooth transition to point 1 if the excitation load
ratio were gradually decreased to its initially lower value. There is of course
the possibility of jump behavior at load levels between those given, not only for
w / w 0 = 1.15 but for larger values of the forcing frequency as well. The two load
level ratios chosen, however, sufficiently illustrate the combination of system
parameters that can lead to erratic jump responses in this nonlinear system.
Subharmonics
The preceding studies for the nonlinear structure modeled by equation (5.81)
showed high amplitude responses or resonance for excitation frequencies w near
wo. This is not too surprising since resonance for w near wo is well-known in its
linear system counterpart (k3 =O) and was shown in Figure 5.4. In nonlinear
systems, however, resonance near w0 may sometimes occur when w is not near
wo. For instance, if high amplitude responses also exist near wo = w jn where
n = 2, 3, ... , then the response is defined as a subharrnonic of order 1/n. Such
responses do not occur in linear systems, but exist under special conditions in
nonlinear systems such as cable-stayed offshore structures in seas with regular,
harmonic wave components.
The subharmonic response most commonly observed is for n = 3, or the
one-third subharmonic (Wilson and Awadalla, 1973). There are several alterna-
tive, classical methods that can be used to show the existence of this and other
subharmonics (Stoker, 1963). For instance for w = 3w 8 , Cunningham (1964)
applied first-order perturbation theory to equation (5.81) in the same, straight-
forward manner illustrated previously in this chapter. Cunningham's results are
summarized as follows. The necessary relationship between the subharmonic re-
sponse of amplitude a8 and its frequency of oscillation w8 (close to wo), when
the system of equation (5.81) is excited by the force p 1 (t) =Pocos 3w 8 t, is
(5.95)
130 SINGLE DEGREE OF FREEDOM STRUCTURES
Define the following three nondimensional parameters which will aid in inves-
tigating this last result relating the one-third subharmonic response amplitude
to the excitation frequency w = 3w s.
f2 S-- Ws.
l (5.96)
wo
Here, the frequency parameter f2s is of order unity and the stiffness parameter
Ks is arbitrarily small (of order e:). When these nondimensional parameters are
used with equation (5.95), the result is
(5.97)
CO = 3COs
Figure 5.11 Effect of nonlinear stiffness on the amplitude and frequency of the
one-third subharmonic response.
Such results are shmvn in Figure 5.11. For accuracy and consistency with fust-
order perturbation theory, neither Ks nor O.. should exceed the values shmvn.
The knees in these curves indicate that for each K, there is a mínimum fre-
quency below which the subharmonic cannot exist. This mínimum, derived
from equation (5.97) using d0 8 jdA, = O, yields the following equation from
which the critica! points (roots) Os Om can be computed:
6 4
Om - Om -- 87 Ks O (5.98)
An approximate formula for the critica! frequencies at the knee of each curve
of constant Ks can be derived by substituting Om = 1 + r:: in equation (5.98),
expanding the frequency terms, and ignoring terms in r::2 , , . . . . The result is
(5.100)
This is an important result because it shows that, for hardening restraints where
k3 > O, the mínimum frequency for a one-third subharmonic to exist is just a
little higher than wo.
Other important results by Cunningham (1964) are summarized. First, the
subharmonic motion is stable only in the portions of the curves of Figure 5.11
that have positive slopes. No sustained subharmonic motion exists in the region
below the broken line shown in this figure. Second, the effect of including linear
damping of the form c1 v in the nonlinear model, equation (5.81), is to decrease
As at constant Ks and Os, and to eliminate subharmonic motion at sorne cutoff
value of Os Oc > Om > l. The value of Oc decreases as the damping
increases. Third, for a subharmonic of arder 1/n to exist for equation (5.81)
where the excitation frequency is nw0 , the highest power of v in the restraíning
force polynominal must be at least of order n. Thus, a one-fifth subharmonic
response cannot exist for equation (5.81), but a one-half subharmonic response
can.
In summary, it is clear from these classical results for a nonlinear system that
a resonance response amplitude As of frequency Ws (near to but a little greater
than wo) exists for excitation frequencies w = 3w 8 • Experimental evidence for
moored ship responses that support this analysis was presented by O'Brien
and Muga (1964) and by Wilson and Awadalla (1973). This evidence will be
discussed in Chapter 10.
A type of open sea mooring system used to load crude oil into tankers is the
single anchor leg mooring (SALM) buoy. One of a wide variety of such buoy
designs is shown in Figure 5.12: a buoy with a long, upright cylinder stabilized
by cables to sea floor anchors and held in place by a short bottom chain to the
pipeline end manifold (PLEM). This PLEM is anchored to the seafloor with
pin piles. \Vhen the oil is not flowing through the flexible hoses to the ship,
the tanks within the buoy are full, and the buoy has just sufficient buoyancy
to keep it afloat. Then the tension in the bottom connecting chain is very
slight. It is under these conditions that the rotational motion of the buoy is
now analyzed. The chosen excitation is that of a single, harmonic, linear water
wave. Responses to a range of excitation frequencies are sought.
The mathematical model of the upright buoy with its geometry and its
loading are defined in Figure 5.13a. The buoy is treated as a rigid body rotating
at angle () about a frictionless pin at its base. The free body sketch of the buoy
shown in Figure 5.13b identifies: the net cable tension force Fe at angle cjJ with
the horizontal; the damping force Fd acting at the mass center G; and the
total wave load p 1 ( t). The center of buoyancy B is assumed to be coincident
with G. The buoy weight and its buoyant force are equal and opposite, and the
net restoring moment due to these forces is zero. The horizontal wave particle
velocity and acceleration are assumed to be much larger than those for the buoy.
When equation (2.3), the equation of plane motion for rotation of a rigid body
about a fixed point O, is applied to the free body sketch of Figure 5.13c, the
result for small buoy rotations () is
(5.101)
NONLINEAR RESPONSES FOR A SALM BUOY 133
M } - - FLEXIBLE
HOSE
~-l:f-SHORT CHAIN
Here, the externa! moments and the virtual mass moment of inertia Jo are
those with respect to the base point O. The vertical dimensions he, hd, and hw
locate the respective forces fd, and p 1 (t). These latter three forces are now
evaluated.
The cable tension force is based on a buoy with four identical, symmetrically
arranged restraining cables where the wave force is in-line with two of these
cables. For small in-plane rotations, the contributions of the two out-of-plane
cables to the buoy's restoring force are negligible. The cables have nearly the
same specific gravity as that of sea water so that their dead weight effects on
the buoy are neglígible. Each cable is a three-strand polypropylene line that is
approximately straight and under a pretension force at () = O. The increment
in the tension force Fe required to stretch this line by an amount 8 while its
opposite line tends to go slack, is given in the form
(5.102)
where ao and bo are material constants derived from experiments. If the buoy
diameter is much smaller than the cable length, it follows from the geometry of
the single stretched cable shown in Figure 5.13c that 8 is approximately
(5.103)
134 SINGLE DEGREE OF FREEDOM STRUCTURES
SLACK
CABLE~
....
Thus, if the variables on the right side of the latter equation are known, the
cable force can be computed from equation (5.102).
The damping force Fd is modeled as linear-viscous and proportional to the
horizontal buoy velocity hdil at its mass center. That is
(5.104)
for which the corresponding natural frequency for the undamped system is
wo= (5.106)
NONLINEAR RESPONSES FOR A SALM BUOY 135
The compatible value of the drag coefficient in terms of the damping factor and
the other system constants is
(5.109)
where its location from the still water line is
1- cosh kd
(5.110)
k sinh kd
It is noted that z is negative because it is below the still water line, and the
coordinate z is measured positive upward from the still water line.
When equations (5.102)-(5.104), (5.108), and (5.109) are substituted into the
original equation of motion (5.101), the result gives the final nonlinear modelas
(5.111)
In this model, the absolute val.ue sign was used on the even-order restoring force
terms to preserve the proper sign of that force, or to assure that the restoring
force is always antisymmetric about () = O. For a fixed wave frequency and water
depth, k is calculated from equation (3.16), and z is calculated from equation
136 SINGLE DEGREE OF FREEDOM STRUCTURES
.5
<:!:fl
.,.,_"'::
N
a.; 4
o
¡...
¡...
<
U.l
Cl
;::;¡
¡...
:J
¡:¡...
:E
<
>-
o
;::;¡
m
o roo 5 8.7 10 15
WA VE FORCING FREQUENCY ro, rad!sec
Figure 5.14 Response amplitude of the SALM buoy with harmonic excitation.
6
1 1 1
.5
d
<;¡:,
..:? -
N 4 r-
¡:¡...
o¡...
¡... 2 ¡- -
<
¡...
z
U.l
:::; o
U.l
u
<
...¡
¡:¡...
1;/l
a
-2 -
:;...
o;::;¡
lll -4 -
-6 1 1 1
o 2 4 6 8
TIME t, sec
Figure 5.15 Time history of peak buoy displacement for a simple wave of height of
10 ft and frequency w = 8. 7 rad/sec.
138 SINGLE DEGREE OF FREEDOM STRUCTURES
A typical time history of buoy response is shown in Figure 5.15. This re-
sponse is for a wave excitation frequency of w = 8.7 radfsec, an excitation
frequency that produced three distinct amplitudes (jumps) as shown in Figure
5.14. The time history clearly shows an amplitude variation between the ex-
tremes of 5.4 in. and 3.2 in. However, the oscillation between these extremes is
not so abrupt as the word jump implies because it takes two or three response
cycles to effect an amplitude excursion. Further, the time history shows that al-
though the excitation frequency is constant, the response frequency is not, since
the time between crossings is not constant. This is characteristic of systems
with nonlinear restoring forces.
PROBLEMS
5.1 The cylindrical buoy shown in Figure 5.16 is weighted so that it has
a low center of gravity. Its diameter is 2.5 ft, its weight is 2000 lb, and the
density of the sea water is 64 lbjft3 . Assume that the frictional resistance of
the water is negligible, that CM = O, and that the surface of the water is
relatively undisturbed while the buoy undergoes vertical oscillations. Let v
be its displacement from vertical equilibrium and let the restoring force equal
the weight of the water displaced by the motion v. Using its free body sketch,
derive the equation of motion for the buoy in free oscillations. Then calculate
its natural frequency and the time required for one full cycle of oscillation.
5.2 The spherical buoy shown in Figure 5.17 is a thin steel shell of mean
radius R. It is weighted so that in static equilibrium it floats half out of the
water, where its mass center is a distance h > 3R/8 below its geometric center.
The mass center of the displaced water ( the center of buoyancy) is at h = 3Rj8;
and Jc is the mass moment of inertia for the buoy with respect to its rotational
axis through its mass center. Using its free body sketch, set up a dynamic
model for small angles () of rolling motion. Neglect the frictional resistance of
the water. From this equation of motion, derive an equation for the buoy's
natural frequency in roll. If R 3.5 ft, h = 1.75 ft, and the period of oscillation
is 2 sec, calculate Jc.
PROBLEMS 139
e
P1
Equilibrium
ce~o) 4/ Metacenter
new position B'. Examples of this shift are shown in Figures 2.13. Discuss
qualitatively how this offset can affect the rocking frequency of the structure.
5. 7 For the gravity platform of Example Problem 5.2, suppose that the
ballast density is changed so that the center of mass and center of buoyancy
are coincident. Discuss qualitatively how this coincidence can affect both the
rocking frequency and the dynamic stability of this structure.
5.8 Suppose that the only motion for the gravity platform shown in Figure
5.2 is horizontal sliding, v = v(t).
(a) From the free body sketch of this structure, derive the equation of motion
for free, undamped horizontal sliding. Model the soil foundation stiffness k1
according to equation (2.76).
(b) Assuming harmonic motion for the horizontal displacement in the form
v vo sin wot, derive an explicit expression for wo similar in form to equation
(5.11).
(e) Based on the soil and structural parameters given in Table 5.1, calculate
wo and its corresponding period To for this platform. Compare your frequencies
for both the lower bmmd and upper bound values for Gs given in Table 5.1
with the respective natural frequencies for rocking motion computed in Example
Problem 5.2.
(d) If both rocking and sliding motion were taking place simultaneously,
which of these two vibration modes do you think would domina te the structure's
free vibrations? Explain.
5.9 Your task is to design, construct, and perform free vibration experi-
ments on a desk-top model of a three-legged or four-legged jackup rig. The legs
should be securely fixed to a salid base, and can be either hinged or fixed at the
deck level.
(a) Design your model so that its two lateral periods offree vibration To, one
in air and one in water, are suffi.ciently long to be measured accurately with a
stop watch. Use any theoretical method you wish to predict To for both cases.
(b) Construct your design.
(e) Perform free vibration experiments for the model in air and then with its
legs mostly submerged in water. Measure To in both cases asan average value
over severa! oscillations.
(d) Compare your measured periods for the two types of experiments, and
compare each to the theoretical values you calculated previously. Explain rea-
sons for any discrepancies between your measured and predicted results.
5.10 Based on the mnnerical values given in Example Problem 5.5 and
the perturbation solution given by equation (5.56), plot the free vibration dis-
placement v(t) as a function of wt for a ship in sway motion. Superimpose on
this plot the harmonic displacement for its linear system counterpart (k3 =O).
Discuss the distortion in the free vibration oscillatíons due to nonlinear cable
restraints.
5.11 Derive the differential equation for the horizontal sliding motion only
of the gravity platform shown in Figure 5.2. The structural geornetry is given in
PROBLEMS 141
the figure, other system parameters are summarized in Table 5.2, and the equa-
tions goveming its soil foundation stiffness and damping are given by equations
(2.76) and (2.77).
(a) Calculate an upper and lower bound for To, the undamped period of free
horizontal vibration.
(b) Deduce a damping factor ( for the upper and lower bound values of To.
(e) Assume horizontal ground excitation by the El Centro earthquake whose
pseudovelocity response is given by Figure 5.8. Based on the numerical V'alues
of To and ( just calculated, estímate the bounds for the peak values of hori-
zontal displacement, the horizontal shear load on the soil foundation, and the
overturning moment. You may smooth the pseudovelocity curves to estímate
these bounds.
(d) The preceding response calculations do not include the damping effects of
the surrounding water. Deduce whether sueh additional damping would increase
or decrease the system responses to this same earthquake excitation.
5.12 Use first arder perturbation theory to derive the amplitude function
g(ii) given by equation (5.85). lJsing this result, deduce equation (5.86).
5.13 Include linear, viscous damping in equation (5.82) and use first order
perturbation theory to derive an algebraic result analogous to equation (5.86).
Then predict the peak response amplitude for the two curves of Figure 5.10, for
( 0.05.
5.14 Predict the excitation frequency w = Wm below which there will be
no amplitude jumps in the system modeled by equation (5.82). Do this in two
ways: first by differentiating equation (5.88) to find a mínimum O; and then by
applying the rule of roots where R =O in equation (5.91).
5.15 Cse first arder perturbation theory to derive the one-third subhar-
monic relationship given by equation (5.95).
5.16 Derive equation (5.97) from (5.95). Then show that the frequency
ratio Om above which a one-third subharmonic response can exist is given by
the lowest real root of equation (5.98).
5.17 For the nonlinear system modeled by equation (5.82), the mínimum
frequency ratio Om below which a one-third subharmonic response cannot ex-
ist is predicted from equation (5.98) or approximately from equation (5.100).
Whicb of these relationships predicts the lower value of Om? Choose sorne
realistic numerical values of Ks to validate your answer.
5.18 Assume a solution to the linear equation (5.105) of the form
O= Ooe-(wot cos Wdt
where Oo is a constant and Wd is given by equation (5.71). Use this solution to
derive the following approximate expression for the damping ratio ( :
1 fh
(c::::-ln-
27r 02
142 SINGLE DEGREE OF FREEDOM STRUCTURES
Here, fh and e2 are any two consecutive peak ampitudes during free vibration.
5.19 Carry out the necessary integrations in equations (5.108) and (5.110)
to verify the expressions given for the buoy wave loading p 1 (t) and its location
z.
5.20 Starting with equation (5.111), calculate an expression for the quasi-
static response est of this buoy, or the value of e for w ~ l. For the quasi-static
load, use the amplitude of the wave loading given by the right side of this
equation. Delete all terms involving iJ and B, and delete all terms containing
e with powers greater than one. Then solve the result for e = est. Use the
numerical values of Table 5.3 to calculate est, and compare your result with
that obtained from Figure 5.14 for a small value of w (w = 0.05 for instance). If
these two results do not coincide, explain possible reasons for the discrepancy.
REFERENCES
Clough, R. W., and Penzien, J., Dynamics of Structures, second ed., McGraw-Hill,
New York, 1993.
Den Hartog, J. P., Mechanical Vibrations, third ed., McGraw-Hill, New York, 1947.
Forsythe, G. E., Malcolm, M. A., and Moler, C. B., Computer Methods for Mathe-
matical Computation, Prentice-Hall, Englewood Cliffs, NJ, 1977.
Gould, P. L., and Abu-Sitta, S. H., Dynamic Response of Structures to Wind and
Earthquake Loading, Wiley, New York, 1980.
Nataraja, R., and Kirk, C. L., Dynamic Response of a Gravity Platform under Ran-
dom Wave Forces, OTC-2904, Proceedings of the Offshore Technology Confer-
ence, 1977.
O'Brien, J. T., and Muga, B. J., Sea Tests on a Spread-Moored Landing Craft,
Proceedings of the Eighth Conference on Coastal Engineering, Lisbon, Portugal,
1964.
Paz, M., Structural Dynamics: Theory and Computation, Van Nostrand Reinhold,
New York, 1980.
Rosenbach, J. B., Whitman, E. A., Meserve, B. E., and Whitman, P. M., College
Algebra, fourth ed., Ginn, Lexington, MA, 1958.
Wilson, B. \V., Ship Responses to Range Action in Harbor Basins, ASCE Transac-
tíons 116, 1951.
Wilson, J. F., and Orgill, G., Optimal Cable Configurations for Passive Dynamic
Control of Compliant Towers, Journal of Dynamic Systems, Measurement, and
Control 106, 1984.
Statistical Descriptions
of Offshore Waves
Bruce J. Muga
INTRODUCTION TO WAVE SPECTRA 145
8r----T----~----~----r----,
S 6
~ 4
~ 2
g¡:.r.;¡ OH-~~~~~+-~-4~-4--~-+--~~
::¡:: -2
~-4-6
-8~--------~----~--~~--~
o w ~ ~ w 100
TIME t, seconds
Figure 6.1 A 100 second section of a 17.07 minute record of surface wave height
recorded at Macquaire Island in the Southern Ocean on May 13, 1989 (Young, 1999).
To apply the Nyquist theorem, let fma:x = 1/(2~t) where ~t is the shortest
time between two consecutive crossings of the 1J = O axis. Thus, an appropriate
sampling frequency is fs 2/(2t::J.t) 1/t::J.t, and the corresponding time interval
between sampling of11(t) is simply t::J.t. For the 100-sec section ofthe wave record
shown in Figure 6.1, t::J.t = 2 sec . However, if the whole 17.07- min record were
considered, then a shorter time would be apparent, or t::J.t 1 sec (Young, 1999).
Consider a Fourier series representation of this whole 17.07 min or 1024 sec
wave record. To reproduce this record, one would need N = 1024 sample points
1J. The Fourier series in this case would oonsist of 1024 terms involving a sum
of sines and cosines, or
N/2 N/2
1J(t) = 2 """ 21rnt """ . 21rnt
L.,; an COS --dt + 2 L.,; bn sm - - (6.1)
n=l TQ n=l To
where To is the 1024-sec period and the mean value of 1J(t) is zero (ao = 0).
The coeffi.cients in equation (6.1) are
an = 1 loro 2n1rt
1J(t) cos - - dt
ro o ro
bn
1
-
loro 1J(t) sin--
21rnt
dt (6.2)
TQ O TO
The total wave energy is proportional to the average of the squares of 1J(t),
and this energy is equal to the sum of the energy content of each of the individual
wave components. This is shown with Parseval's theorem, which states that for
146 STATISTICAL DESCRlPTIONS OF OFFSHORE WAVES
any periodic function of time represented by equation (6.1) where the mean is
zero (ao O), then the variance of r¡( t) is
-
1 i'To [r¡(t)] dt
2
n/2
= 2 L(a~ + b~) =a~ (6.3)
TQ O n=l
(6.4)
It is observed that this spectral function is the resolution of the variance of r¡(t)
into its frequency components. That is, the value of S.,(wn) at each discrete
frequency Wn can be approximated by the sum of the Fourier coefficients, or
(6.5a)
where
2mr 27r
dw=- (6.5b)
To To
Here, dw is the spacing between adjacent harmonics. It is emphasized that
the sampling interval dt must satisfy the Nyquist sampling theorem to avoid
falsely distorting the Fourier coefficients. In practice, an experimentally- based
discrete energy spectrum is usually fit to a continuous function by smoothing.
Examples of smoothed spectra are discussed later in this chapter.
The character of the function S.,(w) depends on the surface wave elevation
record. For example, S 11 (w) for unidirectional swell is a narrow-banded function
centered about the dominant swell frequency. On the other hand, sea waves
in general are most often characterized by a broader-banded spectral function
than that for swell waves.
The interest in wave spectra arises from the need to select a representative
design-loading history acting on the proposed offshore structure. The design
wave spectrum, usually furnished by a specialist in hindcasting or forecasting,
is chosen to represent the worst possible combination of waves which would
lead to the highest loading for the structure. To complicate matters, there
may be many different surface water wave elevation histories characterized by
essentially the same spectrum. In other words, a one-to-one correspondence
between the surface water wave elevation record and the spectrum does not
generally exist. However, all records that yield identical spectra do have the
same statistical properties, although details in the records may vary widely.
The suggested procedure, then, is to select a number of water wave elevation
records corresponding to the design wave spectrum and to analyze the proposed
offfshore structure under these alternative conditions.
CONCEPT OF THE SIGNIFICANT WAVE 147
lt is important that the mathematical model used to represent the sea sur-
face (the time history of wave height) should conform as dosely as possible to
the actual waves. In practice, there are two ways to accomplish this. First,
a purely deterministic method can be used in which the sea surface records
are fitted to a harmonic or Fourier series. This representation is unaffected by
nonlinear wave effects. Second, a statistical method can be used in which the
sea surface records are fitted to a spectral function in the frequency domain.
It is emphasized that the second method is based on the concept of superpo-
sition and is adequate only for linear seas. Various efforts have been made to
include nonlinear phenomena in spectral descriptions but these methods are
not widely utilized. For purposes of structural design, however, linear models
usually provide adequate descriptions of the sea surface.
For further discussions of offshore waves as random processes, the reader is
referred to the works of Longuet-Higgins (1963), Munk (1950), Kinsman (1965),
Pierson and Neumann (1966), and the overview of Young (1999).
The concept of the significant wave was fust published by Sverdrup and Munk
(1947), and later discussed by Wiegel (1949), Bretschneider (1959), and Kins-
man (1965). The significant wave height is a useful parameter for describing an
irregular sea surface and its wave spectra.
What is the significant wave height? The significant wave height is not an
identifiable wave form that propagates like a physical wave as in the classical
wave theories. The significant wave height, H 8 , is defined as the arithmetic
average of the highest one-third of the waves in a wave record. Its associated
statistical parameter, the significant period T8 , is defined as the average period
of the highest one-third of the waves in this wave record. lt is noted that the
symbols H 1¡3 and T 1¡3 are also commonly used to denote Hs and T 8 , respec-
tively.
The significant wave height is an important parameter in statistical analysis
of wave mechanics for a number of reasons. The statistical distribution of wave
heights and most energy spectrum analyses are related to the significant height.
In fact, the major portian of the wave energy of the spectrum surrounds the
significant wave height. Thus, it has been the practice historically to report sea
conditions in the form of significant wave heights. Also, the effects that irregular
seas have on many types of fixed and floating objects and onshore processes
such as littoral transport, have been related to significant wave heights, with an
accuracy sufficient for these engineering applications.
In summary, the irregular sea can be described in an abbreviated format by
two parameters: the significant wave height and the significant period. However,
designers of offshore systems sometimes need to know other statistical features
of the sea, such as the maximum wave height or the maximum water surface
elevation. This led to a number of studies that attempted to establish further
statistical correlations for offshore waves. A brief summary of these correlations
is now presented.
148 STATISTICAL DESCRIPTIONS OF OFFSHORE WAVES
p(H) (6.6)
in which Hrms is the root-mean-square (rms) wave height of a given record. The
square of this latter quantity is defined as
(6.7)
As the notation Jlf2 implies, Hrms is formed by squaring the height of each
wave in a given record, taking the arithmetic average of these quantities, and
then taking the square root of the result. The mean or average wave height is
defined by
Based on the Rayleigh distribution, the average wave height, Ho, the signif-
icant wave height, H 8 , and the 1/10 highest wave height, H 1 ¡ 10 , are deduced
as
These three wave heights are depicted in Figure 6.2, which is a plot of the
Rayleigh distribution given by equation (6.6). Also shown is the height of the
most probable wave, which is at the peak of the curve.
Further, the most probable maximum wave height, Hrna:x, depends on the
duration of the storm or length of the wave record. The following relationship
is often used to approxinlate this maximum wave height:
Here N is the number of waves in the record. When N is not known, a reasonable
approxinlation to this maximum v,rave height is
For very severe storm waves or for waves in very shallow water near the breaking
point, equations (6.10) and (6.11) should be used with caution.
CONCEPT OF THE SIGNIFICANT WAVE 149
J
§
0.8
"""
~ 0.6
~
>-
!::: 0.4
i=l
~
p..
o
o 0.5 1.5 2 2.5 3
WAVE HEIGHT RATIO, HIH,.,.,
For typical irregular sea conditions where the average period ranges from 4
to 10 sec, studies have shown that T 8 is nearly equal to the average period,
For longer average wave periods (» 10 sec), it has been found that the average
period is only 75 percent of the significant period. Based on a comprehensive
study of wave conditions in many locations, the International Ship Structures
Congress (ISSC) concludes that the average period can be taken as 90 percent
of the significant period (Price and Bishop, 1974). It is emphasized that the
average period as defined here does not correspond to the period of the average
wave height.
Wave Height- Wave Spectrum Relationships
A historical advance in the description of irregular ocean surface waves was
accomplished by Pierson (1952), who merged key concepts from classical me-
chanics and the theory of stochastic processes with the energy spectrum in order
to predict the behavior of offshore waves (Kinsman, 1965). In its simplest form,
the energy spectrum allocates the amount of energy of the sea surface according
to frequency. As shown in Chapter 3, a small-amplitude sinusoidal wave has
the form
1 1
E= -pgA2 = -pgH 2
(6.14)
2 8
where H 2A) is the wave height measured from crest to trough.
One of the fundamental premises of the spectral approach is that irregular
waves are the result of the superposition of an infinite number of simple sine
waves of small amplitudes that have a contínuous frequency distribution. This
process can be approximated with a finite number of small-amplitude sine waves
CONCEPT OF THE SIGNIFICANT WAVE 151
having discrete frequencies. Under these conditions the mean total wave energ:y
per unit surface area is given by
E= !8 pg( Hf + H~ + H§ + · · · + H~ + · · · ) (6.15)
H¡
ID¡ ID2 ID3 ID4 IDs ID6 ID7 IDa ID9 ID¡o
WA VE FREQUENCY, ID
(6.16)
152 STATISTICAL DESCRIPTIONS OF OFFSHORE WAVES
There are significant and sometimes subtle differences among the analytical
descriptions of wave spectra as they appear in the open literature. Differences
exist in terminology, notation, and even in the basic definition of Sr¡(w). These
alternative descriptions are depicted in the abscissa and ordinates of Figure 6.4.
"',..._
:¡;:
M '('¡'
N':@ •
i:ii::J ;:,::::>
~co::
~ :@ti~~
:::>¡;¡¡~¡;¡¡
c.::;; ......
~
..J
¡,....,:.~"'
1;l..J~p:¡
.,.,< ..J
~
"':¡: CQ
o;¡p:¡f;l:l§
o o""'
¡;¡¡
;:¡¡::"'
............
"'""' ::J:l:C:C
o..c..OO
~~¡:¡¡¡¡¡
<<XX
Alternativa Definitions
For the abscissa of the energy spectrum, the cyclic frequency f = 1/T (Hz),
the angular frequency w 21rjT (radjsec), and the apparent period Tare used.
Experimentalists most often use f to present data; and those who expound
theory use w.
The ordinate is referred to as power spectral density, or simply spectral den-
sity. The average energy of the wave system is related to the sum of the square
of the wave component heights H 2 or amplitudes A 2 (H/2) 2 • The follow-
ing terminology was suggested by 1vlichel (1967) for identifying the ordinate in
terms of the significant wave height.
l. Amplitude spectrum: The following forms, used by Pierson (1952) and
Pierson et al. (1955) define the spectral density as a function of the square of
DESCRIPTIONS OF WAVE ENERGY SPECTRA 153
the wave amplitude, for which the significant wave height is related to (area)I
under the spectral curve, or
When equations (6.17) and (6.18) are compared, the results are: (area)I =
2(variance) and (area)z = variance.
3. Height spectrum: Other investigators choose to use wave heights, for
which the area (area)3 under the spectral curve is four times the area obtained
when wave amplitudes are used. This is because (wave height) 2 = 4(wave
amplitude) 2 . Therefore, the spectral density is a function of (wave height) 2 , and
the relationships of the significant wave height to the areas under the three types
of spectral curves are
Hs = 2.83vf2(variance) = 2.83vf(area)I
The following precautions should be used when working with wave energy
spectrum prepared by others. First, establish the units of the spectral density.
Is this in Traditional English units of ft 2 -secjrad or ft 2 /Hz, or is it in SI units
of m 2 ·s/rad or m 2 /Hz? Second, establish the frequency units of the abscissa.
Is this in units of Hz for cyclic frequency f, or in units of radjsec for circular
frequency w? Third, determine the specific formulation on which the ordinate
is based. Sometimes this last step can be accomplished by simple examination
of the symbolic notation; or the author of the data can be questioned. In any
case, the Rayleigh distribution coefficient used to predict significant wave height
(or wave amplitude) from the energy spectrum should be consistent with the
nomenclature.
154 STATISTICAL DESCRJPTIONS OF OFFSHORE WAVES
Empirical Forros
Particular forms of deepwater wave spectra in common use are those of
Pierson and Moskowitz (1964) and Bretschneider {1959), which are based on
theory and require sufficient data to fit the constants. The spectra of Pierson-
Moskowitz are termed wind-speed spectra since wind speed is included directly
in the spectral density function. The spectra of Bretschneider are termed height-
period spectra since significant height H 8 and the significant or mean apparent
period Tm = 2Tr/wm are used directly in the spectral density function. It is em-
phasized that in deducing these spectra, use was made of the distribution func-
tions originally derived theoretically by Longuet-Higgins (1952), and supported
by the empirical relations based on the wave data of Putz (1952). These deep-
water spectra have the form of equation (6.15) in which the constants (Ao, B)
differ, depending on whether the height-double or amplitude-half spectrum as-
sumption is made in reducing the data.
Another spectrum in common use is JONSWAP, which investigators have
found appropriate for the design of offshore structures in the North Sea. Un-
like the spectra in the form of equation (6.15), which are referred toas fetch-
unlimited spectra, the JONSWAP spectrum includes the fetch in its formulation,
and is called a fetch-limited spectrum. It is noted that the term fetch, denoted
by X, is the distance from the shoreline of the wave field under consideration. A
fetch-limited wave spectrum is one based on a wave field at distance X that has
reached a steady state or time-independent condition. Such a condition can oc-
cur in deep water if the wind has blown for a sufficient length of time, out to sea,
and in a direction perpendicular to a regular shoreline. These and other factors
that affect the fetch, together with experimental measures of X appropriate for
fetch-limited spectra, are thoroughly discussed by Young (1999). Useful forros
for both fetch-unlimited and fetch-limited spectra are surnmarized.
Wind-Speed Spectra. The Pierson-Moskowitz fetch-unlimited spectra has
the general form
g2 4
S.,(w) 0.00815 e-Bfw (6.21)
w
where g is the acceleration due to gravity. When the wind speed is known, then
(6.22)
where V is the wind speed at a height of 19.5 m above the still water level.
Any consistent set of units for the quantities g, V, and w are appropriate for
equations (6.21) and (6.22), provided that w is expressed in radians per unit
time.
Significant Wave Height-Period Spectra. When the significant wave
height rather than the wind speed is known, the constant B of equation (6.21),
for H 8 in units of m/s, is
(6.23)
DESCRJPTIONS OF WAVE ENERGY SPECTRA 155
The source of the latter relation is the International Towing Tank Conference
(ITTC) (Price and Bishop, 1974), and the resulting spectrum is known as the
ITTC spectra. Shown in Table 6.2 are the ITTC recommended values for
significant wave height and average wave period for severa! wind speeds.
w;,
S r¡ (w ) = 1.25
4 w5
H2
se
-1.25(wm.fw) 4
(6.24)
in which Wm is the frequency at the maximum of the spectrum, where the cor-
responding period is Tm = 27r/Wm. Listed in Table 6.3 are empirical equations
for Wm in terms of the significant wave height H 8 and its probability of occur-
rence. Also listed in that table are values for Wm and T m for the special case of
Hs = 12.2 m, a case considered later in Example Problem 6.1.
It is emphasized that the spectra expressed by equations (6.21) through
(6.24) are amplitude half-spectra and that the square root ofthe area under each
spectral diagram must be multiplied by 4 in order to obtain the corresponding
significant wave height. That is, H 8 = 4vf(area)2.
Table 6.3 Spectral Model Data (Bretschneider, 1959 and Ochi, 1978)
Equation for For H 8 = 12.2 m: Probability
Wm, radjs Wm, radjs Tm, s of Occurrence
0.048 (8.75- ln H 8 ) 0.30 20.9 0.0500
0.054 (8.44-ln Hs) 0.32 19.6 0.0500
0.061 (8.07-ln Hs) 0.34 18.5 0.0875
0.069 (7.77-ln H 8 ) 0.36 17.3 0.1875
0.079 (7.63-ln H 8 ) 0.41 15.5 0.2500
0.099 (6.87-ln Hs) 0.43 14.5 0.1875
0.111 (6.67-ln Hs) 0.46 13.6 0.0875
0.119 (6.65-ln Hs) 0.49 12.7 0.0500
0.134 (6.41-ln Hs) 0.52 12.0 0.0500
156 STATISTICAL DESCRJPTIONS OF OFFSHORE WAVES
There are two general observations concerning signiflcant wave height statis-
tics. First, different wave time histories can have approximately the same sig-
nificant wave height, but still have widely different spectral properties. Ochi
and Whalen {1980) and Ochi {1981) offered a rational explanation and experi-
mental documentation for such observations. Second, extensive global data for
signiflcant wave height based on nine years of satellite measurements taken in
the 1990s were presented as contour maps by Young (1999). These data show
mean monthly values of H 8 , values which could be expected to be exceeded
10 percent, 20 percent, ... ,90 percent of the time. The highest value of Hs
reported was 6 m with a probability of exceedence of 10 percent. Since confined
events such as hurricanes affect such monthly averages very little, these satellite
data have very limited use in the design of offshore structures.
Fetch-Limited Spectra. The JONSWAP spectra has the general form
(6.25)
where
'Y= 3.3 for mean of selected JONSWAP data
'Y = 7.0 for a very peaked spectrum
O' 0.07 for w :5 Wm
O" = 0.09 for w > Wm
Wm = 21r(3.5)(g/V)(X)~ 0 - 33 peak frequency
a 0.076(X)~ -0 22
or a= 0.0081
2
X= gX/V
X fetch length; V = wind speed
50
PROBABILITIES OF
OC CURRENCE 500
0.05
40
400
'O
10 100
FREQUENCY, rorad/sec
The selection of a design significant wave height can be done only after a
hindcast study has bet:>..n completed for the specific site. This hindcast study
should yield both the significant wave heights corresponding to the selected
storm recurrence intervals and the associated probability functions governing
the modal frequencies Wm· With this information and with a suitable spectral
SYNTHESIS OF TIME HISTORIES FROM SPECTRA 159
model, the farnily of wave spectra can be derived. The following example illus-
trates these ideas.
Example Problem 6.1. The hindcast studies for a particular site in the
Gulf of Mexico indicate that H 8 for the 100-year storm is 12.2 m. Assume that
the wave spectral farnily can be described by the two-parameter Bretschneider
(1959) model, equation (6.24), and that the modal frequencies, Wm, the corre-
sponding periods Tm, the probabilities for ocurrence of H 8 , for this family are
those listed is Table 6.3. The results in Table 6.3 were derived by Ochi (1978)
from selected wave data. Note that the sum of the probabilities for occurrence
(or weighting factors) of Hs is equal to unity.
The equation for the most probable spectrum is obtained by substituting
into equation (6.24) the frequency at the maximum of the spectrum: Wm = 0.41
rad/s from Table 6.3, for which the probability of occurrence is the maximum
of 0.25. With Hs = 12.2 m, the spectrum of equation (6.24) becomes
Equation (6.26), which is plotted in Figure 6.5, corresponds to the most probable
spectrum associated with a storm of a 100-year recurrence interval. As such,
this curve is not a design spectrum but a member of the family of spectra to be
considered in design. Other members of the farnily are also shown in Figure 6.5,
with selected frequencies Wm, and the corresponding probabilities of occurrence
as given in Table 6.3, together with their confidence limits.
The ideas illustrated in this example can be extended to more complex spec-
tral formulations such as JONSWAP, which may be more appropriate for other
sites such as the North Sea. The reader is encouraged to consult the most re-
cent references so that the most up-to-date data can be used to generate the
farnily of design spectra. Once these spectra are chosen, they can be employed
for prelirninary dynamic analyses of alternative structural designs, as will be
illustrated in succeeding chapters.
In the analysis of offshore structures, time domain solutions are frequently re-
quired. However, if the wave height excitation is expressed in the form of spec-
tral density, it is necessary to transform this design spectra into an ensemble of
representative time histories. This may be accomplished by utilizing Borgman's
(1969) procedure for wave simulation. This is now presented with slight modi-
fications.
The wave elevation r¡( t) can be represented as
where A 2 (w) is the amplitude spectrum ordinate ande is a random phase angle
picked from a list of random numbers uniformly distributed over the interval
from zero to 21r. The integral is discretized by partitioning the spectrum into
equal portions (areas), instead of taking equidistant points on the frequency
axis. This procedure avoids the presence of periodicities in the resulting time
history.
Consider the partition
(6.28)
where wo is a small, positive value and F is a value beyond which the spectral
ordinate is zero, for all practical purposes. Let
(6.29)
where w~ = kng and the overbar denotes the average value of the parameter.
Let S(wn) represent the cumulative area under the spectral density curve, or
n
S(wn) = L A 2 (wn) .Ó.Wn (6.32)
n=l
Thus
(6.33)
(6.35)
(6.16)
SYNTHESIS OF TIME HISTORIES FROM SPECTRA 161
S(w)
J Ao e-Bfw4 dw
w5
(6.36)
Ao
S(oo) (6.37)
4B
2 Ao (6.38)
a= 4BN
Now determined are the positions of the partition frequencies wo, w1, w2, . .. , w N.
From equation (6.36) where w = F, it follows that
Ao = eBIF4S(F) (6.39)
4B
Because of the equal area partition
(6.40)
It follows that
N B/F =eBfw n
-e
4 4
(6.41)
n
When equation (6.41) is solved for Wn, the result is
The random phase angles en can be generated using one of the many available
codes, but once selected, they are identified with a specific set of frequency
components on a one-to-one basis. On the basis of sorne unpublished work by
this writer, it has been found that an N value of at least 15 to 20 is necessary to
produce time histories with statistical features similar to those predicted from
low-order spectral moments. Time histories of other parameters such as water
particle velocity, acceleration, and pressure can also be derived following the
above approach.
162 STATISTICAL DESCRIPTIONS OF OFFSHORE WAVES
N n-1
LL a2kn' cos( -wn' t +en') cos( -wnt +en)
n=l n'=l
(6.44)
In Chapter 10, this synthesis is illustrated for the case of surface wave exci-
tation of a barge that deploys an OTEC pipeline.
REFERENCES
Barber, N. F., Ocean Waves and Swell, Proceedings of the Institution of Civil Engi-
neers, Marine and Waterways Division, 1950.
Blackman, R. B., and Thkey, J. W., The Measurement of Power Spectra, Dover, New
York, 1959.
Borgma.n, L. E., Risk Criteria, Journal of the Waterways and Harbors Division,
ASCE 89 (WW3), 1963.
Borgman, L. E., Ocean Wave Simula.tion for Engineering Design, Journal of the
Waterways and Harbors Division, ASCE 95 (WW4), 1969.
Borgman, L. E., Confidence Intervals for Ocean Wave Spectra, Proceedings, Thir-
teenth Coastal Engineering Conference, ASCE, Vancouver, B.C., 1972.
Darbyshire, J., The Generation of Waves by Wind, Proceedings of the Royal Society
Ser. A 215, 1952.
Freudenthal, A. M., and Gaither, W. S., Design Criteria for Fixed Offshore Struc-
tures, First Offshore Technology Conference, Houston, Texas, May 1969.
Goda, Y., Random Seas and the Design of Marine Structures, University of Tokyo
Press, Japan, 1985.
Hamada, T., Mitsuyasu, H., and Hose, N., An Experiemental Study of Wind Effect
upon Water Surface, Report of Transportation Technical Research Institute,
Tokyo, June 1953.
REFERENCES 163
Kinsman, B., Wind Waves, Their Generation and Propagation on the Ocean Surface,
Prentice-Hall, Englewood Cliffs, NJ, 1965.
Longuet-Higgins, M. S., and Stewart, R. W., Changes in the Form of Short Gravity
Waves on Long Waves and Tidal Currents, Journal of Fluid Mechanícs, 1961.
Michel, W. H., Sea Spectra Simplífied, Meeting of Gulf Section, Society of Naval
Architects and Marine Engineers, April 1967; also in How to Calculate Wave
Forces and Their Effects, Ocean Industry, May-June 1967.
Munk, W. H., Proposed Uniform Procedure for Observing Waves and Interpreting
Instrument Records, Wave Report 26, Scripps Institute of Oceanography, 1944.
Ochi, M. K., Wave Statistics for the Design of Ships and Ocean Structures, Transac-
tions, Socíety of Naval Architects and Marine Engineers 86, 1978.
Ochi, M. K., Waves for Mooring System Design, Technical Note N1604, Civil En-
gineering Laboratory, Naval Construction Battalion Center, U.S. Navy, Port
Hueneme, CA, March 1981.
Ochi, M. K., and Whalen, J. E., Prediction of the Severest Significant Wave Height,
Proceedings, Sixteenth Coastal Engineering Conference, 1980.
Pierson, W. J., A Unified Mathematical Theory for the Analysís, Propagation and
Refraction of Storm Generated Ocean Surface Waves, Research Division, College
ofEngineering, Department ofMeteorology, New York University, Part 1, March
1, 1952; Part 2, July 1, 1952.
Pierson, W. J., and Moskowitz, L., A Proposed Spectral Form for Fully Developed
Wind Seas Based on the Similarity Theory of Kitaigorodskii, Journal of Geo-
physical Research 69, 1964.
164 STATISTICAL DESCRIPTIONS OF OFFSHORE WAVES
Pierson, W. J., Neumann, G., and James, R. W., Practical Methods for Observing and
Forecasting Ocean Waves by Means of Wave Spectra and Statistics, Publication
No. 601, U.S. Navy Hydrographic Office, 1955. Reprinted 1960.
Price, W. G., and Bishop, R. E. D., Probabalistic Theory of Ship Dynamics, Chapman
and Hall, London, 1974.
Putz, R. R., Statistical Distribution for Ocean Waves, Transactions, American Geo-
physical Uníon 33 (5), 1952.
Putz, R. R., Statistical Analysis of Wave Records, Proceedings, Fourth Coastal En-
gineering Conference, The Engineering Foundation Council on Wave Research,
Berkeley, CA, 1954.
Rayleigh, Lord, On the Resultant of a Large Number of Vibrations of the Same Pitch
and of Arbitrary Phase, Philosophical Magazine 10, 1880.
Rye, H., and Svee, R., Parametric Representation of a Wave-Wind Field, Proceedings,
14th Conference on Coastal Engineering, Honolulu, Hawaii, 1976.
Seiwell, H. R., Sea Surface Roughness Measurements in Theocy and Practice, Annals,
New York Academy of Science 51 (3), 1949.
Sverdrup, H. U., and Munk, W. H., Wínd, Sea and Swell, Theory of Relatíons for
Forecasting, Publication No. 601, Hydrographic Office, U.S. Department of the
Navy, 1947.
Wiegel, R. L., An Analysis of Data from Wave Recorders on the Pacific Coast of the
United States, T'ransactíons, American Geophysical Union 30, 1949.
Young, I. R., Wind Genemted Waves, Elsevier Science Ltd., Oxford, UK, 1999.
7
Statistical Responses for
Single Degree of Freedom
Linear Structures
James F. Wilson
166 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
and for stationary wave excitation. The derived structural responses are analo-
gous to the steady- state time domain responses considered previously. Closed
form solutions are derived by two statistical methods: the classical station-
ary stochastic analysis of Khintchine (1934) and Weiner (1949), and covariance
propagation analysis originally applied by Bryson and Hu (1975) to electrical
control systems. In both methods, the structural response spectrum is predicted
in terms of the wave force spectrum on the structure, and the result leads to
the root-mean-square (rms) structural displacement response and its probabil-
ity of exceedence. These statistical ideas lay the mathematical framework for
Chapters 9 and 10 in which statistical responses are deduced for multi-degree
of freedom structures and for continuous structural elements.
It is assumed a priori that the surface wave height r¡(t), the wave load p 1 (t)
and the structural displacement v(t) or rotation O(t) all have time histories of
the general form shown in Figure 7.2. The variable y = y(t) is used to denote
such a general time history, which is defined as a random, stationary process
of zero mean. A random, stationary process looks essentially the same over a
time interval To, no matter where this interval starts or stops. The interval To
is of sufficient duration to capture the essential character of y(t), and there is
no startup, shutdown, or transient behavior for y(t). A more precise definition
of stationary will be given later in this section. The condition of zero mean is
expressed as
E[y] = -
1 ¡ro/2 y(t)dt =O (7.1)
To -ro/2
y(t)
1'1~ M2 M3 M 4
T0!2 - - - - - f . . . . - - - - - r 0 /2
Figure 7.2 Typical time history of a stationary random variable of zero mean.
A VERACES AND PROBABILITIES 167
From the viewpoint of structural dynamics, equation (7.1) implies that the
fluctuations are about a static, zero mean. For such cases the variance (j y,
defined as the mean square value of y(t), is often used as a measure of these
fluctuations. That is,
(7.2)
This is the probability that y is between a fixed value Yo and (y 0 + !:iy). Specif-
ically, !:iP is defined as the fraction of the total time To that y(t) is in this
bandwidth !:iy. The vertical strips of Figure 7.2 show the typical time incre-
ments /:it 1 , !:it 2, . . . within this bandwidth. Thus
1
!:iP = -(!:ih + !:it2 + ... ) (7.4)
To
. !:iP dP
p(y) = hm(!:iy ___.O) !:iy = dy (7.5)
~~._--4---~~-- ..y
-30'y o 30'y
( y2)
p(y) = py
1
exp -
21ra 2y22
cry
(7.6)
Here cr~ is the variance of the amplitude. Equation (7.7) is known as the
Rayleigh probability density, which was presented in a somewhat different form
for wave height in Chapter 6, or equations (6.6), (6.7), and Figure 6.2.
It is apparent from the above definitions that once the probability density
function and the variance are established for a particular process, the proba-
bility of occurrence within prescribed limits of that process variable (y(t) or its
amplitude a) can then be calculated. These ideas are now illustrated.
= py 1 ¡3<1y
exp
( y2) dy = 0.9974
- - (7.8)
27rcr2 -3ay 2cr y
y
STATIONARY AND ERGODIC HYPOTHESES 169
r3<ta
P[O :S a :S 3aa] Jo p(a)da = 0.989 (7.9)
which is based on the Rayleigh distribution, equation (7.7). It follows that the
probability that any peak chosen at random exceeds 3aa is
This shows that about one peak in 100 exceeds 3aa. Further, it is not diffi-
cult to show that, for both the Gaussian and Rayleigh probability distribution
functions, P is unity when the respective variable y or a extends throughout its
whole range, or
Presented in the following sections of this chapter are two methods for eval-
uating the standard deviation a v or a 9 for a structural displacement coordinate
V or rotational coordinate e. For the first method, it will be shown that solutions
for a v and a 9 can be based on a single measured surface wave height spectrum
Sr¡(w), the wave-to-structure load transfer function G(w), and the structural
response functions h(t) and H(w). For the second method, whích is the newer
of the two and based on control theory for electrical systems, the standard devi-
ation is computed in closed form in terms of a three parameter spectral density
representation of the excitation force. With the standard deviations computed
by either method, appropriate probability density functions can then be used,
as in the examples just presented, to evaluate probabilities of occurrence of
structural displacements and rotations under wave loading.
With the assumption that y(t) is both stationary and ergodic, the statistical
calculations leading to the structural standard deviation in displacement are
simplified enormously. Although these assumptions are rarely checked in prac-
tice, it is nonetheless illuminating to elaborate on these two hypotheses from
an "experimental" viewpoint, as suggested by Muga and Wilson (1970). Todo
this, cut a sample record y(t) such as in Figure 7.2 into J equal charts, each
of time duration ro. Again, To is of suffi.cient duration that it captures the
essential character of y(t). This ensemble of J charts is denoted as
170 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
where i denotes a typical one. Figure 7.4 shows three typical charts of such an
ensemble where each trace begins at time zero. The two broken lines across the
ensemble correspond to arbitrary times t 1 and t2. The stationary hypothesis is
warranted if both of the following criteria are substantially met:
l. Ata fixed time t¡, measure y(t) y(íl(tl) at t t 1 on each chart. The
average y( t) is
(7.12)
The numerical values of equation (7.12) should be about the same for any value
of t1 one chooses, where the value of J is a very large number.
2. Pick a constant time interval r. For a time t 1 and another time t2 , where
(t 2 t 1 ) = r, measure the values of y(il(t 1 ) and y(il(t 2 ) on each chart i and
form the sum
(7.13)
STATIONARY AND ERGODIC HYPOTHESES 171
The numerical values of equation (7.13) should be about the same for any value
of t 1 and t2, as long as (t2 -ti) = T and J is very large.
The ergodic hypothesis is in two parts. The first part of this hypothesis
states: the average value of y as given by equation (7.12), based on a constant
t 1 over an ensemble of charts, is equal to the time average of y over just one
typical chart. This must be true for all values of t 1 . Suppose that the typical
chart, which is representative of the ensemble of charts, is called y(k)(t). Sup-
pose further that the chosen time interval over which y(kl(t) is considered is of
sufficient duration that it represents the behavior of y reasonably well. Then
the time average of y(kl(t) can be approximated as
1 N
N LY(k)(tn) (7.14)
n=l
where y(k) (tn) is the value of y at time tn in the time interval To. The values oftn
are equally spaced, and the total number of measurements is N. If expressions
(7.12) and (7.14) are nearly equal, the first part of the ergodic hypothesis is
approximately satisfied, or
(7.15)
where J and N are both large numbers. In other words, the first part of the
ergodic hypothesis is satisfied if the ensemble average of y for J charts is equal
to the time average of y from a typical chart.
For the second part of the ergodic hypothesis another type of average is
defined by
(7.16)
1 J 1 N
J LY(i)(tl) Y(i)(t2) ~N LY(k)(tn) Y(k)(tn + T) (7.17)
i=l n=l
1 N
Ry(T) e::' N LY(tn) y(tn +T) (7.18)
n=l
which is independent of time but not the time difference T, provided that y(t)
is stationary. More precise!y, Ry (T) is defined by replacing the finite surn in
equation (7.18) by the integral over the duration To where N = To/ Át and
Át ...._. dt in the limit. That is,
Ry(T)
l
lim(To ...._. oo)-
¡ro/2 y(t) y(t + T)dt = E[y(t) y(t + T)] (7.19)
To -To/2
(7.20)
(7.21)
(7.22)
In principie Ry(r) can be calculated from equation (7.19) for a given y(t),
provided that To is sufficiently large, and Sy(w) and cr; can be subsequently
calculated from equations (7.20) and (7.22), respectively. The software package
:Mathematica® (1999) offers a convenient method for calculating the Fourier
STRUCTURAL RESPONSE STATISTICS: PART I 173
transform and the inverse Fourier transform from digitized data. In practice,
however, it is usually more efficient to calculate Sy(w) directly from experimental
time histories (for r¡(t), for instance) by means of an electronic instrument called
a frequency analyzer such as described in Problem 7.4, or by means of computer-
aided Fast Fourier Transform (FFT) methods to evaluate the Fourier coefficients
of equation (6.2) for use in equations (6.5). Newland (1975) elaborated on these
methodologies, and Cooley and Tukey (1965) presented efficient algorithms for
calculating the Fourier coefficients.
Then rewrite equation (7.20) twice using equation (7.19), first substituting r¡ for
y and then p 1 for y. The results for the respective spectral densities are
(7.24)
(7.25)
When equation (7.23) is substituted into equation (7.25) and this result is com-
pared to equation (7.24), the required relationship is deduced as
(7.26)
The derivation of response statistics that follow are based on a linear model
in the form of equation (2.43), written in terms of the displacement coordinate
v. However, similar results can be obtained based on a linear model in the
form of equation (2.81), written in terms of the rotational coordinate B. After
going through the following derivation for the displacement model, the reader
is encouraged to then derive the response statistics for the rotational model.
Given the linear structural model
(7.27)
174 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
for which the load spectral density Sp 1 ( w) is known through equation (7.26), the
task now is to calculate the response spectral density Sv(w) and the variance a;.
Because these results are so important in applications, all of the assumptions and
mathematical details needed for this derivation are included here. Because two
particular response functions for equation (7.27) are required eventually, these
are repeated for convenience. One is the harmonic response function given by
equation (5.62), or
(7.28)
Here the lower limit T = O was replaced by T = -oo since p 1 (t) vanishes for
T < O, leaving the value of the integral unchanged. Further it is recalled that
h(t- r) is the response toa unit impulse at (t- r) =O. For (t- r) < O, the
response v is zero because the unit impulse has not yet come into existence.
Thus for T < t, h(t- r) = O, and the upper limit t = T may be extended to
t = oo without changing the value of this integral. That is
Change the sign of the integral and reverse the limits of integration, or
(7.33)
(7.34)
STRUCTURAL RESPONSE STATISTICS: PART I 175
To show the relationship between h(t) and H(w), the solution v must now
be expressed in terms of H (w). To do this, set
(7.35)
(7.36)
(7.37)
1
kH(w)
1
= ¡= -=
h(t)e-jwtdt (7.39)
The function H(w)/k 1 is the Fourier transform of (27r)h(t). The inverse Fourier
transform of the latter yields
h(t) =
2~ ¡: : 1
H(w)dwtdw (7.40)
(7.41)
where B1 and (h replace T to avoid confusion. Assume that v(t) is stable and
that these integrals converge. The term Rv(T) can then be written as a double
integral where the order of averaging and integration is interchanged. That is
176 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
Next assume that p 1 (t) is stationary and ergodic. The autocorrelation func-
tion for this loading is then independent of time t and that portian of the
integrand of equation (7.42) involving p 1 (t) can be written as
Equation (7.43) is simply the autocorrelation function for p 1 ( T) with a time lag
of ( -e 2 + ei). The last two equations are combined to give
Response Parruneters
The spectral density of the response v is defined as the Fourier transform of
R,(T). That is, from equation (7.20),
(7.45)
After interchanging the order of integration in equation (7.46) and inserting the
following identity in the integrands,
(7.47)
(7.48)
When compared with equation (7.39), it is observed that the first two integrals
on the right of the last equation are H ( -w) 1k1 and H (w) 1k 1 , respective!y. With
equation (7.20), the last integral in equation (7.48) is identified as the power
spectral density of PI (t) with a time shift of ( -e2 + e1). The product of these
first two integrals is
1 1 2
k 2 H(-w) · H(w) = k 2 IH(w)l (7.49)
1 1
STRUCTURAL RESPONSE STATISTICS: PART 1 177
1 2
Sv(w) k 2 IH(w)j Spr(w) (7.50)
1
This last remarkably simple and useful result relates the power spectral
density of p 1 (t) to the power spectral density of v(t) through the complex fre-
quency response function. As shown in Example Problem 7.2, Sp 1 (w) is given
by equation (7.26) in terms of Sr¡(w) and G(w). With thls and the modulus of
the harmonic response function given by equation (5.63), the response spectral
density of equation (7.50) for the linear struetural model of equation (7.27) is
deduced as
(7.51)
(k¡ ~ mw 2 ) 2 + c'fw2
The variance of the response is then
Example Problem 7.3. The free, undamped lateral motion of the three-
legged jackup rig shown in Figure 2.17 has already been investigated in Example
Problem 5.4. Now include light damping, and subject this structure to steady,
unidirectionallinear waves with a significant wave weight Hs =15m and with a
distribution given by the Pierson-Moskowitz spectrum, equation (6.21). Based
178 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
on the result given by equation (7.52) for a single degree of freedom structural
model, calculate the variance o-;for the horizontal displacement of the deck.
Then assuming that the wave height distribution is Gaussian, calculate and
interpret the following quantities based on ±3av: the horizontal deck displace-
ment, the horizontal shear force in each leg, and the overturning moment. State
any further assumptions needed to make these calculations.
The first step is to clarify the mathematical model of equation (7.27). Iden-
tify the coordinate v, the constants for the stiffness k¡, the mass m, and the
damping c1 • Let v be the absolute horizontal displacement ofthe deck where vis
on the average much smaller than the horizontal wave particle velocity u. From
Example Problem 5.4, the undamped natural frequency wo for this structure is
given as the last ítem in Table 5.2, or
In this calculation, the equivalent stiffness and the virtual mass are: k 1 8.18
x 105 lbjft and m 4.42 x 105 slugs (or lb-sec 2 /ft). Lacking the required data
for cb assume ( = 0.05. With equation (5.64), it follows that c 1 2(v'k(iií =
6.01 x 104 lb-secjft.
The second step is to choose a reasonable wave theory and then calculate
G(w), the corresponding transfer function for wave loading of the legs. Assume
that the inertial fiow regime dominates. Then p 1 ( t), or the total horizontal
load on all three legs, Ís found by setting C D 0 and Íntegrating íj given by
Morison's equation (2.14) over the range from z -d to z =O. That is,
(7.54)
This last form implies that the wave forces on the legs are assumed to be statisti-
cally independent of each other so that the horizontal wave particle acceleration
u is essentially the same on all three legs. Let x = O be the location of the water
particles on each leg so that the value of ú for a single wave as given in Table
3.1 can be employed, or
. H 2 coshk( z + d) .
u= -2w sinh kd sm wt (7.55)
where the wave period T was replaced by 21r Jw and the wave amplitude A was
replaced by H /2. With this last result, equation (7.54) can be integrated to
give
(7.56)
side of equation (7.56) and then expressing the result in complex notation leads
to the required transfer function, or
(7.57)
lt follows that
(7.58)
(7.59)
In this example, traditional English units are used so that the numerical coef-
ficient of the exponential term above is (0.0081)(32.2)2 8.40 ft 2 jsec4 . From
equation (6.23), B = 3.11/ Il'; = 3.11/15 2 0.0138. Note that the constant 3.11
(7.60)
The units displayed for equation (7.60) are appropriate only if w is given in
radjsec. A plot of equation (7.60), given in Figure 7.5, shows that for all
practical purposes the wave energy is confined to frequency range of 0.16 to 1.6
radjsec, the lirnits of integration in the variance integral. With equations (7.58)
and (7.60), the variance equation (7.52) is thus
2
(]V=
( 37r 2
-8 pgD CM
) 2
2
11.6 8.40 e-0.0138/w4
5[(k1 - mw 2)2 + e 21w 2Jdw (7.61)
0 _16 w
where
pg = 64.3 lb/ft 3 , water density
D = 12 ft, single leg diarneter
CM = 2.0, assurned inertia coefficient
m = 4.42 x 105 slugs (lb-sec 2 jft), equivalent rnass
k1 = 8.18 x 105 lb/ft, bending stiffness
e1 = 6.01 x 104 lb-secjft, darnping
When equation (7.61) was evaluated nurnerically, the variance was = 0.276 a;
ft 2 , giving a v = O. 526 ft for the rms deck defl.ection. For a Gaussian process,
the probability of exceeding the following dynarnic responses is 0.26 percent:
deck displacernent: v = 3a v = 1.58 ft
horizontal shear load per leg: fmax= k 1 v /3 = 4.30 x 105 lb
overturning rnornent: Mmax = Hmax = 1.14 x 108 ft-lb,
(leg height: e = 265 ft )
Approximate Responses
In sorne applications, the spectral density of the loading Sp 1 ( w) can be ap-
proxirnated as a constant So over a frequency band between the lirnits of w 1
and w2 and as zero outside that frequency band. That is
(7.62a)
(7.62b)
Equations (7.62) define band-limited white noise. The designation white noise
originated with the description of white light for which the spectrurn is nearly
uniforrn over the range of frequency for visible light. In the present appli-
cations, it is noted that So is a one--sided spectrurn since it is based on the
one--sided experimental wave spectrurn S 77 (w). Further, it is assurned that the
undarnped frequency w0 of the single degree of freedorn systern is within the de--
fined frequency band of equations (7.62). With these equations it follows that
the variance of the defl.ection response, equation (7.52), has the following forrn:
STRUCTURAL RESPONSE STATISTICS: PART I 181
So~---.., ~~---------4~00 00
_ _..__ _ (¡) o
.
~
.
-
_
.
.
.
.
_
O IDo
(a) (b) (e)
Figure 7.6 Three idealizations of the load excitation spectrum: (a) band-limited
white noise; (b) white noise with a cut-off frequency; and (e) ideal white noise.
(7.63)
The solution to the latter equation, given by Crandall and Mark (1963), is
1rSo
av2 = - k [I(w2fwo)- I(w¡fwo)], w1 :S wo :S w2 (7.64)
1c1
1 t _ 2((wifw0 )
I(Wí 1wo ) = - an 1
1r 1- (wdwo)2
2
+ ( ln 1 + (wdwo) + 2(wdwo)~ (7.65)
27r~ 1 + (wi/wo) 2 2(wi/wo)~
This solution includes, in addition to band-limited white noise, two other special
cases: white noise with a cut-off frequency Wc, and ideal white noise. These three
cases are depicted in Figure 7.6. For white noise So with a cut-off frequency Wc,
then w1 =O, w2 Wc and the integral term of equation (7.64) becomes
For ideal whíte noise where So is uniform forO :S w :S oo, the integral term is
The latter result follows from the behavior of the arctangent term in equation
(7.65): as w2 ---+ oo, its argument is large and negative, and tan- 1 ( -oo) ---+ 1r.
182 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
Also, as w2 --+ oo, the argument of the naturallogrithmie term approaches unity
and ln (1) O. Thus, for ideal white noise, So is uniform for all frequencies and
1rSo
(7.68)
kici
Example Problem 1.4. For the jackup rig described in Example Problems
5.4 and 7.3, compute the upper bound for the rms deck deflection, av, for each
of the three idealized loading spectra shown in Figure 7.6. In all three cases,
choose
So= !G(w)! 2 S11 (w) (4.76 x 108 )(336.3) = 1.60 x 1011 lb2-secjrad
Here the quantity 336.3 ft 2 -secjrad is arbitrarily chosen as one--half the peak
value of the wave height spectrum shown in Figure 7.5, the value that corre--
sponds to the frequency of w = 0.324 radjsec. The results are summarized.
(a) For the band-limited spectrum of Figure 7.6a, choose the same frequency
limits that were used in the direct integration of equation (7.61), or w 1 = 0.16
radjsec and w2 = 1.6 radjsec. Compute the variance by evaluating equations
(7.64)-(7.66), from which
av = 3.1066 ft
(b) For the cut-off frequency spectrum of Figure 7.6b, choose WJ O and
w2 = Wc = 1.6 radjsec.Compute the vmiance by evaluating equations (7.64)
and (7.65) using (7.67). The result is
av = 3.1079 ft
av 3.1976 ft
These results show a small but progressive increase in the rms deflection as
the band width increases. However, these idealized approximations all led to a
gross overestimate in a v by about an order of magnitude, compared to the valu e
of 0.372 ft computed by direct integration of equation (7.61). The conclusion is
that direct integration gives the best answer, at least for this type of problem
in which the structure's fundamental frequency wo is at the very low end of the
wave height spectrum.
Extensions
There are severa! possible extensions to the classical statistical results ob-
tained thus far in this chapter. These extensions involve the relaxation of certain
restrictive assumptions upon which the variance of the response given by equa-
tion (7.52) was based. Two of those assumptions were: there exists a single,
stationary excitation wave force of a known spectrum Sp 1 (w) and the wave
excitation force has a zero mean value. This assumption of zero mean was ín-
vestigated by Tung (1974), who showed how a single, stationary wave excitation
STRUCTURAL RESPONSE STATISTICS: PART II 183
spectrum is modified when the mean is not zero, which corresponds to the case
of steady current forces and wave forces acting simultaneously on the structure.
In typical examples, however, Tung showed that such simultaneous action hada
definite but relatively small effect on the structural response. Thus the practice
of superimposing the effects of current as a static loading on the structure under
wave excitation seems to be justified, providing that all loading due to vortex
shedding in the current field is negligible.
Another extension is the case of two or more stationary wave excitation
forces that are uncorrelated or statistically independent, and all with a zero
mean value. Suppose that the N such spectral densities are known and denoted
by Si(w), i = 1, 2, ... ,N . From lengthy but straightforward calculations, the
structural response analogous to equation (7.52) then becomes
(7.69)
Modern control theory, developed mainly after 1960 for use by electrical and
mechanical engineers, offers sorne powerful techniques of dynamic statistical
analysis that are applicable to offshore structures. In this analysis, the linear
differential equations representing structural motion the excitation forces are
cast in state-variable form, those forms are transformed to a statistical repre-
sentation called covariance propagation, and the latter result is then solved to
obtain the statistical responses. These ideas are now discussed and illustrated
using a single degree of freedom linear structure subjected to stationary wave
184 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
excitation. The theory is freely drawn from the expositions of Bryson and Hu
(1975), Lin (1967), and Hedrick (1984).
State Variable Form
Equation (7.27), governing the structural motion v v(t), is rewritten as
1
ii + 2(wov + w5v -Pl(t)
m
(7.70)
(7.71)
(7.72)
where G(w) is the load transfer function and S17 (w) is the surface wave height
spectral density. This force excitation spectral density is now arbitrarily fitted
to the following equation:
(7.73)
Here, the constants &, w, and 2; are picked to give a best fit to the right side
of equation (7.72). There are two reasons for picking the latter form. First,
when IG(w)l 2 is constant, then S71 (w) has this general shape of equation (7.73).
Second, that form is precisely the spectral density obtained by passing white
noise w(t) through a linear filter given by
(7.74)
E[w(t)] =O (7.75)
z = Fz+rw (7.77)
where F and r are constant matrices and for brevity the argument t is omitted
from the variables z, w and their components. The state variables are defined
STRUCTURAL RESPONSE STATISTICS: PART II 185
from the two governing differential equations (7.70) and (7.74). These and the
other matrices of equation (7. 77) are as follows:
P1 . ]T
Pl
(7. 78)
m m
(7.79a)
V (7.79b)
z3
m
= Z4 (7.79c)
Z4
~
= -- 2(WZ4- ~ 2 + -W
W Z3
1 ~ 2 Ct 1/2 W A
(7. 79d)
m
o 1 o o
F
[o -w5
o
-2(wo
o
o
1
o
o
1
--2(w l (7.80)
1 2 ~ 1/2 ]
rw [o o o -wcx
m
A
w T
(7.81)
(7.84)
Taking advantage of the linearíty of the expectation operator E , the last equa-
tion becomes
(7.86)
What remains is to evaluate the last two terms on the right of the latter
equation. Begin by defining the state transition matrix cp(t, r) with the following
properties:
d
dt cp(t, T) Fcp(t, T), cp(t, t) = I (7.87)
(7.90)
Use the last result and then invoke linearity to put the expectation operator E
under the integral of equation (7.89). Since w(t) is independent of the integra-
tion variable T, include w(t) under that integral as well. Thus, equation (7.89)
beco mes
The problem in evaluating the last integral is that the impulse occurs at
the end point of the time interval. Use the symmetry property of the white
noise autocorrelation function, or Rw( T) = Rw( -T). The impulse can be ap-
proximated as the magnitude of any symmetric flmction whose time duration e
approaches zero in the limit. Choose a rectangular pulse at T = t of duration e
and of magnitude Qjc. The symmetry property leads to the evaluation of the
integral of equation (7.94) as
¡ t
Qb(t- T)dT =
¡t+t:/2 Q8(t (7.95)
o t-e/2
With this last result and equation (7.94), the third term on the right of
equation (7.86) is determined as rQrT /2, which, after sorne algebra, turns out
to be identical to the last term on the right of equation (7.86). Thus, the
covariant propagation equation (7.86) becomes
• T
Z=FZ+ZF +B, (7.96)
FZ+ZFT +B =O (7.97)
Davison and Man (1968), R. Smith (1968), and P. G. Smith (1971) discussed
such methodologies. In a typical procedure, the eigenvalues ofF are first calcu-
lated, an arbitrary scalar parameter /3 is chosen as two and one-half times the
188 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
real part of the largest eigenvalue, and successive matrix solutions are generated
by the following recursion formula:
(7.98)
The quantities of equation (7.98) are defined as follows, in which I is the identity
matrix:
U= (¡31- F)- 1 (7.99a)
V= U(¡3I+F) (7.99b)
W 2¡3UBUT (7.99c)
(7.99d)
The generation of the successive terms in the series is aborted when the succes-
sive changes in partial sums becomes sufficiently small, or less than one percent.
If the system is not too large, then the following closed form solution to
equation (7.97), based on Laplacian transforms, can be used (Lin, 1967). That
is,
(7.100c)
where .c- 1 is the inverse Laplace transform and s is the Laplace operator.
A Closed Form Solution
The solution to the covariance propagation equation for the single degree
of freedom system described by equations (7. 70)-(7.81) was calculated using
the integral solution of equations (7.100). One result is an expression for the
varíance of the displacement in terms of the three structural constants (m, k1, c1 )
and the three load excitation parameters (0:, w, (). That is
22
4 2 2
~e In+ e 1112 + 22
e /3/4
L._¿
n=l
2an a1 + a2 a3 + a4
(7.101)
where
STRUCTURAL RESPONSE STATISTICS: PART II 189
a= -k¡fm; b=-c¡fm; e=
d= -2(w; e w2 a112 /m
2s1 d + (d2 + 4c) 112 ;
2s3 b + (b2 + 4a) 1 12 ; 2s 4 = b (b2 + 4a) 1 12
el = d(d- b)/[c(a- c)(d- b)- (be- ad)(c + ~ bd a)]
Cz=-CI(c+~ bd-a)j(d-b)
C3 = -C1; C4 = -(1 + aCz)/c
'1' 1 = (s1C1+ Cz)/(si- sz); '1'2 = -(szC¡ + Cz)J(si - s2)
')'3 = (s3C3 + C4)j(s3- s4); '1'4 (s4C3 + C4)j(s3- s4)
ai -si, i 1, 2, 3, 4
Although this integral solution is exact, it is recalled that the results are
an approximation to reality because of the assumptions inherent in the math-
ematical model. In addition to the single degree of freedom approximation for
the structure, the assumed form for Sp 1 (w), equation (7.73), does not exactly
replicate Sry(w). Note that as w----+ O, Sp 1 (w) &, whereas Sry(w)----+ O. However,
these two spectra have similar behavior otherváse: both form a single peak and
both approach zero as w becomes large.
(7.102)
The two terms on the right side of the last equation are known from Example
Problem 7.3: IG(w)IZ = 4.76 x 108 lb 2 /ft 2 , and Sry(w) of equation (7.60), which
has its peak at 672.5 ft 2-secjrad at the frequency w = 0.324 radjsec. Now match
the peaks for each side of equation (7.102) at w = w = 0.324 radjsec, which
leads to
190 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
PROBLEMS
7.1 If y(t) is Gaussian with zero mean and has a variance of O" y, calculate
by numerical integration the probability that y(t) is outside the levels y= ±O"y
and y = ±20" y. Check your results using Gaussian probability tables such as
found in a statistics reference book.
7.2 Suppose that y(t) is Gaussian, exists only over a narrow band of fre-
quencies, and is a smooth function of time. Also assume that each cycle crosses
the mean level y(t) =O so that the maxima always occur for y(t) >O and the
minima always occur for y(t) < O.
(a) Sketch a function y(t) which behaves as defined.
(b) The probability distribution for the peaks of y(t) so defined is given by
equation (7.7), the well-known Rayleigh distribution. Sketch p(a) as a function
of the amplitude a of y(t). Prove that the maximum value of p(a) occurs when
the amplitude is equal to the standard deviation, ora= O"y.
(e) Compute the probability that any peak of y(t) exceeds these two values:
20"y, 30"y. Note that the probability that any peak of y(t) exceeds a is given by
00 2
1 p(a)da = exp (- a 2 )
20"y
7.3 A digital time history of y(t) is available in the form of points y(tn) at
even increments of time, n = 1, 2, ... , n. Outline a computer-aided method that
will generate a probability density function p(y) from these data. How would
you check p(y) to see if it was Gaussian?
7.4 A schematic diagram of an instrument called a spectrum analyzer
is shown in Figure 7. 7. The wave height input r¡(t) is assumed to be a sta-
tionary ergodic random process. This input is filtered by a filter whose har-
monic response function H(w) is a constant H 0 in the narrow frequency band
(wo- !:lw/2)::; w::; (wo +!:lw/2). The filter output y(t) is squared, and its time
average z(t) is calculated from
The mean level of z(t), estimated from the output meter for sufficiently long
time periods To, is
PROBLEMS 191
Input
Wave
Y(t)
21t/ffio
Squaring Device
E[z(t)] = -
1 ¡ro/2 E[y (t)]dt
2
To -ro/2
(b) State the assumption that leads to E[z(t)] = E[y 2 (t)]. Then show that
the mean level E[z(t)] of the output meter is a direct measure of the wave
spectral density, or
S ( ) "' E[z(t)]
r¡ wo - 2H2 ~w
o
7.5 The Pierson-Moskowitz wave height spectrum for a significant wave
height of H 8 = 15 m is given by equation (7.60) and is plotted in Figure 7.5.
192 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
Use this equation and employ numerical integration, with reasonable limits for
the integral, to compute the area A under the curve of Figure 7.5. Then verify
the result discussed in Chapter 6 that H 8 = 4A 112 15 m.
7.6 In Example Problem 1.4, three rms displacement responses crv for the
jackup rig were computed, each based on an idealized model: band-limited white
noise, white noise with a cut-off frequency, and ideal white noise. Using the same
parameters as for this exa:rnple problem, solve equation (7.63) by numerical
integration to obtain crv for each of these three idealized models. Compare your
results to the corresponding results in the text that were derived from closed
form solutions to the integral. Explain possible differences in the results. Also
explain why these idealized models give results for cr v that are about six times
that obtained from numerieal integration of equation (7.61).
7. 7 Reconsider the jackup rig deseribed in Example Problem 5.4, with the
design parameters given in Table 5.2. The statistical responses to this same
strueture were discussed in Example Problem 7. 3. lt is proposed to add more
equipment to the deck of this jackup rig so that the deek weight mdg would
increase from 1.02 x 107 lb to 5 x 107 lb.
(a) "''hat percentage of the Euler buckling load is this new deck load? Use
equation (5.36) to answer this and to explain whether the new deck weight will
increase the chanee of structural buckling.
(b) With the new deck weight, the other parameters of Table 5.2, and
equation (5.36), caleulate the following quantities: the equivalent bending stiff-
ness k1 , the equivalent mass m, the equivalent damping constant c1 based on
( c¡jJ4k 1 m 0.05, and the undamped structural frequency wo.
(e) Based on the same parameters of Example Problem 7. 3, exeept for the
modified values of k 1 , m, and c1 , use equation (7.61) and numerical integration
to compute the variance cr~. \Vould you expect crv to be higher or lower when
compared to the original rig with the lower deck mass? Explain.
(d) Compute the response cr v of the modified strueture to ideal white noise.
Compare this result to that obtained in part (e). What assumptions account
for the differences in these two results?
7.8 Consider the concrete monotower shown in Figures 2.2 and 5.2. This
structure is idealized as a rigid body eonsisting of a rectangular box caisson and
a uniform leg. The structure rotates in the plane with angle e about the base
point O. The foundation is assumed to be linearly elastic with a stiffness ke and
damping ce, modeled respeetively by equations (2.78) and (2.79), in which the
frequency w is identified as the fundamental frequency in free vibration, or wo.
Let Jo denote the virtual mass moment of the strueture's ínertia about poínt O,
and let Af(t) denote the moment on the strueture indueed by wave action. The
governing equation (2.6) for rotational motion thus has the following form:
JoB+ ceB + keO = M(t)
(a) By eomparing the symbols of the above equation to those for tra:r1slational
motion v, or equation (7.27), deduce for rotational motion the harmonic response
funetion H(w) in a form similar to equation (7.28).
REFERENCES 193
(b) This monotower is subjected toa simple wave as described in Table 6.1.
Derive the explicit equations for the two majar components of the wave-induced
moment M(t): the moment on the uniform leg, and the moment on the caisson.
HINTS: For the leg moment, use q of Example Problem 4.4, integrate q over
the leg height to calculate the total horizontal wave load on the leg, compute
the centroid of this load from point O, and form the product of the last two
quantities to give the first component of the moment. The second component
of the moment, that on the caisson, was derived in Example Problem 4.3. Carry
through those calculations, expressed symbolically by equation (4.22).
(e) Based on the results of part (b), derive the equation for the transfer
function G(w) corresponding to the total wave-induced moment M(t) on the
whole structure.
7.9 Assume that the wave field imposed on the monotower described by
Problem 7.8 is stationary and ergodic.
(a) By comparing the symbols in the equation (7.27) to the equation of
motion for rotation () given in Problem 7.8, deduce without calculation the
explicit forms of Se(w) and a~ for rotational motion. These results will be
analogous to equations (7.50), (7.51), and (7.52).
(b) For the quantities H(w), G(w), Se(w ), anda~ corresponding to rotational
motion, write clown a consistent set of units, first in the traditional English
system, and then in the SI system.
7.10 The purpose of this problem is to obtain numerical results for the
responses of the concrete monotower, for which the theoretical results were
obtained in Problems 7.8 and 7.9. The system parameters for this tower are
summarized in Table 5.1, which also lists two valu es for the free, undamped
rocking frequency w 0 based on two soil foundation stiffnesses.
(a) Based on the weaker soil foundation for which G s = 10 MPa and
w = wo = 1.41 radjs, compute the soil foundation parameters ke and ce us-
ingequations (2.78) and (2.79).
(b) Compute the foundation damping factor ( = ce/J4keJo. This damping
factor is analogous to ( = ctfJ4k1 m, the damping factor for equation (7.27).
(e) The monotower is subjected to steady, unidirectional waves with a sig-
nificant wave height H 8 of 15 m and with a distribution Sr¡(w) given by the
Pierson-Moskowitz spectrum, equation (7.60). With this spectrum, the numer-
ical results of part (a), and the equation previously derived in Problem 7.9 for
the variance, compute a~ by numerical integration.
(d) Based on the ±3ae limits, compute the angle of rotation, the horizontal
displacement of the deck, and the horizontal shear force at the base, and the
overturning moment for this 180m high monotower.
REFERENCES
Bryson, A. E., and Hu, Y. C., Applied Optimal Control: Optimization, Estimation,
and Control, Wiley, New York, 1975.
194 STATISTICAL RESPONSES FOR LINEAR STRUCTURES
Cooley, J. W., and Tukey, J. W., An Algorithm for the Machine Calculation of
Complex Fourier Series, Mathematics of Computation 19, 1965.
Crandall, S. H., and Mark, W. D., Random Vibration in Mechanical Systems, Aca-
demic Press, New York, 1964.
Davison, E. J., and Man, F. T., The Numerical Solutions of ATQ + QA -C,
IEEE Transactions on Automatic Control, August 1968.
Gould, P. L., and Abu-Sitta, S. H., Dynamic Response of Structures to Wind and
Earthquake Loading, Wiley, New York, 1980.
Hedrick, J. K., and Firouztash, H., The Covariance Propagation Equation with Time-
Delayed lnputs, IEEE Transactions on Automatic Control, October 1974.
Lutes, L. D., and Sarkani, S., Stochastic Analysis of Structural and Mechanical Vi-
brations, Prentice Hall, Upper Saddle River, NJ, 1997.
Muga, B. J., and Wilson, J. F., Dynamic Analysis of Ocean Structures, Plenum, New
York, 1970.
Tung, C. C., Peak Distribution of Random Wave-Current Force, Journal of the En-
gineering Mechanics Division, ASCE 100 (EM5), October 1974.
(8.1)
(8.2)
in which the superscript T denotes transpose ( the interchange of the row ele-
ments shown to its defined column array). This vector can be composed of a
mixture of displacement coordinates, designated in applications as v1 , v2, . . . ,
and rotational coordinates, designated as 8 1 , 82, .... Each chosen coordinate
describes the motion of a node poínt on the structure, such as the mass center
of a structural element or a junction point on a frame. Each chosen coordinate
must then be tested as follows for independence. Freeze the motion of (N 1)
node points and then check whether the lone remaining node can have motion
when loaded. If and only if motion occurs at that lone node is its a..ssociated
coordinate independent. Then repeat this mental test for each of the remaining
nades, one-by-one, to check for their independence.
(e)
(b)l
Figure 8.1 (a) A jacket template structure; (b) stalk model with externa! loads; (e)
free body sketches of the two masses.
(8.3)
(8.5)
The total load acting at each node is approximated by integrating q(z, t) over
the appropriate portions of the four legs up to each node point. Recall that
the coordinate z has its origin at the still water line and is measured positive
upward. The results are
2
Pl =
Jo
-(d-P 2 ) 2
w
k
( sinh k€2) .
q(z, t)dz =--CM pD 2 - H 1- .
1r
smh k
d sm wt (8.6a)
(8.6b)
in which d is the water depth. For this case, the loading vector is expressed as
(8.7)
EQUATIONS OF 1\JOTION: GENERAL FORl\f 199
o
(8.8)
o
The conditions under which M is diagonal are now discussed and typical calcu-
lations for M are illustrated.
Recall that each generalized coordinate represents either the displacement
or the rotation of a portian of the structure's mass at a numbered node point.
Sometimes a fixed point labeled O is considered a node point. The questíon
remaíns: How does one determine the portian of the structural mass to be
associated with each nade point? Equivalently: How does one formulate the
mass matrix?
The mass lumping method is probably the most popular method of dis-
cretizing the supporting framework and the rigid body portions of an offshore
structure. For a framework, the mass lumping requíres sorne experience on the
part of the analyst. For flexura! motíon of structural frame elements, the an-
alyst can use as a guideline two particular cases discussed in Chapter 5. For
instance, it was calculated in Example Problem 5.3 that, if 37 percent of the
mass of a uniform beam clamped on both ends (a cross member of a supporting
framework) is lumped at midspan of an equivalent massless beam of the same
flexura! stiffness, then the fundamental flexura! frequencies of the two beam
models are the same, for practica! purposes. In an another case, the jackup rig
of Erample Problem 5.4, it was calculated that if 37.5 percent of a cantilevered
beam's mass is lumped at the tip of its massless counterpart (an elastic beam of
the same geometry, restraint, and bending stiffness), then both configurations
have nearly the same fundamental flexura! frequency. In such cases, the mass
not accounted for is of no consequence; but for argument's sake it can be lumped
at a fixed end node point O. In these two cases, then, the criterion for lumping
the mass is based on preserving the fundamental flexura! frequency between the
continuous element and its simple lumped mass counterpart, and this frequency
equivalence is based on the conservation of potential and kinetic energy during
motion.
In practica! cases, however, the end constraints for an element of a sup-
porting framework are not so simple as these two cases just discussed. Thus,
without making further calculations, the choice of the fraction of element mass
to be lumped ata node becomes quite subjective. Nevertheless, the experienced
analyst knows that lumping toa node between 25 percent and 40 percent of the
element mass surrounding that node usually leads to an adequate structural
dynamic model with a diagonal mass matrix.
200 MULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
Example Problern 8. 3. Compute now the virtual rnass for Example Problem
8.1, shov.rn in Figure 8.1. Here, nodes 1 and 2 are on the vertical centerline of the
structure, in line with the base point O, and the system mass is symmetrically
distributed with respect to this centerline. Choose node 1 at the rnass center
of the deck and its equiprnent, at the height + e2) frorn o at the sea floor.
Locate node 2 at the mass center of the horizontal rnembers, at height e2 from
O. The masses lumped at nodes 1 and 2 are approximated as follows:
(8.9)
node 2. The subscripts v 1 and v2 for the other mass terms denote virtual mass
for portions of the legs and nonhorízontal members (or supporting structure)
surroundíng nodes 1 and 2, respectively. J:.or simplicity, assume that all members
below the deck are fully submerged. The coefficient 0.375 is the proportion of
the structure's mass that surrounds a particular node point, deduced from a
somewhat similar cantilevered structure using energy considerations, Example
Problem 5.4. The remaining mass of the whole system can be considered lumped
at the fixed node point O, and is of no consequence. The virtual masses for each
node are approximated by
Recall that the quantitíes in added mass terms are: the added mass coefficient
e A ( which, lacking any experimental data for this geometry, can be approxi-
mated as unity); the mass density of water p; and the volumes of water Vh, V1
and v2 displaced by the respective structural components: the submerged hor-
izontal members, the submerged members surrounding node 1, and the sub-
merged members surrounding node 2. The mass matrix for this example is
thus
(8.11)
K=
r k21
ku
.
kNl
kl2
k22
kN2
k,N
~2N
kNN
l (8.12)
The constant k¡j is defined as the force that is required at node i to coun-
teract a unit elastic displacement ~sJ = 1 imposed at node j, under the condi-
tion that all displacements ~sí O for i f:: j. In this single displacement case,
Qsi = kij (1) = kíj, í :-/:: j. If such a displacement condition is applied sequentially
to each node, then the net force at each node j can be obtained by superposi-
tion. For instance, the elastic force q81 at node 1 is a linear combination of the
forces at node 1 resulting from the unit displacements at all nodes, or
Equation (8.15) can be expressed in matrix notation in the following two ways:
Esl
l ][
ku
l
qsl k¡2 k¡]V
[
Qs2 k21 k22 k2N Es2
qs K~s (8.17)
Example Problem 8.4. In practice, a reliable structural computer package
(SAP 2000, for instance) is used to compute the K matrix. This general proce-
dure is illustrated for the simple jacket template structure of Example Problem
8.1, shown in Figure 8.1. A plane frarne, static force-displacement analysis is
made in the two stages depicted in Figures 8.2. In the first stage, Figure 8.2a, a
unit displacement = 1 is imposed at level 1, and ( 82 O is imposed at level
2. The corresponding forces k11 and k2 1 necessary to achieve those displace-
ments are computed. In the second stage, Figure 8.2b, ( 81 =O and 1;, 82 1, for
whieh the respective forces at levels 1 and 2 are k12 and k 22 . Each such force is
assumed to be positive in the direction of its coordinate. If a force is opposite
to the direction of its coordinate, then that force is a negative nurnber. In this
problem, it is intuitive that k 12 ( = k 2 1) is a negative number. The stiffness
matrix for thís exarnple is
(8.18)
basic beam elements, the junction fixity (in this case, welded joints), and the
value of Young's modulus, the cross-sectional area, and the cross-sectíonal area
moment of inertia for each beam element in the plane frame. In this example,
the deck can be modeled as a rigid body, or a beam vvith high flexura! stiffness
compared to that of the supporting structural elements. Note that a computer
analysis requires units for the geometry and material properties, and that any
compatible set of units can be used. Thus, the unit displacements can be 1
m or 1 ft, for instance, depending on whether the SI or the traditional English
units are employed.
1
t-1-j
(a) (b)
Figure 8.2 Definition of the stiffness influence coefficients for Example Problem 8.1.
In this analysis, the damping force qn;. for the structural node coordinate ~i is
assumed to be a linear combination of the generalized coordinate velocities ~í'
i = 1, 2, ... , N. The constants relating the nodal damping forces to the nodal
velocities are called the damping infiuence coefficients Cij; and the relationship
is analogous to force-deflection relation given by equation (8.15). That is
N
qDi +. ... +e·2)'-,J
é 2.::: Cij~j (8.20)
j=l
204 A1ULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
Equation (8.20) can be expressed in matrix fonn in the following two ways:
(8.21)
(8.22)
The damping matrix can be cast in several different specialized forms, each
of which has the ad'\>antage of easily utilizing available experimental data to
determine the elements Cij. One such form is Rayleigh damping in which e is
proportional to the system's mass and also the system's stiffness. That is
(8.23)
in which a 1 and a 2 are constants. A more explicit form for e based on Rayleigh
damping will be presented la ter in this chapter.
Other specialized forms of e are beyond the scope of the present work.
Those forms include Caughey damping, for which Rayleigh damping is a special
case (Caughey and O'Kelly, 1965; Chopra, 2001); and cornplex stiffness damping
(Clough and Penzien, 1993).
(8.24)
in which the surn on the left represents all forces applied to mi in the direction
of ~i· Those forces, which have lines of action acting through the mass center of
each mi, include the the net restoring force q8 i due to structural stiffness, the
net viscous damping force qDi, and the lurnped '\>alue of the environmentalload
Pi· Note that equation (8.24) is analogous to equation (2.1) for a single degree
of freedorn with one mass and one coordinate.
Suppose a portian of the structure can be modeled as a rigid mass mi whose
rotation is defined by the coordinate f.i· Define Ji as the mass moment of inertia
of mi about an axis perpendicular to the plane of motion and through its mass
center. Then, the form of Newton's law of motion for this rigid body is
(8.25)
EQUATIONS OF MOTION: NEWTON'S METHOD 205
Here, the sum on the left represents all moments applied to that mass, moments
that are positive in the direction of €i and that lead toan angular acceleration ~i·
Such moments arise from forces whose lines of action are not through the center
of mass of the body, moments due to structural stiffness, structural damping,
and environmental loading. The mass center need not be fixed.
Example Problem 8.5. Shown in Figure 8.1c are the free body sketches for
the two lumped masses of the jacket template structure. For each mass mi, and
for each corresponding coordinate ~i =vi, there are three types of in-line loads:
qsi and qni, which both oppose the positive direction of Ei, and the positively
directed load Pi induced by wind, waves, and currents. \Vhen equation (8.19)
is applied to mass mi, the result is
(8.26)
\Vhen the forces qsi and qDi of equations (8.15) and (8.20) are combined with
equations (8.26), the result is the set of two differential equations of motion, or
i = 1,2 (8.27)
[ ;1 o
m2 ][ ]+[
iil
ii2
en
C21
Cl2
C22 ][ ]
V¡
v2
+[ ku
k2l
kl2
k22 ] [ ] [~ ]
V¡
V2
(8.28)
Example Problem 8.6. Newton's method is now used to derive the equations
of motion for the three degree of freedom model of the jacket template structure
shown in Figure 8.3a. This structure is modeled as the stalk configuration of
Figure 8.3b with the mass lumped at nodes 1 and 2, as for Example Problem
8.1. The difference now is that the deck is allowed to rotate with angle B about
node 1 as nodes 1 and 2 undergo horizontal displacements v 1 and v2. Thus,
there are three degrees of freedom and the coordinate vector is
(8.29)
The loading vector, dueto the external environmental forces and moments, is
(8.30)
in which p 1 and P2 are the horizontal forces lumped at levels 1 and 2, and Md
is the net moment about the mass center G of the declc, due to wave slamming.
The mass matrix is
(8.31)
206 MULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
where miand m2 are the sanie as for Example Problem 8.1 and are given by
equations (8.9) and (8.10); and Jd is the mass moment of inertia of the deck
only, taken with respect to an axis perpendicular to the plane of motion and
through G. Another way to express the latter quantity is Jd = mdr 2 where md
is the deck mass and r is the deck's radius of gyration.
-qS3
¡-v¡
-~%
3
:---+-
1 ~M
---T-- e
-
m¡,Jd
1-l
(e)
(b)
Figure 8.3 (a) Jacket template structure; (b) three degree of freedom model with
externa! loads; (e) free body sketches.
The respective generalized stiffness and damping forces for node i are special
cases of equations (8.15) and (8.20) for N= 3, or
i = 1,2,3 (8.32)
i = 1,2,3 (8.33)
Here, the stiffness infl.uence coefficients kij for i, j = 1, 2, 3, are the forces and
moments defined in Figure 8.4. These coefficients are computed from a struc-
tural program based on the three sets of displacement conditions shown, respec-
tively, in Figures 8.4a, 8.4b, and 8.4c: VI = 1, V2 =e =O; VI =e= o, V2 = 1;
and VI = V2 = o, e = l. The damping infl.uence coefficients will be determined
later in this chapter.
EQUATIONS OF MOTION: NEWTON'S METHOD 207
1+ c==:t+===s-k\2
Figure 8.4 Definition of the stiffness influence coefficients for Example Problem 8.6.
Refer again to Figure 8.3 in which the free body sketch for each of the two
masses is shown. For translational motion, the application of equation (8.24) to
each mass gives
(8.34a)
(8.34b)
For rotational motion of m 1 ( the total mass of the deck plus the lumped portion
of the virtual mass of the upper legs, as previously defined), the application of
equation (8.25) gives
(8.34c)
When equations (8.32) and (8.33) are combined with (8.34), the results are
the three equations of motion, or
(8.35a)
(8.35b)
(8.35c)
208 MULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
[f'
o
rn2
o
o
o
Jd l[r ]+[ en
c21
C31
C12
C22
C32
C13
C23
C33
][ l v1
v2
é
3] [
l [S,]
kl2 V1
+[ kn
k21 k22 k~3
k t'2 (8.36)
k31 k32 k33 ()
which has the sa.me formas the matrix equation (8.1). Further, equation (8.36)
reduces to equation (8.28) for its two degree of freedom counterpart in which
the coordinate e is suppressed.
8.3 EQUATIONS OF MOTION: LAGRANGE'S FORMULATION
(8.37)
(8.38)
The third scalar, the system's virtual work 8TV, is the sum of the virtual
work done on each component mass by each generalized nonconservative force
EQUATIONS OF MOTION: LAGRANGE'S FOR1\1ULATION 209
¡t2
l
t5(K- V)dt +
¡t2
t¡
t5W dt o (8.40)
The sum of the time variations of the difference in the kinetic and
potential energies and the work done by nonconservative forces over
any time interval t 1 to t2 is zero.
It is now shown how equation (8.40) leads to the equations of motion for struc-
tural systems.
Derivation for a Simple System
Lagrange's equations are now derived for a simple dynamic system which
is defined by all of the following characteristics: a set of N generalized coordí-
nates ís assigned, one for each degree of freedom (the system is scleronomic); an
independent variation can be given to each of the generalized coordinates with-
out violating the system constraints ( the system is holonomic); the generalízed
conservative forces, such as those due to elastic deformation and those due to
gravity, are all deriV'a.ble from a potential energy function of the form of equation
(8.38); the generalized nonconservative forces are the externally applied forces
and those forces such as viscous fiction that dissípate energy irreversably.
Kow compute the first variations 8K and 8V for equations (8.37) and (8.38).
Substitute these results, together with 8W of equation (8.39), into the varia-
tional equation (8.40), which leads to
+ tgibf,i) dt =o
(8.41)
210 1'V!ULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
For the second sum from the left, interchange the sum with integration and then
integrate by parts, or
(8.42)
o (8.43)
Now choose a particular coordinate i = n for which 8~n =J: O, but for which
8~í O for all remaining N - 1 values of i. Again, this can be done since the
coordinates are independent. It then follows that the sum in equation (8.43)
disappears and the single square bracket that remains in the integrand contains
terms all with the subscript n. To satisfy this equation, that square-bracketed
term must vanish. Since n is arbitrary, this results holds for all values of n.
After a rearrangement of terms in this square bracket, and a change in index
from n to i, the result is the set of N Lagrange equations of motion for a simple
system, or
For ease of reference for applications, the three scalar energies are summa-
rized in their index and matrix forms. For small motion, the kinetic energy has
the following quadratic form:
(8.45)
Note that for a diagonal mass matrix, the elements of M are míj O for i =J: j
and mii = m,¡ for i j.
For small motion, the potential energy can be expressed as
1 N N 1
v =
2
¿¿ kíj .;i~j + vg = 2erK e+ vg (8.46)
J=l 1=1
in which the double sum is a quadratic form, and Vg can also be so exprP.ssed.
The term Vg represents the sum of all gravitational potential energy changes
EQUATIONS OF MOTION: LAGRANGE'S FORMULATION 211
of the system's mass from the system's equilibrium state. This term V9 can be
safely omitted for lumped mass frameworks in which each mass has a negligible
rise and fall along the gravity vector. (The gravity vector points toward the
mass center of the earth). However, V9 can be important in determining the
dynamic stability of a relatively rigid gravity platform rocking on an elastic
foundation, as will be demonstrated in Chapter 9.
The virtual work in the present context is
N N
8W = l.: 9i 8~i = L(Pi - qDi)8~i = u:.r g = 8f.r (p- qD) (8.47)
i=l i=l
in which p is the vector representing the externally applied loads and qD is the
vector for viscous damping, given by equation (8.22).
(8.48a)
(8.48b)
1 2 1 2
= 2ml vl + 2m2v2
(8.49)
(8.50)
The virtual work for the nonconservative forces is evaluated from equation
(8.4 7):
(8.51)
212 MULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
With the values of qm and qD2 from equation (8.20), the generalized forces,
which are the coefficients of the virtual displacernents, are as follows:
(8.52)
oK o
OVl
With equations (8.52) and these last calculated results, the equations of rnotion
(8.48) become
(8.53a)
(8.53b)
When the viscous damping forces in the above equations are rearranged to the
left sides, then it is observed that equations (8.53a) and (8.53b) are identical to
the results obtained using Newton's Method, or equation (8.27) for i 1 and
í 2, respectively.
Comments
At this point, it is appropriate to add several comments and recommenda-
tions concerning the equations of structural motion formulated in this chapter.
First, it is highly recornmended that the stiffness rnatri:x K be cornputed us-
ing a well-tested, cornmercially available structural software package, SAP 2000,
for instance. In analyzing the supporting framework for tall offshore structures
in relatively deep water and with very rnassive decks and deck equiprnent, it is
recornrnended that the chosen sofhvare account for the prestress of the structural
elernents, in particular, the axial cornpression of the vertical beam elernents. Re-
call that, for the jackup rig analyzed in Chapter 5, the leg's flexural stiffuess was
reduced by the dead weight cornpressive load of the deck. However, if the static
Euler buckling load for the tower is rnuch srnaller than the deck load and its
equiprnent, then such prestress with its accornpanying reduction in the flexural
stiffness of the frarnework can be ignored.
EQUATIONS OF MOTION: LAGRANGE'S FORAiULATION 213
Second, it is not difficult to show that the Lagrange equations (8.44) can
be equivalent to the matrix form of (8.1), provided that the dynamic system is
simple, that the motion is small, and that K and V are expressed in quadratic
form. To show this equivalence, let K, V, and 8W have the forms of equa-
tions (8.45)-(8.47), except let Vg = O. \Vhen the terms of equations (8.44) are
evaluated using equations (8.37)-(8.39), then the result is
M~+Ke =g (8.54)
(8.55)
The last comment is that, once the equations of motions are formulated,
with all time-varying environmental loads and the constant coefficients identi-
fied, then those equations can be solved directly to determine the time-varying
responses ~i = ~i(t), i = 1, 2, ... , N. Todo this, the analyst has a wide choice of
a computer software packages, including PSI-Plot (1999) and Mathematica®
(1999). For such computations, 2N initial conditions must be specified, or
(8.56a)
eco) (8.56b)
In general, steady state solut.ions ~í(t) with light damping are sought, and
those responses occur in the numerical solutions if the run time t is sufficiently
long. In such solutions, the damping eventually eliminates the initial transient
responses so that the choice of initial conditions expressed by equations (8.56)
is of no consequence. Recall that for a single degree of freedom system, light
structural damping was based on the parameter ( c¡j2..¡¡;¡;;m in which ( was
in the measured range of 0.05 to O. l. See equations (5.58) and (5.64). Using
this latter information, rough estimates of cij Cji can be made to obtain the
steady state numerical solutions. A more exact way to relate ( to Cij is discussed
later in this chapter.
Although numerical solutions can generally be computed by what is sorne-
times called the bTute joTce method, it is quite appropriate and often more
physically meaningful to obtain closed form solutions to the equations of mo-
tion by the classical normal mode method. The remainder of this chapter is
devoted to this latter method: the computation of the system's characteristic
frequencies in free vibration Wn and their corresponding modal vectors Xn, and
the superposition of these modal vectors to obtain the structure's shape in terms
of the steady state response vectors ~i = ~í(t).
214 MULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
The values of Wn and Xn are computed from the condition of free, undamped
structural motion. That is, C p =O, for whích the governing equatíon (8.1)
becomes
M~ +K~=O (8.57)
Frequencies
Assume a harmonic solution to equation (8.57) in the form
(8.58)
(8.59)
Since ejwt is arbitrary, then the bracket term in the last equation is zero, which
can be expressed as
(8.60)
(8.61)
(8.62)
][ !:J [: l
kn- w~m1 kl2 k1n
k21 k22 -w~m2 k2n
(8.65)
It is noted that, since each frequency Wn is distinct, and ~In 1, then there is
a unique modal vector enfor each frequency.
After the N modal vectors are calculated in this way, those vectors are
normalized with respect to the mass matrix to form the new modal vectors Xn·
That is
(8.66)
enMen (8.67)
en is always a positive real number, which follows since 1\tl is positive
definite and symmetric.
Orthogonality of the Modal Vectors
To achieve the uncoupling of the equations of motion (8.1), it is first neces-
sary to show that the modal vectors Xn are mutually orthogonal with respect
to both M and K. This orthogonality is defined as follows:
(8.68)
(8.69)
in which 8nr 1 for r = n and 8nr = O for r =/= n.
Consider the two cases for r n. For M: when equations (8.66) and (8.67)
are combined, it follows directly that
(8.70)
216 MULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
(8. 71)
When this last result is premultiplied by x'}; and equation (8.70) is used, then
,,2
x;:Kx, Mx n ""'n (8.72)
Consider the two cases for r cf n. For M: let n r in equation (8.71) and
take the transpose of both sides to obtain
(8.73)
(8.74)
x~Kxn (8.76)
Thus, this last result, together with equation (8.70), completes the proof of the
orthogonality statement (8.68). For K: the right side of equation (8. 76) is w~
for n r by equation (8.70), but is equal to zero for n # r by equation (8.78).
This completes the proof of the orthogonality statement (8.69).
Derived in this section are steady state solutions to the structural equations of
motion (8.1), which include both forcing vector p and system damping. These
solutions utilize the free, undamped modal vectors Xn and their associated un-
damped frequencies w 11 , both derived in the last section. These solutions depend
on two important matrix forms: the modal shape matrix X and the modal
damping matrix C. Following the definition of these forms, a transformation
is introduced that uncouples the equations of motion, allowing for the normal
mode solutions to be displayed in closed form. The method of solution is then
FORCED, DAMPED MOTION 217
(8.79)
(8.80)
XN2
Note that the nth normalized modal vector forros the nth column of X, where
(8.81)
(8.82)
(8.83)
in which I is the N x N identity matrix, wíth 1 for all diagonal elements and
zero for all off-diagonal elements. Recall that the notation diag(function of n)
represents an N x N matrix whose argument defines the n consecutive diagonal
elements (n = 1, 2, ... , N), with zero for all off-diagonal terms.
Modal Damping Matrix, C
Modal damping is assumed in the forro first proposed by Lord Rayleigh
(1945) and previously given by equation (8.23). That is
(8.84)
where a¡ and a2, the Rayleigh constants, are fixed for a given dynamic system.
vVhen the latter equation is premultiplied by xr and postmultíplied by X, then
(8.85)
With the orthogonal properties of equations (8.82) and (8.83), the last result
becomes
(8.86)
218 MUL11-DEGREE OF FREEDOM LINEAR STRUCTURES
The nth modal damping factor (n is defined in terms of the nth diagonal term
in equation (8.86), where
(8.87)
(8.88)
This last result shows that, for an N degree of freedom system for which
the N frequencies w 1 ,w2, ... ,WN are known, then the two Rayleigh constants
a1 and a2 are uniquely determined if any two values of (n are specified. For
instance, if (k and Crn are specified, then equation (8.88) yields the following
two simultaneous equations from which a 1 and a2 can be calculated:
(8.89)
Then the remaining N 2 values of ( n for all n :fe m can then be calculated
from equation (8.88). In many applications, the first few modes will domínate
the motion, and in such cases it is reasonable to choose ( 1 and ( 2 ( or k 1
and m 2) as the arbitrary numerical damping factors. Note that, as Wn
(and thus n) becomes large, C:n --+ wna 1 /2 and for all practica! purposes the
damping increases linearly with frequency. This is consistent with experimental
observations in which the higher modes have diminished amplitudes and are
difficult to detect because of the increased damping at high frequencies.
Uncoupling the Equations of Motion
Consider the governing equations (8.1) in which the externalloading vector
p is an arbitrary function of time at each nodal point, or
(8.90)
Define the solution vector for equation (8.90) by the following modal coordinate
transformation:
N
e= Xy or en LXnk Yk (8.91)
k=l
In this last equation, it is noted that the coefficients of y, y, and y are defined
by equations (8.82), (8.87), and (8.83), respectively, which leads to
(8.94)
(8.95)
Note that the right side of the last equation is a scalar quantity.
Steady State Solutions
It is recognized that the scalar equation (8.95) is in the form of the single
degree of freedom model, equation (5.58). The steady state Duhamel integral
solution, equation (5.56), was derived for the latter model in Chapter 5. For the
present case, this integral solution can be derived in the same way. The result
is
Yn r)] dr (8.96)
(8.97)
(8.98)
From the transformation equation (8.91), the steady state solution for the modal
coordinates in component form is thus
[:~ l~ [ ][
xu
l
X12 X1N Y1
X21 X22 X2N Y2
(8.99)
Summarized below is a ten-step procedure for obtaining the steady state modal
response solutions to the governing equation (8.1):
l. After formulating the N-degree of freedom model of the structure, deter-
mine numerical values for the elements of the M and K matrices.
2. Evaluate the loading vector p(t).
220 MULTI-DEGREE OF FREEDOM LINEAR STRUCTURES
PROBLEMS
8.1. Consider a two degree of freedom lumped mass model of the jackup
rig shown in Fig. 2.17 and described in Example Problem 2.8. Choose the two
independent COOrdinateS as V and e, the displacement and rotation of the deck,
respectively. Let Pd and A1d be the arbitrary loadings corresponding to v and
e. For the deck, the mass moment of inertia about its mass center is Jd. Neglect
all changes in elevation of the deck. Include damping.
(a) \Vrite clown the coordinate vector, the loading vector, and the damping
matrix.
(b) Define in words and symbols the two nonzero elements of the mass ma-
trix.
(e) Each leg of the three-legged structure has a bending stiffness El anda
length e. Use classical beam theory to compute the elements of the 2 X 2 stiffness
matrix.
(d) Show a free body sketch for each of the two masses.
(e) Use Newton's method to formulate the two equations of motion in terms
of V ande. Include linear viscous damping in the form of equation (8.20).
8.2. For the structure described in Problem 8.1, formulate the expressions
for the kinetic energy, the potential energy, and the virtual work of the noncon-
servative forces. vVith these three scalar quantities, use Lagrange's method to
derive the two equations of motion for the structure.
REFERENCES 221
8.3 Suppose that the two frequencies in free vibration for the structural
model of Problem 8.1 are computed as w 1 and w2 and that the modal damping
factors are estimated from experimental data as ( 1 = 0.05 and ( 2 = 0.07. \Vith
these damping factors, compute the values of the Rayleigh constants a 1 and a2
in terms of the two frequencies.
8.4 Suppose that the jacket template structure shown in Figure 8.3a is
modeled as a four degree of freedom damped system with the independent coor-
dinates (v 1,v2, fh, 02). Here, the first two coordinates are defined in Figure 8.3b;
the angles 01 and 02 are the rotations of the deck and the mass m 2 ; and J 1 and
h are the mass moments of inertia for the deck and for m 2 , respective!y. There
are three loads only: the respective horizontal wave loading p 1 and p 2 at v1 and
v2, anda wave slamming moment A1d on the deck.
(a) Write down the loading vector and the damping matrix.
(b) Identify the elements of the mass matrix. Inelude the virtual mass where
appropriate.
(e) Sketch the four diagrams, analogous to the three diagrams of Figure 8.4,
that define the elements of the stiffness matrix.
(d) For each of the two lumped masses, construct a free body sketch, anal-
ogous to the sketches in Figure 8.3c.
(e) Use Newton's method to derive the four equations of motion for this
structural model.
8.5 For the structure described in Problem 8.4, formulate the expressions
for the kinetic energy, the potential energy, and the virtual work of the noncon-
servative forces. With these three scalar quantities, use Lagrange's method to
derive the four equations of motion for the structure.
REFERENCES
Caughey, T. K., and O'Kelly, M. E. J., Classical Normal Modes in Damped Linear
Dynamic Systems, Journal of Applíed Mechanics 32, 1965.
Clough, R.\V., and Penzien, J., Dynamics of Structures, second ed., McGraw-Hill,
New York, 1993.
SAP 2000, Integrated Structural Analysis and Design Software, Computers and Struc-
tures, Inc., Berkeley, CA, 1997.
222 MULTI-DEGREE OF FREEDOA1 LINEAR STRUCTURES
Synge, J. L., and Griffith, B. A., Principles of Mechanics, third ed., McGraw-Hill,
New York, 1959.
Whittaker, E. T., A Treatise on the Dynamics of Particles and Rigíd Bodies, fourth
ed., Cambridge University Press, Cambridge, UK, 1937. Reprinted 1989.
9
Applications of Multi-Degree
of Freedom Analysis
James F. Wilson
224 APPLICATIONS OF MULTI-DEGREE OF FREEDOM ANALYSIS
Mathematical Model
A fixed leg platform or jacket template structure, discussed in sorne detail in
Example Problems 8.1-8.5, is shown again in Figure 9.1. As in these previous
examples, the stalk model of this structure is used, as defined in Figure 9.2a,
a two degree of freedom model with the lumped virtual masses m 1 and m2,
located by the two independent horizontal displacement coordinates (~ 1 ,~ 2 ).
The corresponding equations of motion for each mass, derived previously as
equations (8.28), are
C12 ] [ ~1 ]
C22 ~2
] (9.2)
Listed in Table 9.1 are the numerical values for the characteristics of the struc-
ture and of the incident harmonic water wave. The analysis of the structure's
dynamic response begins with a calculation of the undamped natural frequencies
and mode shapes.
D i..., .... }-
D~- .....
+ w
I
1 w J
1
l
SECTION A-A
(a) ¡
Figure 9.2 Stalk model of the fixed leg platform: (a) definition of coordinates; (b)
components of the modal vector { 1 ; (e) components of the modal vector { 2 .
Frequencies
The two undamped structural frequencies for free vibration are computed
by solving the characteristic determinant, equation (8.61), for the two positive
roots w. The governing equation is
(9.3)
which reduces to
Using the quadratic formula, the two positive roots w2 to this equation are
calculated as: wi
= 7.322 rad2 js2 and w~ 123.04 rad2 js2 , from which the two
frequencies are deduced as
(K o (9.7)
m this case, en = ¡1 €2 nY
ror n 1, 2 and the components €2n can be com-
puted from the special case of equation (8.63), or
(9.8)
(9.9)
The reader can verify that the same numerical results for ~ 2 n can be obtained
from the second of equations (9.8) as from equation (9.9). For the numerical
values of this problem, then ~ 21 = 0.341 and ~ 22 = -4.38, which correspond to
the frequencies of w1 = 2.706 and w2 = 11.09 radjs, respectively. These modal
vectors, which have no units, are thus
(9.10a)
(9.10b)
e e
Shown in Figures 9.2b and 9.2c are sketches of 1 and 2 , respectively. :.Jote that
sin ce the components ~ 11 and ~ 12 were arbitrarily chosen as unity, a comparison
of magnitude between these two vectors is not meaningful.
The normalized modal vectors Xn are computed using equations (8.66) and
(8.67). That is,
(9.11)
A FIXED LEG PLATFORM 227
en=~nM~n (9.12)
e2
1 [1 0.341] [
4 9
·g 3.~3 ] [ 0.~41 ]
X 106 5.054 X 106 kg
e¡ 2248 kg112
o
e~= [1
€2
- 4.39] [
= 8063 kg 112
·g
4 9
3.13 ][ 1
-4.39
] X 106 65.01 X 106 kg
With these results and equation (9.11), the two normalized modal vectors are
calculated as
(9.13a)
The modal shape matrix X, defined previously by equations (8. 79) and (8.80)
as the assembly of the modal vectors Xn, is thus
4.45
X= [ xu X12 ] =[ 1.52
(9.14)
X21 X22
Example Problem 3. 2, the wave length A can be expressed in terms of the water
depth d, the wave perlad T, and the height H. In this case, the dispersion
relation has the following form:
9.81 27r. 61
A T 15.4 7r A tanh ->.- (9.15)
2
The latter equation, implicit in A, was solved using Mathematica® (1999) with
the subroutine RootFind, with the result that >. = 312 m. From this, the wave
number becomes k 21r / >. = 0.0201 m-l. These two wave parameters are listed
in Table 9.1.
In computing the structuralloading associated with this wave, the following
assumptions are made: ( 1) the motion of the structure is much smaller than
the motion of the wave, so that Morison's equation (2.14) applies; (2) the ftow
is predominately in the inertia regime so that the structural loading term of
Morison's equation that involves CM dominates the fluid drag term that involves
CD; (3) the four vertical legs (Ne 4) plus the two horizontal cross braces
(Nc 2), which are normal to the ftow at node point 2, account for most
of the structural wave loading; (4) because of their relatively small diameter
compared to the legs, the wave loading of the cross bracings is mainly fluid
drag, a loading that is relatively small compared to the inertia loading on the
other six members to which CM applies; (5) since the wave length A = 312
m is much larger than the distance w = 30 m between the vertical legs in
the direction of wave propagation, the phase of the wave can be neglected, or
x = O in the expression for the wave acceleration ü . \Vith these assumptions,
ü = ü(z, t) of Table 3.2 has the form
21r2 H cosh k(z + d) .
Ü= sinh kd sm wt (9.16)
The loading per unit length of the four vertical legs, and of the two cross
braces normal to the wave direction are, respectively
(9.17a)
(9.17b)
The respective total loads lumped at nodes 1 and 2 are computed by integrat-
ing these loadings over the appropriate structural members. As a conservative
measure, all of the wave loading on the four legs from the sea surface to node 2
is lumped at nade 1 located at the deck level. Also, all of the wave loading on
the legs extending from node 2 to the sea floor is lumped at nade 2. Note that
the two cross braces normal to the flow are located node 2 also. With these
assumptions, the nodal loads can be expressed as follows:
-(d-ez)
P2(t)
1 -d iie(z, t)dz + wifc[-(d (9.18b)
When u of equation (9.16) is substituted into equations (9.17), and those results
are integrated according to equations (9.18), the nodalloads become
1 1
b3
2
k a1a3(sinh 2kd sinh 2kf2); b4
2
k a¡aa sinh 2ke2 + wa3a4 cosh 2ke2
(9.20)
These coefficients, when evaluated using the system parameters of Table 9.1,
lead to the following explicit results for the nodalloads, in units of newtons:
P2(t) -6.534 x 106 sin 0.408t 0.432 x 106 sin 0.816t N (9.21b)
With this loading, together with the normalized vectors Xn, the steady state
solutions to the governing equations of motion (9.2) can be computed using
equations (8.96)-(8.99) and the procedure outlined in Section 8.6. The two
scalar products in the integral solution (8.96) are computed using the vectors
of equations (9.13) and (9.21), or
With these results, y 1 = y 1 (t) and Y2 = Y2(t) were computed by numerical inte-
gration of equation (8.96) at each of the following times t = 0.2, 0.4, 0.6, ... , 30.0
S. For each of these times, the structural displacements el
and e2 were cal-
culated using the transformation of equation (8.99), or equivalently equation
(8.91). That is
(9.24a)
(9.24b)
S
.
~
¡z¡
~o.
2
r/J.
¡;a -0.1
TIME t, seconds
Figure 9.3 Responses for the fixed leg platform toa harmonic wave.
Shown in Figure 9.3 are numerical results for these two structural steady
state displacements over a time of 30 s, or for approximately two cycles of the
wave loading. The absolute values of the first peaks are:
and the respective subsequent peaks change very little from these values. Also,
after the first cycle the responses become more smooth, a result of the light
damping.
This same problem was a1so solved using linear wave theory instead of Stoke's
second order wave theory. For linear wave theory, the wave loading is given by
equations (9.19) with b3 = b4 = O, and with b1 and b2 defined by equations
A FIXED LEC PLATFORM 231
(9.20) as before. Computations showed that the response curves were nearly
identical to those of Figure 9.3, and that the absolute values of the first peaks
were:
For this problem, then, linear wave theory is adequate for preliminary dynamic
design.
Response to Earthquake Excitation
The two-mass model of the offshore platform is shown in Figure 9.4a, now
subject toa horizontal ground displacement v 9 = v 9 (t) that simulates one type
of earthquake excitation. As shown in Figure 9.4b, the coordinates ~ 1 and ~ 2
now represent the displacements of m 1 and m 2 relative to the rigid base of the
structure, and it is these displacements that give rise to the elastic restoring
force and the damping force. In the absence of other external excitations, the
equations of motion are derived from the general equations (9.1) by replacing
e
the acceleration vector for each lumped mass by the new absolute acceleration
vector (e+ 1v9 ), in which the unit vector in this two-mass example is
1 = [1 (9.27)
In this last result, the negative sign on the right was omitted since the sign of
~ is of no consequence in this problem. Further, this last result also applies to
stalk models in plane motion with horizontal, rigid base excitation in which the
number of degrees of freedom N > 2, provided that the unit vector 1 has the
same dimension as N. For stalk models that include soil-structural interactions,
see Clough and Penzien (1993), Chapter 27.
(b)
(9.29)
The peak values of the ~ vector, or emax, are sought for a given experimental
earthquake time history ii 9 . This is done by fust forming the product WnYn and
then computing the maximum value of the above integral for each frequency
Wn· That is, compute the pseudovelocity of the nth mode, defined by
(9.30)
As a numerical example, let the structure of Figure 9.4 have the character-
istics given in Table 9.1. The undamped frequencies Wn and the modal shape
matrix X are given by equations (9.6) and (9.14), respectively. Choose the El
Centro earthquake as the design condition since the pseudovelocity Snt; has been
computed for this case as a function of damping and structural or modal period
To Tn (see Figure 5.8). Compute To = Tn for each frequency as
2rr 2rr 2rr
Tt = Wt = 2.706 = 2.32 s; T2 = 11.09 = 0.567 S (9.32)
Let ~ 1 ~ 2 0.05 and use Figure 5.8 to obtain the respective pseudovelocities
for these two periods:
~1
rnax ]
[ ~2max
[ 4.45
1.52
1.24 ]
-5.44 X
10-4 kg -1;2 [ 0.235
Ü ~.0687 ] m x
0.258 ] m (9 _34)
[ 0.133
Thus, the displacement of the deck at mass m 1 is 0.285 m, and the displace-
ment at the 38 m height at mass m2 is 0.133 m, both relative to the bottom of
A }.;IONOPOD GRAVfTY PLATFORM 233
the legs. Using these results, the maximum horizontal shear loads at m 1 and
m2 are deduced as
Ímax
0.735
[ -1.15
-1.15 ] [ 0.258 ]
3.59 0.133
X l07
[ i:; ] 5
x 10 N
(9.35)
The sum of these two horizontal shear loads is 2.20 x 106 N, which is an upper
bound of the shear load shared by all four legs at the base of the structure.
Further, an upper bound on the base overturning moment dueto these shear
loads, also shared by all four legs of lengths l\ /!2 = 38 m, is given by
(9.36)
The responses computed in this numerical example are upper bound values
since they are based on Su; and S 2t; derived from the maximum of the Duhamel
integral, irrespective of the time of occurrence. Thus, the maximum shear loads
were assumed to be in phase. Nevertheless, thls type of calculation serves a.<;
an economical, approximate check on results derived from the more involved
models where the modal displacements are matched in time.
STATIC
r----t---¡ EQUILIBRIUM
LINE --"i
F(t)
daK aK av
-av- -av+ =gv (9.37)
(9.39)
\Vhen the last three equations are used with equations (9.37), and then with
(9.38), the respective equations for plane motion become
(9.42)
The above two equations can be written in standard form in the following way:
Multiply equation (9.42) by (-ha) and add this result to equation (9.43) to
give
mha
[1- --(ho
Jo
(9.45)
In the last two equations, which are now in standard form, the parallel axis
theorem was used, where
Jo Ja+mh 20 (9.46)
236 APPLICATIONS OF MULTI-DEGREE OF FREEDOM ANALYSIS
[~ Jo
Ja
-hcc1
e
1
_ rnha C()
Ja
C() ][~ ]
-~:a (ko- moghc + mbghb)
(ko - moghc + mbghb)
= [ [1 - ~:e (+
ho - he) J F - ~:a M pe
Mpc (ho + hc)F
]
(9.47)
(9.48)
With the numerical parameters of Table 9.2 substituted for K and M in the
governing equations (9.47), this determinant becomes
1
(3.59- 0.114wo- 0.615w 2) x 109
( -66.3 + 2.1lwo) x 109
(0.0772 + 0.0107w 0 ) x 10 12
(3.60- 0.497w 0 - 0.880w 2 ) x 10 12
¡--
0
(9.49)
Assume that the soil foundation stiffnesses k1 and ko are especially sensitive
to the structure's fundamental frequency w1 , which is the smallest positive root
that satisfies this determinant. It is logical then to let w 0 = w 1 . With this con-
dition imposed, a trial and success procedure is used to extract the frequencies
from equation (9.49). As a first trial, choose w 0 = 1.41 radjs, the result ob-
tained for this same structure under pure rocking motion. See Example Problem
5. 2. The characteristic determinant then reduces to
Using the quadratic formula, the smallest positive root of this polynomial is
calculated as w1 = 1.23 rad/s. Next, choose wo = 1.24 rad/s, for which the
polynomial from equation (9.49), when solved for the lowest two positive roots,
yields the convergent results to three significant figures, or
n 1,2 (9.53)
in which the component form of the nonnormalized modal vector is
(9.54)
The last two equations, when combined and written in component form, are
(9.55)
With this equation, with the numerical '\talues given in equations (9.51), (9.52)
and Table 9.2, and with ~ 11 = ~ 12 1, the two modal vectors are evaluated as
el [~11
A
.;21]
T
[1 0.039üf' (9.57a)
\ 1
\ 1
\ 1
\ 1
\ 1
\ 1
1
' \ 1
+v +V
MODE 1 MODE 2
Figure 9.6 Mode shape envelopes for the monopod gravity platform.
238 APPLICATIONS OF MULTI-DEGREE OF FREEDOM ANALYSIS
The modal vectors of the last two equations are shown in Figure 9.6. For
mode 1 corresponding to the lowest frequency, both components of el are pos-
itive, which indicates a positive or right displacement for v (the ~ 11 term) and
a positive or clockwise rotation for () (the ~ 21 term), signs which are consistent
with the chosen coordinate directions. The broken line is also possible sínce
~ 11 could have been chosen as ( -1) instead of ( + 1), with the result that both
v and e would then be negative. The motions of mode 1 are in phase since
e
both components of 1 have the same sign. The motions of mode 2 are out of
e
phase since the signs of the 2 components are always opposite. In free vibra-
tion, the actual motion is a combination of both mode shapes, which depend
on the structure's initial displacement and velocity. Once in motion, the mode
shapes continually change as potential and kinetic energies are transferred in
the foundation restraints.
With the frequencies and mode shapes in hand, the deterministic responses
of this monotower to given time histories of loading, F(t) and Mpc(t), can be
calculated in a straightforward way using the normal mode solutions derived
in Chapter 8 and illustrated for the fixed leg platform at the beginning of this
chapter.
This particular problem gives sorne interesting insights into the frequency
behavior of multi-degree of freedom systems. For instance, it is observed that
the fundamental undamped frequency is depressed by about 12 percent, from
1.41 rad/s for its single degree of freedom counterpart of Example Problem 5.2,
to w1 = 1.24 rad/s for the two degree of freedom analysis. This is characteristic
of linear systems: as more flexibility is incorporated by allowing more degrees
of freedom, w 1 decreases. In either model, the effect of foundation damping
and viscous damping due to the surrounding water is to depress the undamped
frequency.
Nataraja and Kirk (1977) analyzed a silnilar structure. They calculated
a value of 1.02 rad/s for the fundamental frequency of a three-legged gravity
platform modeled to include leg flexibility and the same type of soíl foundation
elasticity ( k 1 and ke) -as used here. Three factors in the Nataraja and Kirk
model account for their lower value of w1 : their structural mass was somewhat
higher, their model included soil foundation damping, and their model had
added degrees of freedom because of leg flexibility.
Dynamic Stability
There is a vast literature on criteria and methods for determining the dy-
namic stability of coupled, linear systems such as the gravity platform. The
classical works of Liapunov (1907) and Ziegler (1956) are especially noteworthy.
For present purposes, however, the criteria are relatively simple. The dynamic
stability of a linear, undamped structure in free vibration about its static equi-
librium position can be tested by investigating the nature of the roots wr as
calculated from the characteristic determinant, equation (9.3). If every wt ís
real and positive, then the characteristic frequencies given by Wi JW't are real
and positive, and the system undergoes stable, bounded, harmonic oscillations.
STRUCTURAL RESPONSE STATISTICS 239
(9.59)
wr
which is unbounded as time increases. If is complex, one of the roots produces
this same unboundedness. Computer codes that extract roots of a determinant,
such as subroutine Eigenvalues of Mathematica® (1999), are generally available
to the engineer, so it is a relatively straightforward task to check the dynamic
stability of a structure.
With respect to a gravity platform on an elastic soil foundation, dynamic
instability leading to toppling would occur for critica! combinations of its mass,
its center of buoyancy, its center of gravity, and its foundation stiffness. For
instance, for sufficiently high val ues of platform mass and ha, it is visualized that
a moderately weak soil foundation could offer an insufficient restoring moment to
resist both the angular momentum of the structure and the overturning moment
due to its deadweight. Then the structure would topple.
In conclusion, it is noted that if the criterion for dynamic stability is observed
(that all values of wr are real and positive), then the static criteria discussed
in Chapter 1, applied to gravity platforms, will also be satisfied. It is now well
recognized that a dynamic stability analysis includes the results obtained from
a static stability analysis; but the converse is not true.
As discussed in Chapters 6 and 7, the wave data available to the analyst and
designer of offshore structures are most often in the form of a surface wave
height spectrum, Sr¡(w). In Chapter 6 a method was presented for representing
this spectrum as harmonic waves forms. These forms, when converted to struc-
tural forces through suitable transfer functions G(w), are then used to calculate
the time history of structural response. An alternative approach was discussed
in Chapter 7 where the spectral density of the response and its variance were
calculated directly from Sr¡(w) and G(w) for the single degree of freedom case.
This latter approach is now employed for the case of a linear structure with
N degrees of freedom. Summarized in the following ten steps are the critica!
assumptions and the methodology leading to an analytical form for the spectral
density defined as S(~k' w) and the variance u 2({k), calculated for each indepen-
dent physical coordinate ~k. The load transfer functions in this case are defined
by the vector G(p,w).
l. Set up the mathematical model and derive the equations of motion for
the structure in the form of equation (9.1). Identify the coefficient matrices M
andK.
240 APPLICATIONS OF MULTI-DEGREE OF FREEDOM ANALYSIS
2. Calculate each component Pk(t) of the loading vector p(t). Select linear
wave theory. Use Morison's equation and diffraction theory where appropriate
to determine the components of loading.
3. Calculate each component function G(pk, w) of the load transfer vector
G(p,w).
4. Calculate the undamped frequencies wk and the normalized mode shape
matrix X. The consecutive columns of X are modal vectors x 1, x2, ... , XN. The
kth modal vector has components x1k, X2k, ... , X N k·
5. Define the generalized loading vector in modal coordinates as q(t) = q.
Then calculate the corresponding transfer functions G (iik, w) for each component
iik of q where
iik x[ p(t) (9.60)
Thus the generalized load transfer function for the kth mode is
(9.62)
Calculate Hk(w), the harmonic response function for the kth mode, by substi-
tuting the following quantities into the last equation:
(9.63)
The results, including the modulus, are
(9.64)
(9.65)
7. Assume that p(t) is a stationary ergodic process. It follows that q(t)
will be stationary and ergodic also since the components of p(t) and q(t) are
related by the linear transformation (9.60). Using the analysis in Chapter 7 for
the single degree of freedom system in the form of equation (9.62), it follows
that the spectral density of Yk is
(9.66)
Here S (iik, w) is the spectral density of the generalized force component iJ.k.
8. Assume linear wave theory. From Example Problem 1. 2, deduce the
following analogous relationship between the spectral density of the kth load,
S(pk,w), the wave height spectrum S 11 (w), and the transfer function G(pk,w):
(9.67)
STRUCTURAL RESPONSE STATISTICS 241
~k L XknYn (9.72)
n=l
It follows that
In the sums of this last result, there are N autocorrelation functions of the form
The following two results are based on the definition from equation (7.20):
S(~k,w) = 1
r: R(E;,k,r)e-jw-rdT (9.77)
242 APPLICATIONS OF MULTI-DEGREE OF FREEDOM ANALYSIS
-1
21T
¡=
-00
.
R,.,(r)e-Jwr dr (9.78)
Now substitute equation (9.65) into (9. 70), change the index from k to n, and
substitute this rewritten forro of S(yn,w) into equation (9.79). The result is
., . ~ x~n lx~G(p,w)l
2
_
S(~k' w) - S'l(w) L.,¡ ( 2 - 2)2 + (2( , )2 (9.80)
n=l Wn W nWnW
10. Calculate the variance for each physical coordinate ek· Using the defini-
tion given by equation (7.22) and S(ek,w) of the last result, compute
(9.81)
With the wave height spectrum S11 (w) in the Pierson-Moskowitz or JON-
SWAP forro, and the limits of integration in equation (9.81) replaced by 0.16
and 1.4 radjs, the variance of each coordinate can be obtained by numerical inte-
gration. Assuming that p(t) is Gaussian with zero mean, then the responses will
also be Gaussian with zero mean, since the system is linear. Thus, the rms value
of the kth physical coordinate is given by the square root of equation (9.81).
For practica! purposes, the extreme limits of ~k are ±3a(~k), k = 1, 2, ... , N.
For these extreme values, if the static stresses and deflections of the members
are within the allowable limits, then the structure is assumed saje, without con-
sidering other loadings and fatigue failure. For design purposes, it is generally
acceptable to superimpose the effects of the static or steady drag loadings due
to winds and currents.
The statistical responses. derived in the last section are now computed for the
two degree of freedom jacket template platform modeled in Figures 9.1 and 9.2.
The numerical parameters of this platform are summarized in Table 9.1. The
results of the first three steps of this calculation were obtained in Section 9.1
and are summarized as follows in terms of these numerical parameters:
l. The mathematical model is defined by the following two equations of
motion. The forro of damping is defined later.
o
3.13 X 106 ][i~ ]
A FIXED LEG PLATFORM: STATISTICAL RESPONSES 243
7.35 X 107 ~1
8
-1.15 X 108 ] [ ] = [ PI ] (9.82)
+[ -1.15 X 108 3.59 X 10 ~2 P2
2. Assume that linear wave is sufficiently accurate for this analysís. The wave
loading vector is thus given by equations (9.19) and (9.20) with b3 b4 =O, or
2
pP21 ] = -a:.,.---:-w-:-H--e--:- [ sinh 61k - sinh 38k ] sin wt (9.83)
[ sinh61k sinh38k + j3kcosh38k
Here, the numerical values of the constants a and ,8 are
NeCArgPDe
7f 2
= 97980kg/m; j3 = Dc)
Ne (De
2
9.169 m (9.84)
From each load component, the corresponding component of the transfer func-
tion is deduced from the definition that the real part of G(p, w) is equal to
p(t)j H, or
2
. w [ sinh61k sinh38k ] jwt (9.85)
Ja ksinh6lk sinh38k + .Bkcosh38k e
3. The two undamped frequencies and the normalized modal matrix, given
by the respective equations (9.6) and (9.14), are
X=[ X12 ]
X22
[ 4.45
1.52
(9.87)
With the results of these three steps, the statistical responses are then cal-
culated directly from steps 9 and 10, or equations (9.80) and (9.81) of Section
9.3. These calculations do require a wave height spectrum, which is now chosen
as the Pierson-Moskowitz spectrum of equation (7.59), with a corresponding
significant wave height of Hs 15 m. That is, for g 9.81 mjs2 ,
= -O. 780 m 2 · s j r ad
4
S 11 (w ) 5
-e -0 ·0138/w (9.88)
w
Using the normalized modal vectors of X, equation (9.87), the components
of the transformed transfer function in equation (9.80) are computed as follows:
The squares of the moduli of these last two results are, respectively,
2
lxfG(p,w)l = (4.45P + 1.52Q) 2 (9.91)
2 2
lxfG(p,w)l = (1.24P- 5.44Q) (9.92)
aw 2 X 10- 4
P = ksinh k (sinh61k- sinh 38k) (9.93)
61
aw 2 X 10- 4
Q= ksinh k (sinh 38k + f3k cosh 38k) (9.94)
61
Note that a and f3 are constants given by equations (9.84). Also, the wave
number k is related to the wave frequency w by dispersion equation (3.16),
which in this example is
120
"O
!:!
---"'
N
S
~
~
'-'
~
Vo
.....
~r
E-<
u¡¡.:¡
o..
en
¡¡.:¡
en
z
oo..
r:/l
00
FREQUENCY, ro radls
Figure 9.7 Response spectra for the displacement coordinates of the fixed leg
platform.
(9.98)
(9.99)
Assuming a Gaussian process, there is only a 0.026 percent chance that each
response exceeds the ±3a limits, or that .; 1 exceeds 0.250 m and that .;2 exceeds
0.0878 m. Note that these latter two displacements of 0.250 m and 0.0878 m
are close to the respective values of ,; 1 = 0.1937 m and .; 2 = 0.0805 m derived
in Section 9.1 as the peak responses toa single incident harmonic wave with a
significant wave height of 15 m.
246 APPLICATIONS OF MULTI-DEGREE OF FREEDOM ANALYSIS
PROBLEMS
9.1 For a water depth of d 300m, plot the value of wave number k as
a function of wave frequency w where 0.1 < w < 2 radjs. Use the relationship
w 2 = gk tanh kd. For what range of frequency is k given by w 2 / g to within 5
percent?
9.2 Calculate the damping matrix C for the fixed leg platform of Section
9.1. Assume ( 1 ( 2 0.07. Indicate the units of C.
9.3 Derive in detail the result of equation (9.28), the mathematical model
for an N degree of freedom stalk structure subjected toa horizontal base exci-
tation. Why is the negative sign on the right of equation (9.28) ignored in the
response solution of equation (9.30)?
9.4 For the gravity platform modeled in Section 9.2, calculate X and show
the units for each component vector Xi.
9.5 Suppose that the gravity monotower of Section 9.2 is subjected to the
horizontal El Centro earthquake. Calculate the maximum values of V and e'
the peak absolute value of horizontal displacement for the deck, the peak shear
force at the base, and the peak overturning moment. Assume that the modal
damping for each mode is 0.05.
9.6 Show that if one of the frequencies Wi of the characteristic determinant
of an undamped linear system is a complex number, then the system's behavior
is divergent.
9. 7 For an N degree of freedom linear structure, vvrite a computer program
to calculate the coordinate displacement spectra, the corresponding variances,
and the rms values. Use the results of equations (9.80) and (9.81) where the
following quantities are specified as input: S 71 (w), X, G(p,w), (k, and Wk, k=
1, 2, ... , N. As a numerical example, check the results for S(,k,w) and a((k)
obtained for the linear structure in Section 9.4, for k 1, 2.
REFERENCES
Clough, R. W., and Penzien, J., Dynamics of Structures, second ed., McGraw-Hill,
New York, 1993.
Mansour, A. E., and Millman, D. N., Dynamic Random Analysis of Fixed Offshore
Platforms, OTC-2049, Proceedings of the Offshore Technology Conference, 1974.
~ataraja, R., and Kirk, C. L., Dynamic Response of a Gravity Platform Under
Random Wave Forces, OTC-2904, Proceedings of the Offshore Technology Con-
ference, 1977.
REFERENCES 247
Ziegler, H., On the Concept of Elastic Stability, Advances in Applied Mechanics, Vol.
4, H. L. Dryden and T. von Karman, editors, Academic Press, New York, 1956.
10
Continuous Systems
James R Wilson
The dynamic models of offshore structures discussed so far have involved only
a finite number of independent coordinates and ordinary differential equations
of motion. In the single degree of freedom systems, one coordinate was chosen
to describe the dominant structural vibration mode in the plane. In the multi-
degree of freedom systems, examples included the rigid gravity platform with
coordinates to describe sliding and rocking motion andjacket template platforms
with coordinates to describe the motion of discrete masses lumped at node
points.
For line components such as rather long beams, pipelines, and cables, alter-
native continuous system models may provide more precise and sometimes more
economical descriptions of component motion. Since a partial differential equa-
tion is used to characterize the motion of a continuous line component, solutions
are generally more involved mathematically than for a corresponding lumped
system. However, if the continuous models are chosen judiciously, closed form
expressions can be derived for the characteristic frequencies and mode shapes of
line components, which then lead to upper and lower bounds on their dynamic
responses.
Two classes of continuous line components are analyzed in this chapter. The
first component is designated as a beam for which bending stiffness and longi-
tudinal tension are incorporated in the model. Examples include: pipelines
for dredging manganese nodules from the sea fioor as depicted in Figure l. lOa;
pipelines for ocean thermal energy conversion as shown in Figure l.lüb; gath-
ering lines and risers; and the long structural bracing members of the various
offshore platforms. The second structural component considered here is the ca-
ble which resists tension but whose bending stiffness is negligible. Examples
include the steel and synthetic fiber rapes and steel chains used to stay buoys,
fioating platforms, compliant towers, and ships.
Excitation of line components comes about in severa! ways. For instance,
a flexible cylinder in a steady current may undergo adverse transverse motion
and can be destroyed by the periodic shedding of vortices. This can occur if
one of the lower natural frequencies of the cylinder is coincident with that of
the vortices. There is also direct transverse excitation due to waves, and there
248
CONTINUOUS SYSTEMS 249
is end excitation of line components attached to ships and towers that move
with the waves. Two types of end motion for a mooring line, transverse and
longitudinal (or parametric) excitation, are depicted in Figure 10.1.
/ Ship Motion
Dlf)./
'\
''Dlfl
......,....,...""1""7
Water Line
1=-; X
•
Transverse Excitation
L .
X
... . Dit)
Parametric Excitation
Governing Equations
The beam or cable rnodel is shown in Figure 10.2. This line cornponent ha.'>
a virtual rnass per unit length of m, a length e, and a fixed longitudinal axis
x intersecting the end points. The transverse dynamic displacernent frorn its
equilíbriurn position at v =O is v = v(x, t) which is assurned to be srnall enough
so that its slope () &v / &x ís always rnuch less than one. A further assurnption,
which is also consistent with classical beam theory, is that transverse planes
at equilibriurn, or v = O, rernaín planes during rnotíon, when v # O. The
infinitesimal elernent of length dx shows the bending rnornent M, the transverse
shear load V, and the tension load P, all of which are subject to small changes
across the elernent. The transverse excitation force per unit length is q = q(x, t),
and the linear damping force per unit length is cv, which correspond to the
average elernent loads qdx and cvdx, respectively, acting at the elernent's center.
For a srnall slope, first-order changes in (), A1, V, and P are appropriate. That
is
av 82 v
1\J + d.A1
aMd
+--X
() + d() ::::: ax + axz dx ; M
ax
av aP
V+ dV ::::: V+ &x dx; P + dP ::::: P + ax dx (10.1)
ij (x,t)
~~~:~ty
//
/
X
Element,
\'
shown below
ijdx
(10.3)
Next sum moments about point O, the lower left comer of the element,
and equate this sum to zero. With this approximation, rotational inertia, or
rotatory inertia as it is sometimos called, is neglected. For the relatively low
frequencies encountered in beams and cables of ocean structures (below 100Hz),
the rotational energy is much less than that due to transverso motion, which
justifies this assumption. It follows that
the terms involving q, e, and dV are multiplied by (dx )2 , and are thus
neglected as higher-order terms. Equation (10.4) thus yields the shear load as
V= fNvf (10.5)
ax
For elastic members, elementary beam theory gives
(10.6)
(10.7)
When the last result is combined with equation (10.3), the final result is ob-
tained, or
(10.8)
For submerged beams or cables, a more realistic form of damping than the
linear approximation c(8vj8t) in equation (10.8) is velocity squared damping,
defined by
82 v 82 v
-P 8x2 +m 8t2 ij(x,t) (10.11)
Here m can vary with the longitudinal dimension, but it is constant for most
applications.
2. For an undamped beam of constant stiffness El, a constant m, and a
tension P which is independent of x, equation (10.8) becomes
El 84v (10.12)
8x 4
The procedures for calculating the free, undamped frequencies and mode shapes
for cables and beams described by these last two models are now illustrated.
Cable Frequencies and Mode Shapes
Consider the free, undamped vibrations of a flexible cable of constant iñ
and constant tension P Po. The corresponding equation of motion is deduced
A10DELING BEAMS AND CABLES 253
EPv
Poaxz (10.13)
Let X = X ( x) be the general expression for the mode shape, and assume
harmonic motion with w denoting the frequency parameter, or
V= Xejwt (10.14)
where (') denotes the operator djdx. The frequency parameter is given by
w "(~ (10.16)
where '! has yet to be calculated. With each end of the cable fixed, then
v(O, t) =O or X(O) O·
'
v(i, t) O or X(i) o (10.17)
where the end conditions on X are determined from the end conditions on v
through equation (10.14).
The general solution of equation (10.15) in terms of two arbitrary constants
DI and Dz is
When each condition of equation (10.17) is applied to equation (10.18), the two
results are
(10.20)
It follows from equation (10.19) that D 2 = O. With this, the last equation
implies either the trivial result that DI O or that no motion exists. The
alternative is that sin "(i O, which leads to
n7r
"( = e' n = 1, 2, ... (10.21)
With this result and equation (10.16), it follows that the cable frequencies are
given by
- n7r ¡p¡;
Wn- f Vfh' n= 1,2, ... (10.22)
254 CONTINUOUS SYSTEMS
where the subscript non the frequency parameter is a reminder that there are
multiple frequencies. For each characteristic value "( there is a frequency Wn and
a corresponding mode shape X = Xn given by equations (10.18) and (10.22),
where D1 = Cn and D2 =O. That is
n1rx
Xn = Cn sin-/!- , n = 1, 2, ... (10.23)
in which the coefficients Cn are arbitrary. The first two mode shapes are shown
in Figure 10.3. It is observed that the fundamental frequency w1 corresponds
to the half sine wave, and the next highest frequency w 2 corresponds to a full
sine wave.
• X •x
l 1- l/2 + l/2 -1
Mode One Mode Two
Figure 10.3 The first two mode shapes for both a fixed end cable and a simply
supported beam.
This mathematical model predicts that ever increasing frequencies are pos-
sible as n becomes larger and larger. In reality, damping and cross-coupling
effects omitted in the mathematical model reduce these higher frequencies and
accompanying mode shapes to insignificance. Thus it is the practice of many
engineers to take n = 20 as a physically realistic upper limit for the purposes of
analysis and design of offshore cables and other continuous components as well.
Beam Frequencies and Mode Shapes
The free undamped vibrations of a uniform beam with a negligible axial
tension are described by equation (10.12) for Po = ij =O, or
(10.24)
The characteristic frequencies and mode shapes depend on the support or bound-
ary conditions. For purposes of illustration, the beam's supports are chosen to
be at x = O and x = 1! only. Further, each end is subjected to any one of the
following three sets of conditions in which Ve designates the deflection at the
end x = O or at x = 1!:
l. Simple support, or zero displacement and zero moment at a pin or roller:
Ve O; OVe"'"' O (10.26)
8x
3. No support, or zero moment and zero transverse shear at the end:
O; o (10.27)
Let X= X(x) denote the mode shape and w the frequency parameter. For
harmonic beam vibrations of the forro given by equation (10.14) applied to
equation (10.24), the characteristic beam equation becomes
X"" - a 2 X =O (10.28)
(10.29)
where D1, D2, D3, and D 4 are constants. This solution, together with the appro-
priate end conditions lead to the calculation of a, the free vibration frequencies,
and the mode shapes. This procedure is now illustrated.
(a)
Leg
(b)
l, .. l
(e) jl========f:t-x
---lloJoil
k 4---- 1
1
Figure 10.4 (a) Top view of the cross brace; (b) simple support model; (e) clamped
end model.
A lower bound on the frequencies corresponds to the brace model with the
least constraints, or with the simple supports as shown in Figure 10.4b. The
associated boundary conditions are those of equation (10.30), each applied at
each end, or
(10.35a)
(10.35b)
When equations (10.35a) and (10.35b) are added, it follows that D4 =O, from
which Dz O. The last two equations thus reduce to
Since sinh aC =/: O for aC > O, then it is necessary that sin aC = O to satisfy
equation (10.38), or
nJr
a=---¡-, n=1,2, ... (10.39)
With this last result, the beam frequencies from equation (10.29) become
_ n21r2 {Ei
Wn- f_2 Vffi' n = 1,2, ... (10.40)
in which Wn replaces w.
Since sin aC = O and thus D 3 = O from equations (10.36), the only remaining
nonzero constant is D 1 . For each Wn there is a corresponding mode shape
X = Xn at an arbitrary amplitude D1 = Cn given by equation (10.33) or
n1rx
Xn = Cn sin-C-, n = 1, 2, ... (10.41)
In this case the mode shapes are identical to those for the fixed end cable, shown
in Figure 10.3. However, the frequencies of this beam vary as n 2 rather than as
n for the cable.
Calculate next the upper bound frequencies, or those corresponding to the
brace with clamped ends, Figure 10.4c. The appropriate boundary conditions
are those of equation (10.31), each applied at each end, or
(10.43a)
(10.43b)
From these last results, the determinant of the coefficients of D1, D2, D3, and
D 4 is formed, expanded, and set equal to zero. The simple result is
(cos a:C) (cosh a:C) 1 (10.44)
It is easily verified that a: 1 = 4.730 is the first nonzero root of equation (10.44)
and that this is just 0.37 percent lower than the approximate value given for
n = 1 given by
a:nl! = (n + 0.5)rr (10.45)
For n = 2, 3, ... , equation (10.45) yields successive roots of equation (10.44)
which are even more accurate than this. Thus the upper bound frequencies for
this problem are given by equation (10.29) with (10.45), or
7r2 fEi
Wn = (n + 0.5) 2 f2V fu' n 1,2, ... (10.46)
The corresponding mode shapes for this fixed end beam are found by con-
secutively eliminating the arbitrary constants D2, D3 and D4 from equations
(10.43) and writing the result in terrns of D 1 only. For the nth mode, let D1
= Cn, and the result is
Xn(x) Cn[sina:nx sinha:nx + !3n(cosa:nx- cosha:nx)] (10.47)
where a:n is given by equation (10.45) and
(10.48)
The first two of these mode shapes are shov.rn in Figure 10.5. Unlike the mode
shapes for the cable and the simply supported beam, these shapes have zero
slope at each end.
-x
This result shows that for n = 1, the fundamental frequency for the most
constrained case is 2.25 times higher than that for the least constrained case.
Thls factor decreases rapidly as n increases, where the factor is approximately
1.2 for n = 5 and 1.1 for n = 10.
(b) \
' 1
'x'
V l 1
0
1
Figure 10.6 Transverse end excitation of a submerged cable: (a) attached toa ship;
(b) attached to a buoy.
v(O, t) O (10.50)
''? = m w2 (10.54)
Po
The general solution to equation (10.53) is given by
(10.55)
D-~ (10.56b)
- sin ;ye
1
From equation (10.54) and the expression for the natural frequencies Wn given
by equation (10.22), it follows that
we w
;ye = = n1f- (10.57)
VPofifi Wn
This solution shows that the transverse cable displacement at any point O < x <
ebecomes unbounded if the excitation frequency W coincides with Wn, since then
the term sin n1r =O in the denominator of equation (10.58). Had light damping
been included in the mathematical model from the very beginning, the peak
response for w = Wn would have been bounded, but still amplified compared to
the imposed end displacement amplitude v 0 . As discussed in Problem 10.4 at
the end of this chapter, this resonance phenomenon can be observed in a simple
laboratory experiment.
CABLE RESPONSES 261
Pararnetric Excitation
Longitudinal or parametric excitation of a relatively taut cable is depicted
in Figure 10.1b. This type of excitation may also occur in the verticallines or
chains shown in Figures 10.6 when the ship or buoy undergoes heave motion in
regular waves. In such cases the average cable tension is P0 , the amplitude of
the harmonically fluctuating force is P 1 < Po, and the excitation frequency is
w. Parametric excitation is thus defined a.q
P Po+ P 1 coswt (10.59)
The mathematical model for cable motion is chosen as equation (10.11) in which
the bending stiffness, damping, and all transverse loadings are neglected. With
equation (10.59), the equation for transverse motion is thus
EPv EJ2v
-(Po+ P1 coswt) ox2 +m ot2 O (10.60)
~ [
~ (Po+ P 1 coswt)
(n1r)2
-¡- Yn(t) +m rfly nrrx =O
dt n ] sin-f- 2
(10.63)
Since the sine term of the last equation is not zero for all values of x, then its
coefficient must vanish, or
m
rflyn(t)
dt 2 + (n1r)2
-¡- (Po+ P1 coswt) Yn(t) o (10.64)
This last equation can be transformed to a standard forro using the following
four parameters:
T wt; éi n /3 = Pr w~ (10.65)
n Po w2
(10.67)
YJ'r)
,
'
',
' '\
(a) (b)
Figure 10.7 Dynamic system behavior: (a) stable or bounded response; (b) unstable
or unbounded response.
ii,.
Figure 10.8 Haines-Strett stability plot (after Lubkin and Stokes, 1943).
unstable solutions Yn ( T). Those results are shown in Figure 10.8, which is sorne-
times called the Heines-Strett diagram. If the parameter set (O:n, 73n) líes in the
cross hatched regions, Yn ( T) is stable, but if this parameter set is elsewhere, the
responses are unstable. Further studies have shown that the addition of light
damping increases the parameter range for stability where the cross hatched
regions are extended to the broken lines of Figure 10.8.
SWAY
PLAN
VIEW
SIDE ~
VIEW~
~
..••
--
Figure 10.9 A spread mooring configuration for an LST in the Gulf of Mexico.
The data for the overall mooring restraint system and the ship's surge and
sway frequencies were given in Example Problems 2.1 O and 5. 5. Pertinent data
on bow chain No. 2 are as follows. The actual unit mass of this chain is increased
by 3 percent to give its virtual unit mass.
The first step is to calculate the natural frequencies of the bow chain. From
equation (10.22) and the preceding data, it follows that
30,000
- -- = 1.11n radjsec, n = 1, 2, · · ·
1.1 2
For resonance due to transverse excitation, one would expect peak bow chain
responses near Wn or near 1.11, 2.22, ... radjsec. Since Figure 10.11 shows no
significant peaks at these frequencies, it is concluded that there is no significant
bow chain resonance response.
264 CONTI]\¡'1JOUS SYSTEMS
- - - - March, am
············· March,pm
Sept.
- ..""? ••••••
i 120
- - - - - - March, am
"'
~ ............ March,pm
Sept.
~
.:í
Figure 10.11 Force amplitude spectra for port bow chain No. 2 of an LST (after
O'Brien and Muga, 1964).
For n = 1, then (0: 1 ;;3 1 ) = (2.92;1.46). This coordinate set líes in the cross
hatched region of Figure 10.8, and thus the transverse chain dísplacements (and
tension forces) are bounded. This same conclusion is reached for (an;i3n) based
on n = 2,3, .... However, if w 1.1 radfsec, which is at the high end of the
driving frequency for this mooring chain, then an e::: n 2 and i3n e::: 0.5n2 . For
n 1, then (0: 1 ;;3 1 ) = (1;0.5), for whích Figure 10.8 shows unstable motion
in the absence of damping, but stable motion in the presence of small, realistic
damping. For n = 2, 3, ... , the motion is stable even without damping.
The conclusion of the preceeding calculations is that the peak responses in
Figure 10.11 are not due to either transverse or parametric resonance of this
particular bow chain. How, then, can these two major response peaks recorded
in the month of :vlarch be reconciled with the single peak of wave excitation in
Figure 10.10? The answer can be found by studying the motion of the whole
ship.
Referring again to Example Problems 2.10 and 5.5, the ship motion v = v(t)
in either surge or sway can be modeled by
where the coefficients on the left side are given by equation (2.72) or (2.73),
where Po is the wave force in line with v, and where w ~ 0.65 radfsec. The
results of Chapter 5 showed that peak responses of such a nonlinear system
under harmonic excitation occur not only at the excítation frequency w 0.65
radfsec but also for wf3, which is near w0 • Since w0 = 0.144 rad/sec for sway
(Example Problem 5.5), then the peak at w 0.2 radfsec can be explained as
a subharmonic shíp response of arder one-third. The existence of the one-third
subharmonic for equation (10.68) is shown in Section 5.5, where the amplitude
parameter is shown in Figure 5.11. Thus, the lower frequency (or longer period)
ship response reflected in mooring chain No. 2 arises from the group behavior of
all the mooring lines and comes about because of the nonlinear restoring force
constant k3 of equation (10.68), during sway motion of the ship. Whether drift
currents ata frequency of about 0.22 radfsec existed during these sea tests and
also contributed to the lower resonance peak of Figure 10.11 is not known.
As for cables, submerged beams and pipelines are subjected to three main types
of excitation: transverse loading due to vortex shedding, transverse end motion,
and parametric E'xcitation. Considered fust in this section are deterministic
responses of uniform beams with common types of end supports and with an
arbitrary transverse load per unit length ij ij(x, t). Following this is a stability
study of a simply supported beam subjected to parametric excitation. This
section concludes with a calculation of the statistical responses of beams to sta-
tionary, ergodic excitation. Modal analysis is employed throughout and closed
form solutions are sought.
266 COJ\i'TINUO US SYSTE11iiS
Transverse Excitation
The undamped motion of a uniform Bernoulli-Euler beam with negligible
longitudinal tension (P = O) is given by equation (10.12), or
84v
El 8x4 + q(x, t) (10.69)
It is not difficult to visualize that the first and third end conditions of Table 10.1
represent the limits of constraints for the cross beams of offshore structures. The
second end condition is discussed in Problem 10.2 at the end of this chapter. The
fourth and fifth conditions can represent cases of support for pipeline segments
during deployment, and the last condition can model an unconstrained, floating
pipeline such as proposed to transport water along the Pacific coast of the United
States.
The free vibration mode shapes Xn = Xn(x) and their correspondíng fre-
quencíes Wn are employed in the following modal analysis of forced beam vi-
bratíons. The quantities Xn and Wn are calculated for one of the six sets of
boundary conditions listed above by using the solution to equation (10.28),
which is given by equation (10.33), together wíth the appropriate transformed
boundary conditions chosen from equations (10.30)-(10.32). The modal analysis
also requires the following two conditions of orthogonality for Xn:
l
lo XmXn dx O, for m =/= n (10.70)
(10.72)
BEAA1 RESPONSES 267
X'n111 - a 4n X n =O (10.73)
4 w2m
a·
'
= -•-
El'
i =m, n (10.74)
(10. 75)
Integrate the right side of the last equation by parts four times. The first
integration, for instance, gives the result
e(Xm X""
n - X n X'"')dx
m = [Xm X"'
n X 111 ]x=O
X nmx=f. + 1e(X'mn
X 111 - X'nX"')dx
m
1o o
(10. 76)
Integrate the integral on the right side of equation (10.76) by parts. Repeat this
procedure twice more to give
where
With equations (10.72) and (10.73), the integral on the right hand side of equa-
tion (10. 77) becomes
(10.79)
\Vhen this last equation is substituted into equation (10.69), the result is
00
Multiplying each term in this last equation by Xrn and then integrating the
result over (0, C) gives
oooo ¡;oo ¡;
El lo
O
L
n=l
XmX~"Yn(t)dx + iñ rL
lo n=l
XInXniin(t)dx = r Xm q(x, t)dx
lo
(10.84)
Assume that the series of the last result is uniformly convergent over the in-
terval (0, e), which allows for the order of integration and summation to be
interchanged. This leads to
L
00 r=
lo [ElX1nX~11 Yn(t)dx+ihXrnXnYn(t)]dx= lo XmiJ(x,t)dx
1!
(10.85)
n=l O O
With the orthogonality conditions of equations (10.70) and (10.71), each term in
the sum of equation (10.87) vanishes except for m n, in which case equation
(10.87) reduces to
in which
Pn(t) = -n
1
mvo o
le Xnif(x,t)dx (10.89)
Co = 1X~dx
e
(10.90)
Assume that the beam is initially at rest, or that v(x,O) = 8v(x,t)j8t =O.
Using equation (10.82), the initial rest position implies that Yn(O) Yn(O) =O.
The solution to equation (10.88) is in the form of the Duhamel integral for ( O,
or
Yn(t) = 1-
Wn
1t O
Pn(T) sinwn(t- r)dr (10.91)
where Po is approximated as the sum of the ballast weight (in water) applied
at the lower end (x O) and one-half of the pipeline's weight (in water). Note
that in reality the hinge at x = 1! carries all of the pipeline's weight, whereas the
hinge at x =O cardes none of the pipeline's weight. The approximation given
for Po becomes more accurate for increasingly high ratios of ballast weight to
pipeline weight.
270 CONTINUOUS SYSTEMS
p = Po + Ji cos &t
El, m, t
(a) (b)
Figure 10.12 (a) Barge-dredge pipe system; (b) model of the dredge pipe.
1r!I (10.93)
g PpV
where Pp and E are the mass density and Young's modulus for the pipeline of
length f (Timoshenko and Goodier, 1951). Applying equation (10.93) toa steel
pipeline 3000 ft in length, we find w¡; = 17.7 radjsec, whích is more than ten
times the híghest excitation frequency expected to be imparted by the barge
through wave action. (Motion of a typical barge in waves is discussed in Section
10.4.) Although we » w holds true for this steel pipeline, the inequality may fail
for pipelines manufactured of polymeric materials such &'l polyethylene. Thís
is because the ratios E/ Pp for polymeric materials are generally much smaller
than for steel.
With thís mathematical model, the governing equation of motion is then
equation (10.12) with the parametric excitation load of equation (10.92), or
84 v . 82 v 82 v
EI (Po+ P 1 coswt) ax 2 + iñ at 2 O (10.94)
The four homogeneous boundary conditions corresponding to simple end sup-
ports are
2
v(O, t) V (e ,t ) = 8 ax2
v(e,t) =
0 (10.95)
It is easily verified that all four of these latter conditions s¡üisfy the following
form chosen as a solution to equation (10.94):
00
v(x, t) ""'
L..- Yn(t) sin n1rx (10.96)
n=l
BEAM RESPONSES 271
Using the same procedure as for parametric excitation of a cable, results anal-
ogous to equations (10.63) and (10.64) are obtained when equation (10.96) is
combined with (10.94). The following parameters are then applied to the re-
sulting second order ordinary differential equation:
r wt (10.97)
+ im;P
Po (mr)2
e ; (10.98)
(10.99)
It is noted that the last result was derived in Example Problem 10.1, equation
(10.40), which expresses the free transverse vibration frequencies of a simple
beam. The result is again the Mathieu equation (10.67), but with differently
defined coefficients, or
o {10.100)
Thus, for a given set of system parameters (Cin; T3n) specified by equation (10.98),
the stability of Yn(t) and therefore the stability of the pipeline is determined
from the Haines-Strett plot of Figure 10.8.
(10.101)
F:rom Figure 10.8 it is observed that pipeline instability occurs near Cin = 0.25
and 1.0. With equations (10.101), the barge heave frequencies at which such
instabilities occur are deduced as
In this case, the coordinate pairs always líe below the dashed lines on the Haines-
Strett plot. Without damping, instability would occur near &n = 2.2, 4.0, 6.2,
272 CONTlNUOUS SYSTEMS
and so on. However, since natural damping always exists, this pipeline is stable
for w : : : Wn under parametric excitation.
Statistical Wave Excitation
Most of the background analysis needed to calculate the statistical dynamic
responses of beams to stationary, ergodic wave excitation has already been de-
veloped. Outlined now are the basic assumptions, methodology, and results
of this statistical analysis. In this section, the reader is encouraged to carry
through the details of the mathematical manipulations. Further expositions on
this topic are given by Clough and Penzien (1993) and Gould and Abu-Sitta
(1980).
The mathematical model chosen for analysis is the uniform Bernoulli-Euler
beam without longitudinalloading (P O) but with light, linear viscous damp-
ing. From equation (10.8) this model is
(10.104)
where the loading per unit length ij = q(x, t) is stationary, ergodic and Gaussian,
defined by the spectral density Sq(w).
Equation (10.104) is reduced toa familiar form by expanding both the dis-
placement solution v = v(x, t) and the loading ij = q(x, t) in terms of the
undamped normal modes Xn = Xn(x) and the generalized modal coordinates
Yn = Yn(t) and fin= fin(t). That is, let
00
00
The procedure now is analogous to that used previously to calculate beam re-
sponses with deterministic loading where equation (10.88) was derived starting
with equation (10.82) and the undamped beam model given by equation (10.69).
That ís, substitute equations (10.105) and (10.106) into equation (10.104); mul-
tiply each term by Xm Xm(x); from equation (10.73) let X~" = a~Xn; in-
tegrate each term of the resulting equation over the interval (0, f); interchange
the order of integration and summation under the assumption that the series
are uniformly convergent; and apply the orthogonality condition of (10.70) and
(10.71) to the result, which is valid for any of the six sets of beam support con-
ditions of (10.81). Thus all terms in the series vanish except for m n, which
leads to
(10.107)
BEAM RESPONSES 273
Define
(10.108)
With the last equation and the natural frequencies Wn given by equation (10.74),
equation (10.107) becomes
(10.109)
(10.110)
Here Hn(w) is the harmonic response function for the nth mode, derived from
equation (10.109) by letting
(10.111)
(10.112)
w2 w. ] -1
2 +2(n-J (10.113)
Wn Wn
The last ingredient needed in equation (10.110) is Sqn(w), the spectral den-
sity of the generalized force. To derive this term, first observe that the beam
loading ij(x, t) is related to the surface wave height r¡(t) by
where the transfer function G(w) is calculated for the chosen wave theory and
flow regime as discussed in Chapter 4. Rewrite equation (10.115) by replacing
ij = ij(x, t) with its expanded forro given by equation (10.106). Now multiply the
274 CONTINUOUSSYSTEMS
resulting equatíon by Xm = Xm(x), and integrate each term over the interval
(O, e). Applying the orthogonalíty conditions (10.70) and (10.71), the result is
1 re
lin(t) = Co Jo Xn q(x, t)dx (10.116)
1'
Co lo X~dx (10.117)
1
[!,' Xn IG(w)ldx]\(w) (10.119)
Now combine equations (10.114) and (10.119) with equation (10.110) to obtain
the time average of the response spectrum, or
(10.120)
where Co is given by equation (10.117) and modal damping has the form of
equation (10.108). The time average for the variance of this displacement is
given by
(10.121)
and the rms value of the displacement is the square root of a~ since v( x, t) was
assumed to have a zero mean. A space average response for this variance for x
in the interval (0, é) is defined as
DEPLOYMENT OF AN OTEC PIPELINE 275
(10.122)
The results of thls section are summarized. Begin by identifying the beam's
properties El, m, C, and its constraint condition of (10.81). Compute the natu-
ral frequencies and corresponding mode shapes for free vibration as outlined in
Section 10.1. For a selected wave theory and fiow regime, calculate G(w) as dis-
cussed in Chapter 4. Then select the wave height spectrum S,.,(w) and assume
the modal darnping factors Cn subject to the constraint of equation (8.88). Eval-
uate numerically the statistical responses given by equations (10.120)-(10.122).
Generally, the upper limit for n can be taken as 20, and the limits of integration
(0, oo) of equation (10.121) are (0.05, 1.5) radjsec for the commonly used wave
height spectrum.
M0 , End Mass
X
hiiiiiii/IIIJJIIJJIJ/111
Mathematical Model
The mathematical model of the OTEC pipeline attached to a barge is shown
in Figure 10.13. To the extent that classical beam theory is valid, the transverse
displacement v v(x, t) for the pipeline can be approximated by solutions of
the Bernoulli-Euler form given by equation (10.8). That is
2
4
El 8 v - - a (P-
8v) +m-
8 v
2
ij(x, t) (10.123)
8x 8x 8t
where damping is neglected. Both El, the bending stiffness, and m, the virtual
mass per unit length of the pipe including its contents, are constant. However,
P P(x, t), the longitudinal tension, and ij ij(x, t), the transverse wave
loading per unit length, depend on both x and time t. Motion is restricted to
the plane in which ij is a maximum, which is in the direction of the waves.
Consider the boundary conditions. The transverse motion at the top is
assumed to be that of the barge in sway, or
Kov(O, t)
(10.125)
ax
The transverse motion at the bottom is minimized by the addition of a ballast
mass Afo which is pinned at x = L. Thus
(10.126)
which express, respectively, the condition of zero end moment and the compat-
ibility of the transverse shear force at the pin with the motion of the ballast
mass.
At location x the average tension load on the cross section depends on two
dead weight terms and on the barge heave motion, H(t). That is,
where M 1g is the ballast weight in water, m1 g is the weight per unit length of
the pipe and its contents in water, and H(t) is the barge heave acceleration. lt is
asumed that H(t) is approx:imately the same at the top and bottom of the CWP
since the natural period of longitudinal pipe oscillations (T1 ) is significantly
smaller than the shortest excitation period (T2). For a steel pipe 1000 m long,
T1 0.3 s is smaller by about an order of magnitude than the shortest realistic
excitation period of T2 3 s.
Pipe Excitations by Barge and Waves
Assume that the motions of the barge from which the pipeline is deployed
are completely determined by the ocean waves. That is, the barge motion is not
affected by the pipe motion, which is a reasonable assumption since the barge
mass and the wave forces on it are both much larger than the pipe mass and
the pipe's wave forces. Choose the modified Pierson-Moskowitz spectral density
function for the sea surface elevation r¡( t), or
(10.128)
where A and B are constants and the wave frequency has the range w1 S:: w S::
w N. Partition the spectrum into N components for which the nth component
is Sr¡(wn). As discussed by Borgman (1969), the sea surface wave amplitude at
each central frequency Wn = nilw is given by
(10.129)
where (rp 1 , <jy 2 , ..• , <ÍYN) is a set of random phase variables distributed uniformly
over the interval (0, 2n). Approximate the empirical weighting function as
for O< Wn 5 Wb
for wb
Here Wb is related to the barge length B 0 by the empirical result given by Kim
et al. (1971): Wb cfBci/ 2 , where e is a constant. With H(t) obtained from
equation (10.130), the description of P(x, t) is complete.
The characterization of barge sway motion is somewhat more subtle than
that for heave motion. In the usual type of pipe-laying barges considered here,
second arder sway motions or slow drift oscillations domínate the first arder
sway motions. Thus it is appropriate to ignore the first arder sway motions, or
the motions having the same frequencies as the incident waves, with amplitudes
proportional to the first power of the wave amplitudes. In this analysis the
amplitudes of sway oscillations are assumed to be directly proportional to the
incident wave amplitudes. As is usual for slow drift motion, the associated
periods are based on the envelopes of these wave amplitude time histories. With
these assumptions, the sway motion of the barge sway is given by
N
S(t) 7J(wp) ·sin (e¡wpt + cPp) + L e27J(wn) ·sin (e¡wnt + <Pn) (10.132)
p+l
(10.133)
Here the coefficients eM) eD and C,1 represent inertia, drag, and added mass;
p is the mass density of water; D is the pipe diameter; and u is the horizontal
water particle velocity. Vortex shedding is neglected. By the superposition of
N simple, deepwater waves, ú from Table 3.1 with x O becomes
N
•
U = "L..,¡" Wn2 rJ ("'-'n ) · exp .knh(kz +d d) Sin
. (
Wnt + cPn
) (10.134)
n=l Slll n
DEPLOYMENT OF AN OTEC PIPELINE 279
where 41T 2 jTl = w;, A = ry(wn), cosh kn(z + d) ::::::: exp kn(z + d), d is the water
depth, gis the acceleration dueto gravity, and kn w;¡g.
The foregoing formulation of Morison's equation was derived from laboratory
tests of groups of circular cylinders in oscillatory flow. These tests showed that
whenever the ratio of cylinder diameter D to water wave length >. exceeded 0.2,
the Kuelegan-Carpenter number is small, and the drag-related viscous term
becomes unimportant. In such cases, Morison's equation (10.133) is no longer
correct since the flow field is modified by the presence of the cylinder. In such
cases, a diffraction analysis based on potential flow theory is used.
In the present case of the OTEC pipe, this ratio D j >. approaches 0.2 for
the shorter wavelengths in the wave excitation spectrum. Therefore, although
a diffraction analysis is not required, it is clear that the inertia term dominates
the excitation history. Separate calculations not included here indicate that the
drag-related term accounts for only about 10 percent of the maximum induced
loads. This observation was not particularly surprising since it is in agreement
with the results of other investigators, notably Hogben (1976) and Hogben and
Standing (1975). Therefore an attractive and obvious alternative, which is the
one employed here, is to neglect the drag term in equation (10.133) and at the
same time increase the remaining inertia term by 10 percent. The effect of this
approximation on pipe stability is negligible.
The excitations needed for solutions v(x, t) of equation (10.123) are defined
by equations (10.127)-(10.134). The solutions lead to the bending moment re-
sponses given by
M(x,t) (10.135)
From these moments, the critica! bending stresses are calculated using elemen-
tary theory.
For convenience, all 25 system parameters for this OTEC pipeline system
are summarized in Table 10.2, together with typícal numerical values used for
exploratory solutions. In a given ocean location, the values of d, g and p remain
essentially constant. The remaining 22 parameters fall in the following four
categories:
Numerical Results
Using the numerical values of the 25 system parameters listed in Table 10.2,
computer solutions were obtained to equation (10.123), subjected to the bound-
ary conditions of equations (10.124)-(10.126), and the local pipe tension given
by equation (10.127). A numerically stable implicit finite difference method
was employed, using the Crank-Nicolson approximation (Carnahan et al.,1964).
Fifty spatial steps in the interval O ::; x ::; L and 300 time steps were used in
solving the difference equations. The central difference forro was used except
for the shear-moment equation (10.126), which was expressed in tenns of for-
ward and backward differences. In all calculations, zero initial conditions were
assumed: the pipe was vertical and at rest at time t = O. The explicit difference
equations and a series of test runs performed to validate the computer program
for this problem were described in detail by Pandey (1980).
A total excitation time of 600 s was used for each of the 17 pipe lengths,
ranging from 61 to 1073 m. This simulation time was sufficient to encompass
DEPLOYMENT OF AN OTEC PIPELINE 281
143 heave cycles at the dominant heave frequency of WN 1.5 rad/s, and to
encompac;s about 3.6 dominant sway cycles of frequency e 1 wp 0.0372 radjs.
Since trial simulation times longer than 600 s produced no further increases in
peak responses, the 600 s simulation was considered in the present calculations
to give a steady state upper bound on the responses, where the deployment rate
was about 1.7 mjs.
Table 10.3 Peak Pipeline Responses During Deployment for a 600 s Sim-
ulation Time, for K 1.13 x 108 N·m/rad (Asterisks Represent Unbounded
Responses)
o~--~----~,--~~
(b) '' \
\
\
1
1
lOOr 1
1
1
~
/
/
/
/
/
2001- 1
1
1
~ M 0 = 227 OOOkg
\
\
3001-
''
''
''
' \
\
200 4001- \
1
1
1
1
/
/
244~--~--~--~--~--~ /
/
-4 -2 o 2 4 6 5001- /
/
Displacement v (x,(J, meters
/ "
549~~~~/~~--~----~
-100 -50 o 50
Displacement v (x,tc), meters
Figure 10.14 Effect of ballast on critica! pipeline displacements: (a) L =244m; (b)
L = 549 m.
Shown with asterisks in Table 10.3 are the pipeline lengths at which the
responses are unbounded. Note that the larger ballast mass (227000 kg instead
of 45 400 kg) increased the system stability for lengths of 183 m and 545 m, but
this additional ballast did not alleviate unbounded responses at lengths greater
than 610 m.
Shown in Figures 10.14a and 10.14b are typical stable pipe profiles corre-
sponding to the peak transverse displacement v(L, te) occurring during the 600
s simulations. Figure 10.14a shows that an increase of a factor of five in ballast
mass does little either to straighten the pipe to relieve its peak moment occur-
ring near x = 75 m, or to reduce its peak deflection at x = L. These two profiles
correspond to critical times te of 146 s and 264 s for the light and heavy ballast,
DEPLOYMENT OF AN OTEC PIPELINE 283
respectively. In Figure 10.14b, where L = 549 m, the profile for the lighter
ballast is not shown since those responses are unbounded, as shown in Table
10.3. The heavier ballast did stabilize the system, however, where te = 580 s. In
these studies, an unbounded response was defined whenever the ratio v(L, te)/ L
exceeded unity.
Shown in Table 10.4 are the effects of K, the rotational stiffness restraint at
the barge end, on the dynamic pipe response and stability. These results show
that, for a ballast mass of M o = 45 400 kg, an increase in K by a factor of 100
increased the range of pipe length for which the system remained dynamically
stable. For instance, the system was stable at L = 183 m for the higher rota-
tional restraint but not for the lower rotational restraint. Also, the higher K
reduced the peak deflection in the range up to L = 183 m, but beyond this
length the stable deflections often exceeded those corresponding to the lower K
valu e.
Table 10.4 Effects of Torsional Restraint on Peak Pipe Responses for a Ballast
of Mo = 45 000 kg and a Simulation Time of 600 s
in sorne cases by imposing a relatively fast deployment rate. However, these ex-
ploratory studies indicate that subsequent anchoring to the sea fioor of the end
ballast and of selected points along the pipeline using cable stays is needed
to maintain dynamic stability if a barge wave excitation, based on a Pierson-
Moskowitz wave spectrum, continues even for just ten minutes. However, if
the barge rotational restraint is chosen carefully and possibly changed during
deployment, stable pipelines can be achieved for longer times.
PROBLEMS
Figure 10.15 Three uniform, submerged structures: (a) a fixed-free pile; (b) a
fixed-hinged pile; (e) a pipeline with torsional restraint.
10.1 The submerged, uniform, cylindrical pile shown in Figure 10.15a has
full fixity at the sea fioor and is unrestrained at the top. Neglecting axial
loading due to self-weight, set up the determinant from which the undamped
bending frequencies can be calculated. From the transcendental equation de-
rived from this determinant, calculate the lowest three values of ant, where the
corresponding frequencies are given by
W n--
10.2 Calculate the lowest three undamped frequencies for a submerged pile
pinned toa deck structure, as shown in Figure 10.15b. Express the results in
the same formas in Problem 10.1.
10.3 The uniform, submerged pipeline shmvn in Figure 10.15c is hinged
to a barge and is restrained from rotating about that hinge by a linear spring
support of constant k. The restraining moment at x =O is
2\1(0, t) =k ov~; t)
The lower end is unrestrained. Account for self-weight by assuming a mean
value for the pipe tension, or P :::::; W /2 = constant where W is the pipe's
PROBLEMS 285
weight in water. Set up the determinant from which the transverse bending
frequencies can be calculated and from this derive the transcendental equation.
Discuss briefly how one could obtain a computer-aided solution for the con-
secutive roots of this equation. How are these roots related to the frequencies
Wn?
Weight, Weight,
l mg = P¿¡ mg=P¡¡
String: m, l
p= Po + P¡ cos (J)/
(a) (b)
Figure 10.16 Experimental setup: (a) Problem 10.4; (b) Problem 10.5.
k2
O:n '::::' 4' k= 1, 2, 3, ...
_ 2wn 2wn Wn
W = k =2wn,Wn'3'2''''
1 W1 (rrjé)JPo/iñ 1
1 2
1 3
1 4
2 2
2 3
2 4
W3 =
3 2
3 3
3 4
10.6 Based on the findings of Trogdon et al. (1976), designa series of water
tunnel experiments in which a submerged, tensioned cable undergoes combined
vortex and parametric excitatíon. Consider how the new hypothesized constant
k, related to the amplitude of the vortex-shedding force and the coupling effect
of longitudinal excitation, can be deduced from a set of experimental measure-
ments.
10.7 A stranded steel mooring cable has a density of0.2lb/in. 3 , a diameter
of 2.0 in., anda length of 500ft. If the cable tension has a mean value of 100,000
lb, what current velocities (perpendicular to the cable's longitudinal axis) will
give ríse to vortex shedding? The kinernatic viscosity and density of the seawater
at 45°F are 1.8 X w-- 5 ft 2 / sec and 0.0372 lb/in. 3 ' respectively. Take m, the
virtual rnass per unit length of cable, as the surn of its actual rnass per unit
length and the rnass of the seawater that it displaces per unit length.
10.8 A uniform cross beam of an offshore structure, modeled as the simple
beam of Figure 10.4b, is subjected to the harmonic wave loading q q0 sin wt.
(a) Calculate the steady state displacement v(x, t) based on Bernoulli-Euler
beam theory and the normal mode method. (b) Compute the series expression
for the dynamic bending moment at midspan and also the dynamic shear load
at the points of fixity. Neglect P, the longitudinalloading.
10.9 Solve Problem 10.8, but replace the simple end supports with clamped
ends as shown in Figure 10.4c. Neglect P.
REFERENCES 287
10.10 Consider a uniform cable under constant tension, fixed at each end,
and subjected to an arbitrary transverse load per unit length of q(x, t). Use
the normal mode method to predict the transverse cable displacement v(x, t),
assumed in the form of equation (10.82). Follow the same procedure used to
obtain the beam responses to the same loading, which leads to the results given
by equations (10.88) through (10.91). Show that Xn(x), n = 1, 2, ... for the
cable are orthogonal functíons.
10.11 Refer toa text such as Clough and Penzien (1993) or Timoshenko
and Goodier (1951) to derive the equation of longitudinal vibrations for a uní-
forro, elastic pipeline. Let u = u(x, t) be the longitudinal displacement of a
material point at position x on its longitudinal axis. For a pipeline without
end constraints, deduce that 8uj 8x = O on each end. Then, following the same
method used in Section 10.1 to derive the free transverse vibration frequencies
for a fixed end cable, to calculate the free longitudinal vibration frequencies for
the unconstrained pipeline. Check that your analytical result for w1 agrees with
equation (10.93).
10.12 Based on the equation of longitudinal vibrations for the pipeline
derived in Problem 10.11, calculate the free vibration frequencies for a uniform
pipeline clamped at each end. Compare your value of w1 to w.: of equation
(10.93), and discuss briefly the reasons for the difference in the two results.
10.13 The horizontal cross brace modeled in Figure 10.4c is located at z =
hin water of depth d. The beam is subjected to transverse wave excitation that
is modeled by linear wave theory. For S 11 (w), assume that the Pierson-Moskowitz
wave spectrum applies, where the significant wave height is 10m. Choose the
transfer function G(w) given by equation (4.30). See Example Problem 10.1 for
the explicit forms of frequencies and mode shapes. \Vrite a computer program
to calculate the midspan values of the displacement response spectrum, the
variance of v and the rms value of v. Specify carefully all input data. Test the
program by choosing a realistic numerical example. If S 11 (w) is Gaussian, how
are these results interpreted in terms of expected peak displacements?
10.14 For the horizontal beam of Figure 10.4b, derive an expression for
the spectral density of the bending moment in terms of the spectral density of
its displacement given by equation (10.120). Where along its length would the
peak rms value of the bending moment occur?
REFERENCES
Barr, R. A., and Johnson, V. E., Evaluation of Analytical and Experimental Methods
for Determining OTEC Plant Dynamics and CWP Loads, Proceedings of the
Offshore Technology Conference, 1979.
Borgman, L. E., Ocean Wave Simulation for Engineering Design, Journal of the
Waterways and Harbors Division, ASCE 95 (4), November 1969.
Carnahan, B., Luther, H. A., and \Vilkes, J. O., Applied Numerícal Methods, Vol. 2,
Wiley, New York, 1964.
288 CONTI~VOUSSYSTE~1S
Clough, R. W., and Penzien, J., Dynamícs of Structures, second ed., McGraw-Hill,
New York, 1993.
Gould, L. P., and Abu-Sitta, S. H., Dynamic Response of Structures to Wínd and
Earthquake Loadíng, Wiley, New York, 1980.
Green, W. L., Gray, D. W., Landers, E. A., and Calkins, D. E., Hybrid OTEC Plants-
An Assessment of Feasibílity, Proceedíngs of the Offshore Technology Conference,
1980.
Griffin, A. and Mortaloni, L.R., Baseline Designs for Three OTEC CW Pipes, Pro-
ceedíngs of the Offshore Technology Conference, 1980.
Hogben, N., Wave Loading on Structures, Proceedings of the Conference on Behavior
of Offshor·e Structures, BOSS, Trodheim, 1976.
Hogben, N., and Standing, R. G., Experience in Computing Wave Loads on Large
Bodies, OTC 2819, Proceedíngs of the Offshore Technology Conference, 1975.
Hove, D., and Grote, P. B., OTEC CWP Baseline Designs, Proceedíngs ofthe Offshore
Technology Conference, 1980.
Lubkin, 1., and Stokes, J. J., Stability of Columns and Strings under Periodic Forces,
Quarterly of Applíed Mathematic.~, 1(1), 1943.
McGuiness, T., Griffin, A., and Hove, D., Preliminary Design of OTEC CW Pipes,
SB-1, Proce.edíngs of the Offshore Technology Conference, 1979.
O'Brien, J. T., and Muga, B. J., Sea Tests on a Spread-Moored Landing Craft,
Proceedíngs of the 8th Conference on Coastal Engineeríng, Lisbon, Portugal,
1964.
Pandey, P., Dynamic Analysis of Pipelines for Ocean Thermal Energy Conversion,
Master's Thesis, Department of Civil Engineering, Duke University, Durham,
NC, 1980.
Scotti, R., and Galef, A., An Assessment ofthe Reliabilíty ofthe OTEC CWP Design
Analyses, ID-1, Proceedings of the Offshore Technology Conference, 1980.
Timoshenko, S., and Goodier, J. N., Theory of Elasticíty, McGraw-Hill, New York,
1951.
Trogdon, S. A., Wilson, J. and Munson, B. R., Dynamics of Flexible Cables un-
der Combined Vortex and Parametric Excitation, .Journal of Dynamic Systems,
Measurement and Control, 1976.
Whitney, A. K., and Chung, J. S., Vibrations of Long Marine Pipes dueto Vortex
Shedding, Journal of Energy Resources Technology, 1981.
vVilson, J. Muga, B. J., and Pandey, P., Dynamics during Deployment of Pipes for
Ocean Thermal Energy Conversion, Proceedings of the 1st Offshore Mechanics,
Arctíc Engineeríng, Deepsea Systems Symposium, New Orleans, 1982.
11
Behavior of Piles Supporting
Offshore Structures
Lymon C. Reese
290 BEHAVIOR OF PILES SUPPORTING OFFSHORE STRUCTURES
estimation of the relative density and friction angle is usually made from the
blow counts found in the field.
Single piles are found in onshore structures and occasionally in an offshore struc-
ture. But most offshore structures are founded on groups of piles. If the piles are
four or five diameters apart, pile-soil-piles interaction is negligible and the piles
can be designed as if they were isolated. If the spacing is closer, an efficiency
factor is employed to reduce the capacity of piles in the group. The mechanics
of the transfer of load from the pile to the soil for the single pile can be modified
to include a reduction coefficient.
A rational model for the pile-soil system is shown in Figure 11.1 (Reese and
Van lmpe, 2001). The pile is shown as a spring to reflect that all piles, regardless
of material, will deform under load. This spring has a constant stiffness with
depth but its stiffness can vary with length without causing analytical difficulty.
1/'A\V 17 11\\'(/
fzb-
Wz
z ttc= Wz
tC Wz
tC Wz
tC
(a) (b)
(e) (d)
T
Z¡
l
Pile Movement Wz
For a particular applied load, values of settlement Wb at the base of the pile
can be found by procedures given above. The corresponding values of unit end
bearing load q can be found by dividing the load at the base of the pile by the
area of the base. The full qfwb curve can be obtained by analyzing all of the
loadings, provided the pile settles sufficiently. If a number of such experiments
in a variety of soils for a variety of piles are performed, correlations can be
developed between soil properties and the load-transfer curves.
Referring to Figure 11.2a, the load test should be performed to cause plung-
ing, where additional deflection results in no increase in load. The result will be
the development of the load-transfer curves to ultimate values, as shown by the
curves for the depths z 1 , z2, and z3. The ultimate values of fuu can be correlated
with the shear strength of the soil in arder to develop equations for computing
the ultimate load in skin friction Q8 • If the end bearing has reached an ultimate
value, the ultimate value of q can be correlated with the shear strength at the
base of the pile and yíelds the ultimate load in end bearing Qb. The capacity
of the píle under axial load Q is Q 8 + Qb.
The procedure for computing the capacity and settlement of a pile in clay,
found offshore at many locations, is illustrated in the paragraphs below. Let
(11.1)
and
(11.2)
for the three strata, starting with the top: 0.83, 0.61, and 0.48. With these
numerical values, equation (11.1) becomes
The reasonable assumption is made that the pile plugged at sorne point during
driving and thus the end bearing can be computed as if the pile were solid. The
following result is found by substituting values into equation (11.2).
Qb = (2.5/2)(9.0)11"(1.5) 2 = 80 tons
Thus, the total load the pile can sustain was computed to be 760 tons or 1520
kips. The safe load would be found by using an appropriate factor of safety.
The load-settlement curve for the pile can be computed by implementing
load-transfer curves. With regard to the load transfer in skin friction for piles
in clay, Coyle and Reese (1966) examined experimental data and proposed the
results shown in Table 11.1.
With regard to the load transfer in end bearing for piles in clay, the work
of Skempton (1951) is used, where the end bearing of a plate loaded in clay is
shown to correlate with the laboratory stress-strain curve. He noted that the
settlement wb at one-half the ultimate unit end bearing of the base is equal to
Wult
(1L3)
2
In the absence of a laboratory stress-strain curve for the soil, the following
values of E5o can be taken as a function of the unconfined compressive strength
(consistency) ofthe clay: soft ( <0.5 tonsjft 2 ) 0.0.2; medium (0.5 toLO tons/ft 2 )
0.01; and stiff (>LO tons/ft 2 ) 0.005. For the example problem, the value of Qu
at the base of the pile was 2.5 tsf, so the value of c5 o was selected as 0.005, with
Wult/2 equal to 0.36 in. Numerous stress-strain curves for soíl have been plotted
on log-log paper and found to be a straight line, many with a slope of about
SINGLE PILES UNDER AXIAL LOADING 295
0.5 up to the failure stress. Therefore, the other points on the load settlement
curve for the base can be computed by the following equation:
(11.4)
Based on Qb = 40 tons for Wb = 0.36 in., the coefficient <; was computed as 66.7
tons/ .Jill.. With this coefficient, sets (Qb tons , Wb in.) can be computed from
this equation.
The data are at hand for computing the axial load-settlement curve for the
pile. An examination of the values of load-transfer in side resistance shows that
the curves indicate movement-softening. Depending on the characteristics of
the soil at the site, the engineer might assume that movement-softening does
not occur and modify the curves somewhat. Comparison of the curves for side
resistance end bearing reveals that the side resistance is generated with much
smaller pile movements than the end bearing. Therefore, to illustrate the com-
putation procedure, a movement of the base of the pile is selected as 0.05 in.,
giving a load in end bearing of 15 tons or 30 kips using equation (11.4).
With regard to the computed load transfer in side resistance, the accurate
approach is to select small increments of length, perhaps as small as one or two
feet. However, the procedure can be shown by taking increments along the pile
equal to the three strata of soil; therefore, the fust increment is from the tip of
the pile at 1680 in. to the top of the lower stratum at 1200 in.
For the first trial, the assumption is made that the movement at the mid-
height of the stratum is the same as at the base: 0.05 in. Using the data on
load transfer in side resistance and in movement to achieve the value of relative
load transfer (ignoring movement softening), the value of unit load transfer was
7.33 lb/in. 2 . The load transfer in side resistance between 1680 in. and 1200
in. was computed to be 398 kips. Thus, the load at the bottom of the 480 in.
section was 30 kips and the top was 428 kips. The elastic shortening from the
tip to the rnidheight of the section, at 1440 in., using elementary mechanics was
computed to be 0.0091 in., yielding a midheight movement of the lower section
of 0.0591 in. rather than 0.05 in. Employing the new midheight movement, the
load transfer in the lower section was computed to be 435 kips, compared to the
398 kips for the first trial. Another iteration was done and the midheight move-
ment of the lower section was found to be 0.0614 in., which yielded a revised
load transfer for the lower section of 440 kips, compared to the 435 kips for the
previous iteration. Convergence was assumed, which gave a load at 1200 in. of
461 kips and a computed movement of 0.0863 in.
Employing the above procedure, computations were done for the top two
strata of soil, and loads and deformations were accumulated, with the com-
puted top load of 1369 kips and a top movement of 0.4423 in. The procedure
could be continued for other values of tip movement in order to obtain the
full curve for load versus settlement for the top of the pile. Shown in Figure
11.3 is a computer-generated top load versus settlement curve. As can be seen,
the ultimate load is 1520 kips. This plot is very instructive. Assuming that
the load-transfer curves are acceptable, the engineer can see readily the settle-
296 BEHAVIOR OF PILES SUPPORTING OFFSHORE STRUCTURES
ment necessary to develop any portian of the load transfer, which is valuable
information in selecting the safe load on the pile.
6~----------------------------~
Figure 11.3 Computed curve of axial load versus settlement for a pile with a length
of 140ft dríven into clay soils.
(a) (b)
T ==:::L::=:,
~y
z
===i2=y
~
X
Figure 11.4 Model for pile under lateral loading showing typical p-y curves.
d4y á2y
El dx4 + Px dx2 - P =O (11.5)
where EI is the flexural rigidity of the pile, y is the pile deflection at position
x along its length, Px is the axial load, and p is the soil reaction force per unit
length.
298 BEHAVIOR OF FILES SUPPORTING OFFSHORE STRUCTURES
Equation {11.5) can be solved using a digital computer (Reese and Van Impe,
2001, p. 29); however, nondimensional methods can sometimes be employed to
yield an acceptable solution for cases where El is constant and there is no
axial load. Both methods of solution give all the necessary design information
including the moment, deflection, and shear at desired lengths along the pile.
The methods described herein have received wide acceptance and are used in
many design offi.ces around the world.
Response of Soil
p
(11.6)
y
The value of p from the last equation can be substituted into equation (11.5)
and a solution obtained for the values of y with respect to points along the pile.
Because Epy is a nonlinear function, equation (11.5) can be solved by iteration
using procedures developed for piles in a variety of soils and rack (Reese and
Van Impe, 2001, p. 49). The recommendations are based principally on the
results of full-scale experiments which are augmented with theory to the extent
possible. Reese and Van Impe (2001, p. 259) showed the comparison of results
from experiments with results from analysis for a sizable number of cases. The
v<tlidity of the analytical method has been well established within a reasonable
degree of accuracy.
Consider solutions for lateral loading for pile, embedded in a soft clay be-
low the water surface, a condition encountered frequently at offshore locations.
Matlock (1970) presented procedures for developing p-y curves for soft clays be-
low the water surface for two loading conditions: short-term static and cyclic.
Those procedures, somewhat simplified, are now summarized.
1.00
o I.....L...-----l..----
0 1 8 yly50 3
Figure 11.5 Characteristic shapes of the p-y curves for soft clay below water
surface: (a) static loading; (b) cyclic loading (Matlock, 1970).
SINGLE PILES UNDER LATERAL LOADING 299
2. Compute the ultimate soil resistance per unit length of shaft, Pu: using
the smaller of the values given by the following two equations:
1
Pu = ( 3 + ~Z
)'
+ b0.5 Z ) cb (11. 7)
Pu 9cb (11.8)
where )'1 is the average effective (submerged) unit weight from ground surface
to the p-y curve, x is the depth from ground surface to the p-y curve, e is the
undrained shear strength of the undisturbed clay soil at depth x, and bis the
pile width. Compute at each depth where a p-y curve is desired, based on the
shear strength at that depth.
3. Compute the deflection, Yso, at one-half the ultimate soil resistance. Use
the following equation:
(11.9)
l. Construct the p-y curve in the same manner as for short-term static
loading, for values of p less than O. 72p .
2. Equate equations (11. 7) and (11.8) and compute the depth z Zr where
the transition occurs. If the unit weight and shear strength are constant in the
upper zone, then
6cb
(11.11)
+0.5c
3. If the depth to the p-y curves is greater than or equal to Zr, then the
value of p is equal to O. 72p for all values of y greater than 3Y5ü·
4. lf the depth to the p-y curve is less than or equal to Zr, then the value of p
decreases from O. 72pu at y 3y5o to the value given by the following expression
at y 15Y50·
(11.12)
---
~+M
(a) +V
- +p
~+S
--
+x '-"' --
(b)
(11.13)
302 BEHAVIOR OF PILES SUPPORTING OFFSHORE STRUCTURES
where El is the flexura! rígidity of the pile and kpy is a constant relating the
secant modulus of soil reaction to depth (Epy kpyx).
3. Compute the depth coefficient Zmax as follows:
Z _ Xmax
max- T (11.14)
4. Compute the deflection y at each depth along the pile where a p-y curve
is available. Use the following equation:
PtT 3 MtT 2
y =Ay El + ByEI (11.15)
(a) 5 (b)
~_p.
~;.. 4 X 4
E;
~
....:-
¡;;:
~·
<U 3 A 8T3 ....:- 3.
.(3 Y,¡ ¡;;: 1 X
!.;::::
YEJ <1i
....... 2 •<:> M.T2
<U
o !E ~ B/f!t
u (.)
o
¡;;:
o u
·.::: ¡;;:
(.)
<U
!:;:;
o o
·.::: o
(.)
<U <U
o -1 !:;:;
<1i -1
Zmax= 2 o Zmax=
-2 -2
o 2 3 4 5 o 2 3 4 5
Depth Coefficient, Z xff Depth Coefficient, Z = x/T
Figure 11.7 Case 1: Pile deflection produced by: (a) lateral load at the mudline; (b)
moment at the mudline.
5. Select from a p-y curve the value of soil resistance p that corresponds to
the pile deflection value y at the depth of the p-y curves. Repeat this procedure
for every p-y curve that is available.
6. Compute a secant modulus of soil reaction using equation (11.6).
Plot the Epy values versus depth.
7. From the Epy versus depth plot of Step 6, compute the constant, kpy,
which relates Epy to depth (kpy = EpyX), giving more weight to the Epy values
near the ground surface.
SINGLE PILES UNDER LATERAL LOADING 303
8. Compute a value of the relative stiffness factor T from the value of kpy
found in Step 7. Compare this value of T to the values of T assumed in Step 2.
Repeat Steps 2 through 8 using the new valu e of T each time until the assumed
value of T equals the calculated value of T.
9. When the iterative procedure has been completed, the values of deflection
along the pile are known from Step 4 of the final iteration. Values of soil reaction
are computed from the basic expression: p = EpyY· Values of slope, moment,
and shear along the pile can be found by using the following equations:
PtT 2
S As El + (11.16)
(11.17)
(11.18)
(a) (b)
o o
-0.5 -0.5
-...::"'
-1.0 ~ M¡f-rr--
...;:- ...;:- -1.0
~v sor
X
e
'o
~ 5 HM,T
-1.5 E: ·u -1.5 El
E ~
lS S
~
o SA=A.sP,T
u -2.0 El 8 -2.0
~ !J)
Q..
o -2.5 §- -2.5
ü3 zlllllX 2 ü3 t\
-3.0 -3.0 i Zmax = 2
-3.5 ~.....___.__......._____.._........___¡ -3.5 '--~-~--'--~--'
o 1 2 3 4 5 o 2 3 4 5
Depth Coefficient, Z x!T Depth Coefficient, Z xiT
Figure 11.8 Case I: Slope of pile caused by: (a) lateral load at mudline; (b) moment
at mudline.
(a) (b)
0.8 l. O
~
"~:: 0.6
M.h X
¡¡
CCl 0.8
~B~-
t X
lpt ....:-
É<!) M_.=AmftT S:: MB=B,M,
'<:) 0.4 <!)
'<:) 0.6
lE lE
g 0.2 o 0.4
<!)
-o
u
~
E
u
e:
<!)
E
o
0.2
o 5
~ -0.2 ::;E JO
-04
.o -0.20
2 3 4 5 2 3 4 5
Depth Coefficient, Z = x!T Depth Coefficient, Z = x!T
Figure 11.9 Case I: Pile bending moment produced by: (a) lateral load at the
mudline; (b) moment applied at the mudline.
(a) (b)
0.2
1.2
,. o
"~::
....r 0.8 ~~ t::tS
....:-
S::
<!) 1¡~ ;::
<!) -0.2
'!3 0.4 'G
E J;í = AvFJ !,;::::
M;trv
<¡...<
<!) <!) -0.4
o
u o '\ '\ u
o
!;¡ \ \\ . !;¡ -0.6 V¡¡ M
-0.4
Vn= Bvy
~ ;
<!) <!)
.S:: \ \ ""'·, .S::
00 Zmax=2.,/ ,_, 00 -0.8
-0.8
o 1 2 3 4 5 o 2 3 4 5
Depth Coefficient, Z = x!F Depth Coefficient, Z x/T
l. Perform Steps 1- 3 of the solution procedure for free head piles, Case l.
2. Compute the deflection y at each depth along the pile where a p-y curve
is available by using the following equation:
F PtT 3 (11.19)
YF Y El
The deflection coefficients Fy can be found by entering Figure 11.11 with values
of Z, using the appropriate curve according to the value Zmax·
3. Proceed in steps similar to those for the free head case.
SINGLE FILES UNDER LATERAL LOADING 305
t:..:..." 0.8
¡:f
·M 0.6
e IU
o
u 0.4
t::
o
·.;::
(..)
IU 0.2
ti:
IU
Ci
o
-0.2
o
Depth Coefficient, Z = x/T
Figure 11.11 Case II: Deflection of pile fixed against rotation at the mudline.
4. Compute Mt, the moment at the top of the pile, from the following
equation:
(11.20)
The value of FMt can be found by entering Table 11.3 with the appropriate
value of Zmax·
Table 11.3 Moment Coefficients at Top of Pile for the Fixed Head Case
3 -0.97
4 -0.93
5 and above -0.93
5. Compute values of slope, moment, shear, and soil reaction along the pile
by following the procedure for the free head pile.
Case III: Pile Head Restrained Rotation
This case can be used to obtain a solution when the rotational restraint is
known for the pile on entering the superstructure. Such a case is illustrated by
the example that follows.
l. Perform Steps 1- 3 of the solution procedure for free head piles, Case I.
306 BEHAVIOR OF PILES SUPPORTING OFFSHORE STRUCTURES
ko = Mt (11.21)
where Jlvft is the moment at the top of the pile and St denotes the slope there.
(Note that St sometimes denotes soil sensitivity, but in context, there should
be no confusion of symbols.)
3. Compute St, the slope at the top of pile, as follows:
Here Ast is the slope coefficient As found in Figure 11.8a; and Bst is the slope
coefficient Bs found in Figure 11.8b.
4. Solve equations (11.21) and (11.22) for Mt, the moment at the top of the
pile.
5. Perform Steps 4-9 of the solution procedure for free head piles, Case I.
This completes the solution of the laterally loaded pile problem for three sets
of boundary conditions. The solution gives values of deflection, slope, moment,
shear, and soil reaction as a function of depth. The example problem in the
next section illustrates the use of this method.
The nondimensional method as presented above has the following limita-
tions: the influence of an axial load was not considered; the pile must have a
constant value of El; and perhaps of most importance, the soil must be of one
type and preferably have a shear strength that increases linearly with depth
from zero at the mudline. In spite of these lirnitations, the nondimensional
method can give good answers to a considerable share of cases of lateralloading
of piles. Furthermore, this hand solution can reveal explicitly the influence of
·various parameters.
Computer-Aided Solutions
Computer codes have been written to elirninate the limitations in the nondi-
mensional method by solving equation (11.5) using finite-difference techniques.
If a pile is divided into n increments of length h, then n + 1 equations of the
following form can be written:
+ Px(
h2 Ym-l o (11.23)
Two imaginary points must be introduced at the top of the pile and two at the
bottom, leading to n + 5 unknown deflections. Two boundary equations at the
AN EXAMPLE: LATERAL PILE LOADING 307
bottom of the pile and two at the top can be added to the n + 1 standard
equations. Solutions of these n + 5 simultaneous equations yield the deflections
along the length of the pile, from which slope, bending moment, shear, and soil
reaction may be readily computed. Equation (11.23) is written for a constant
El; however, the equation can be rewritten to account for a change in pile
stiffness. This expanded equation presents little additional difficulty in writing
the computer program.
A number of methods can be used to salve the simultaneous, finite difference
equations. A method of Gaussian elimination discussed by Reese and Van Impe
(2001) was found to be efficient. The mesh size h and the number of significant
figures in the computation procedure must be controlled in arder to achieve
acceptable accuracy. Reese and Van lmpe (2001) present severa! case studies.
The definite advantages of a computer-aided solution are summarized:
l. Changes in flexura! stiffness of the pile can be introduced at any depth.
2. The pile length can be changed as desired.
3. The p-y data can be introduced in several ways.
4. The Epy value can be changed from point to point as dictated by the soil
response.
5. An axial load can be specified and accounted for in obtaining the shape
of the deflected pile.
t
Figure 11.12 Aplane truss or bent from an offshore platform.
Problem Description
The problem is illustrated in Figure 11.12, which shows aplane truss or bent
from an offshore platform. A template or jacket, which consists of a welded pipe
framework, is first set on the ocean fioor. In this particular case the legs of the
jacket are assumed to penetrate into the soft surface soil. The penetration is 140
in.; mud sills are set below the bottom bracíng to keep the jacket from sinking
further.
After the jacket is set, the piles are stabbed into the jacket legs and driven
to a predetermined penetration. The annular space between the pile and the
jacket is then grouted. (Sorne designers prefer not to use a jacket-leg extension
and not to grout the annular space in order to reduce the bending moment at
the bottom panel points. In this case, the wall thickness of the pile must be
greater than in the case being considered.)
(a) (b)
h= 240 in.
Figure 11.13 (a) Section of jacket showing pile and superstructure at mudline; (b)
interaction between pile and jacket leg with assumed equation for rotational restraint.
Shown in Figure 11.13 are apile anda portion of the superstructure. Two
boundary conditions are indicated: the lateral load Pe and the rotational re-
straint ke. As indicated in the figure, the restraint against rotation at the pile
head is provided by the superstructure and can be computed using equation
(11.21), or
6
k
8
= Mt "'"' 3.5Efc = (3.5)(30 X 10 )(14,117) = 6 18 109 . -lb/ ad
St - h 240 · x m. r
(Note that hin the above computations is the panellength and not the increment
length as defined earlier.) A constant pile stiffness El must be employed in the
AN EXAMPLE: LATERAL PILE LOADING 309
+
'r-l --- --+ \
\
t\¡~ ¡ --,~
'
1
-J l)
flocculated below
85'
l.
11
--- -+
1'
,1
d
,t
\ •~
.,
..0 with gas,. 95 '-107'
17 1
with shell frag-
:::::: 160
o ments at 95'
stiff below 95'
11
••
+-; --+
1~ f\ .,.,\
wi th sand lsyers ,
127' to 180'
..
19
f¡f-· 1
' ',
~ 200 1
silty clay below
1so'• -207' ,.
\ 1\1.......
-
Gray fine sand
silty at 216' & ""'" *Number ofblows of a 175 lb weight
,..,..
!0/12"
d.ropped approximately 5 ft , required to
-
246'
240 produce (except as noted) a 24 in.
with clay pockets penetration of a 2-1/8 in. ID sampler.
& shell fragmenta 30/6"
at 265' ,.,.... (Sampler OD = 2.5 in. to 23 ft penetration
fine to medium sand,
lDil'
and 2.25 in below 23 ft penetration.)
slightly silty
280 below 275' :Jl/3"
30/6"
Completed at 305.5 ft penetration l
L..
r305.5' (EL at 396.5 ft)
'
Legend for Shear Strength Plot
• unconfined compression • unconsoliated-undrained triaxial
+ miniature vane • consolidated-undrained triaxial
Figure 11.14 Soil profi.le with properties for the example problem.
310 BEHAVIOR OF PILES SUPPORTING OFFSHORE STRUCTURES
nondimensional solutions, and the stiffness of the upper portian of the pile is
selected. Experience has shown that the behavior of the upper portian of a pile
under lateral loading has a significant effect on the result of a solution.
Solutions Using Nondimensional Parameters
The p-y curves are nonlinear, which leads to a nonlinear solution with respect
to the lateral load Pt; therefore, a series of solutions should be generated for
a range of values of Pt. An approach that is useful is to increase the Pt by
multiplying the service load by the factor of safety and to check the behavior
of the pile under the increased load. However, it is wise to employ loads below
and above the design load because nature of the p-y curves is such that in sorne
cases a slight increase in lateral load can cause a large increase in defl.ection and
bending moment.
The soil conditions employed in this analysis are shown in Figure 11.14.
The soil conditions are representative of those in the eastern Gulf of Mexico
and offshore from the Louisiana coast. The water depth is 91 ft. The undrained
shear strength and the water content vary with depth in a manner like that of
a normally consolidated day. The soil profile was simplified somewhat, and the
following values of undrained shear strength and submerged unit weight were
selected for the analysis:
no effect on its behavior. For the following solutions, a pile length of 1500 in.
was selected.
12
-::0
~8 700 in.
o
o........
(!)
(.) 6 500 in.
ª......
......
V'.J
V'.J
(!)
4
400 in.
-
~
......
o
r/1
2
300 in.
200 in.
100 in.
50 in.
o
o 2 4 6 8 10 12 14
Deflection, y in.
COMPUTATION SHEET
-(1.75)(200)Mt/(4.23 x 10ll)
y =1,135Ay - 0.880By
The Computation Sheet and Table 11.5 show the first trial for a lateral load
Pt = 60,000 lb. A relative stiffness factor T of 200 in. was selected. A value
of the pile-head moment Mt was computed and the nondimensional expressions
were used to compute a lateral defl.ection y at the depth of each of the p-y curves.
The p-y curves of Figure 11.15 were entered with the computed defl.ection to
obtain a value of p, with which the values of soil modulus Epy were computed.
The values of soil modulus are plotted in Figure 11.16, and the best straight
line passing through zero is fitted through the plotted points. Somewhat more
weight is given to the points near the mudline in fitting the line. The slope of
the line is the value of kpy, the variation of the soil modulus with depth. The
computed values of kpy and Tare shown at the bottom of Table 11.5. Note that
the first trial value of T = 200 in. yielded a computed value of T = 221 in.
AN EXAMPLE: LATERAL PILE LOADING 313
Second and third trials were made follmving this same computation proce-
dure, and the resulting values of Epy are plotted in Figure 11.16. A second
trial value of T of 250 in. yielded a computed -v'alue of 244 in. A plot of the
tried values of T against the computed values that were obtained is shown in
Figure 11.17. As shown in this figure, T will converge toa value of about 239 in.
When the third trial was made with a tried value of T of 239 in., convergence
was indeed found to be this value.
Epy, lb/in~
0 or---~loo~--~2~o~o___3~0~0--~4Too~--5~oo
~
\'
50
'\~ + First tria!
• Second tria!
\. o Third tria!
100
·\\\~F;nt truoJtria!
.s \1 Second
1
....~ \x-Third tria!
.~1 \ '
i3
Q.
<U
o 200
\\
250
\1
\\
\
\\
\ ••
\\
o \ +
350
\1
\ 1
1
\ \
Figure 11.16 Trial plots of Epy values.
below a depth of 140 in., and the p-y curves were computed from the soil prop-
erties that existed at each depth. Computations were made for values of Pt in
increments of 20 kips up to a value of 160 kips.
260.-----.-----r-----.-----r----.
T(tried)
106M, in.-Ib
-16 -12 -8 -4 o 4
J
Bottom ofjacket-leg
extension
7
600 1
.5 1
>< 1
.;¡ 800
0.. 1
(!)
o 1
1
1000 ,'
1
1200
Figure 11.18 Bending moment versus depth for two methods of solution.
AN EXAMPLE: LATERAL PILE LOADING 315
The bending moment along the length of a pile is of major concern for
an offshore structure and pile deflection in most cases is less important. Figure
11.18 shows these computed values of bending moments with the two procedures
for a lateral load of 60 kips. There is reasonable agreement in the results of
the two methods, with the computer results showing somewhat more bending
moment, perhaps because the use of the axial load of 500 kips. The position of
the jacket leg extension is indicated, which allows the moment at the top of the
30 in. section to be found. The maximum values of combined stress in the 30
in. section and the 33 in. section were computed, assuming that the jacket leg
extension was fully grouted so that the axial load at the mudline was taken by
both the 33 in. and the 30 in. sections; and it was found that the most critical
stress occurred at the top to the pile or where the pile joined the jacket.
Y¡-==-.,.... '"llv
120 , ,.
V> ,, "
~
0..
,"
/
¡:¡_,- 80 J'
J'
1
__.._ Computer
- -X-- Nondimensional
o
o 4 8 12 16 20
Deflectíon, .Y¡ in.
o 2 3 4 5
Moment, 107 M1 ín.-lb
Figure 11.19 Deflection and bending moment at the pile head based on two
computation methods.
The computed values of pile deflection and maximum bending moment for
the top of the pile for the two methods of computation are shown in Figure
11.19. The computed values agree reasonably well, although the results from
the computer are somewhat greater. If mild steel with a yield strength of 36 ksi
is used in the design, the computed lateral load to cause a plastic hinge at the
top of the pile can be found. The axial stress fa can be computed as follows for
the areas of 51.05 in. 2 and 69.92 in. 2 of the respective sections of 33 x 1/2 in.
and the 30 x 3/4 in. Thus
The allowable bending stress would then be 31.83 ksi. The section modulus of
the section (14,117/16.5) yields a bending moment at failure of 27,233 in.-kip.
By entering Figure 11.19 with the bending moment, a lateral load of about 105
kips is found. The computed pile-head defl.ection at that load was about 6. 7
in. A portion of the lateral load could be taken by dynamic loading and sorne
load factor, according to the design criteria, to find the loading under service
conditions.
Conclusions
In addition to the comments made in the presentation about the importance
of obtaining the best possible prediction of p-y curves, the following points can
be made:
l. The nondimensional solution can yield solutions that agree fairly well
with the computer solution if the soil profile shows favorable ;ralues of shear
strength with depth (sands and normally consolidated clays).
2. The nondimensional solution can (a) serve as a check to computer so-
lutions and (b) reveal clearly the nature of the lateral-load problem and the
importance of various parameters.
3. The plot of bending moment with respect to depth shown in Figure 11.18
clearly demonstrates the importance of analyzing the piles and superstructure
as a unit to ensure the best estímate of rotational restraint at the top of the
pile.
4. The engineer can achieve the most efficient design of the piles by taking
into account the length of the extension of the jacket, the details of grouting
between pile and jacket, the wall thickness of the pile with respect to depth,
and careful estimation of loadings on a platform as a function of time.
Earthquakes
Several steps are taken in order to design the foundation for a pile-supported
structure in a seismic region. The location of the fault with respect to the
structure must be found and the expected magnitude of the event must be
selected. Then, the characteristics of the soils and rocks at the site must be
considered in order to perform microzonation. With such data at hand, the time-
dependent, free-field motion of the supporting soils and rocks at the building site
can be computed or estimated. The engineer can then decide if there is a chance
of liquefaction of any loose granular soil below the water table. If liquefaction
appears to be likely for the selected earthquake, steps must be taken to improve
the supporting soil or to design a structure that remains stable, even though
sorne of the supporting soil liquefies.
If the free-field motion of the soils at the site are known as a function of
depth, a fully rational solution can be undertaken which will require extensive
and complicated computations. The validity of the results of sueh computations
RESPONSE OF FILES TO DYNAMIC LOADING 317
PROBLEMS
11.1 Compute a point on the curve of axial load versus settlement for the
example in the text using a base settlement of 0.02 in. (instead of 0.05 in.).
Compare the result with the computer solution in Figure 11.3.
11.2 Refer to the American Petroleum Institute's (1993) Recommended
Practice and compute the allowable load that can be sustained by the axially
loaded pile in the example.
11.3 Given a normally consolidated clay with an undrained shear strength
that increased from zero at the mudline to a value of 1250 psf at a depth of
100 ft, an average submerged unit weight of 45 pcf, and an c 5 o of 0.02, make
necessary computations and plot the p-y curves for both statíc and cyclic loading
for depths of O ft, 6 ft, and 12 ft for a pile with a diameter of 24 in. and a wall
thickness of 1.0 in. The pile is assumed to be at an offshore location; select an
appropriate value of the submerged unit weight in your computations.
11.4 Assume a fixed-head, open, steel-pipe pile with a diameter of 30 in.
and a wall thickness of O. 75 in. and assume further that the pile behaves as a
long pile. Compute the magnitude of the lateral load at the mudline to yield
a maximum bending moment of 20 ksi if Epy = kpyX , where kpy = 12 lb/in. 3 .
Assume no axial load anda constant El with depth. How long should the pile
be? Use the nondimensional method.
11.5 Repeat Problem 11.4 with the assumption that the head of the pile
is free to rotate.
11.6 For Problems 11.4 and 11.5, find the rotational restraint such that
the maximum negative moment at the pile head and the maximum positive
moment are equal. Use the lateral load found in Problem 11.5.
11.7 Use data from Problem 11.3 and assume a lateral load is applied at
10 ft above the mudline. Salve for the energy from a docking boat that can be
sustained by the pile if the maximum allowable pile bending stress is 30 ksi. Use
the nondimensional method of analysis and assume the number of repetitions
is small so that the p-y curves for static loading are appropriate. (Hints: The
lateral load to cause the maximum allowable bending stress is about 40 kips.
Because several computations are required, the work may be divided among
severa! students.)
REFERENCES 319
REFERENCES
Blaney, G. W., Kausel, E., and Roesset, J. M., Dynamic Stiffness of Piles, Proceedings
of the 2nd Intemational Conference on Numerical Methods of Geomechanics,
Blacksburg, VA, 1975.
Coyle, H. M., and Reese, L. C., Load Transfer for Axially Loaded Piles in Clay, 1-26,
Joumal of the Soil Mechanics and Foundations Division ASCE 92, SM2, 1966.
Det Norske Veritas, Rules for the Desígn, Constructíon and Inspection of Offshore
Structures, Veritsveien 1, 1322 Hovek, Norway, 1977.
Ensoft, Inc., Analysis of a Group of Piles Subjected to .4xial and Lateral Loadíng,
Austin, TX, 1999.
Emrich, W. J., Sampling of Soils and Rock, Performance Study of Soil Sampler for
Deep Penetration Marine Boríngs, ASTM STP 483, 1971.
Fisk, H. N., Nearsurface Sedíments on the Continental Shelf Off Louisiana, Proceed-
ings ofthe 8th Texas Conference on Soil Mechanics and Foundation Engineeríng,
Bureau of Engineering Research, University of Texas, Austin, 1956.
Hetenyi, M., Beams on Elastic Foundation, University of Michigan Press, Ann Arbor,
1946.
Hvorslev, M. J., Subsurface Exploration and Sampling of Soils for Civil Engíneering
Purposes, Waterways Experiment Station, Corps of Engineers, Vicksburg, MI,
1949.
Matlock, H., Correlations for Design of Laterally Loaded Piles in Soft Clay, OTC
1204, Proceedings of the Offshore Technology Conference, 1970.
McClelland, B., Engineering Properties of Soils on the Continental Shelf of the Gulf of
Mexico, Proceedings of the 8th Texas Conference on Soil Mechanics and Founda-
tion Engineering, Bureau of Engineering Research, Uníversity of Texas, Austin,
1956.
Poulos, H. G., Behavior of Laterally Loaded Piles: Pile Groups, Journal of the Soil
Mechanics and Foundations Division ASCE 97, SM5, 1971.
Reese, L. C., and Van lmpe, W., Single Piles and Pile Groups Under Lateral Loading,
Balkema, Rotterdam, 2001.
Reese, L. C., and Wang, S. T., TZPile for Windows, Ensoft, lnc., Austin, TX 1996.
Roesset, J. M., and Kausel, E., Dynamic Soil Structure lnteraction, 2nd International
Conference on Numerical Methods of Geomechanics, Blacksburg, VA, 1975.
Skempton, A. W., The Bearing Capacity of Clay, Proceedings of the Building Research
Congress, Division 1, London, 1951.
Tomlinson, M. J., Foundation Design and Construction, fourth ed., Pitman, London,
1980.
Zeevaert, L., Foundation Engineering for Difficult Subsoil Conditions, second ed.,
Van Nostrand Reinhold, New York, 1983.
CONVERSION TABLE
Customary U.S. (English) Units to SI (Metric) Units
321
Index
Damping matrix, 203
Added mass coefficient, 21
Damping ratio, 116
Anchor, 10
Diffraction coefficient, 88, 89
Archimedes' principle, 28
Dirac delta, 118
Autocorrelation, 172, 175, 187
Drag coefficient, 23, 93
Barge motion, 278 Duhamel integral, 120, 174
Beam dynamics: Dynamic magnification, 40, 117
frequencies, mode shapes, 254-258
Earthquakes, 36, 120-123, 231
responses, 265
Energy:
Bernoulli-Euler equation, 251, 276
kinetic, 106-108, 208, 211
Buckling, 110
potential, 106, 208, 211
Buoys:
Ensemble average, 171
cylindrical, 131-138
Ergodic process, 171
spherical, 102, 138
Euler's buckling, 110
Buoyancy, 28-32
Expectation operator, 166, 186
Cables:
Fast Fourier transform (FFT), 173
arrays, 46
Fatigue, 15
dynamics, 253, 259, 260
Fetch, 154
sag, 45
Fluid momentum, 26
stiffness, 44, 49, 50, 134
Fourier series, 145
Catenary, see Cables
Fourier transform, 174
Clay, 299
Free vibrations:
Compliant tower, 3
linear, 101, 214, 236, 253, 254
Conservative system, 209
nonlinear, 110
Convolution integral, 119, 174
Froude-Krylov forces, 89, 90
Coordinates:
generalized, 196, 208 Gaussian process, 167-169
normal, 215 Gravity platform, 3, 20, 30-32, 55,
Covariance propagation, 185-187 102, 233-236, 243-246
Current, 34, 35 Hamilton's principle, 209
Cylinders: Harmonic response function, 116, 174
flexible, 27, 28, 40, 85
rigid, 22 Ice loading, 38
Impulse response function, 118, 174
Damped motion, 28, 86 Inertia coefficient, 21, 93
Damping: ITTC, 155
linear viscous, 134
Rayleigh, 204, 217 Jacket template platform, 3, 5, 197,
velocity squared, 251 206, 243-246
323
324 INDEX