215233908
215233908
By
SUDIPTA SANKAR BANDYOPADHYAY
Bachelor of Civil Engineering
Jadavpur University
Calcutta, India
1~0
Master of Engineering
Indian Institute of Science
Bangalore, India
1971
Thesis Approved:
1032356
ii
ACKNm~LEDGMENTS
advice any time the author needed it. The author is very grateful to
him.
Dr. Alex L. Matheson, for his assistance in mathematics.
Fellow graduate students, notably Mr. N. Ravishankar and.Mr. R.
Viswanathan, for their friendship and assistance.
Ms. Charlene Fries, for her excellent and meticulous typing.
The author is greatly indebted to his parents and to his wife for
their sacrifices and encouragement.
iii
TABLE OF CONTENTS
Chapter Page
I. INTRODUCTION 1
2.1 Introduction . . . .. . 6
2.2 Static Load Solutions • o' 7
2.3 Dynamic Load Solutions 11
II I. ANALYSIS •, 16
3. 1 Governing Equations . ~ . . 16
3.2 Application of the Complex Fourier
Transformation . . . . . 20
3.3 Partial Fractionalization 22
3.4 Inverse Transform 24
IV. RESULTS AND DISCUSSION . . . . . . 32
4.1 Viscoelastic Models With Four Elements 32
4.2 Elastic Response . . . . . . . . . . . . . . . 33
4.3 Elastic Plate on Kelvin Foundation . . . 34
4.4 Numerical Results . . . . . . . . . . . 36
4.5 Discussion of Results . . . . . . . . . . 36
4.6 Comparative Study of Different Viscoelastic
Models . . . . . 47
4.7 Numerical Example 52
V. SUMMARY AND CONCLUSIONS 56
5.1 Summary . . . . 56
5.2 Conclusions . . . . . . . 57
5.3 Suggestions 'for Future Work 58
iv
TABLE
Table Page
I. Values of Maximum Deflection . . . . . . . . . . . . . . . . . 55
LIST OF FIGURES
Figure Page
1. Viscoelastic Models . ... . 4
v
NOMENCLATURE
Al constant
A2 constant
8 constant
81 constant
82 constant
83 constant
84 constant
c root of the characteristic equation I
c constant
cl constant
c2 constant
c3 constant
c4 constant
d root of the characteristic equation
0 constant
o· flexural rigidity of the pavement
Dl constant
02 constant
03 constant
04 constant
vi
E constant
f root of the characteristic equation
F constant
g root of the characteristic equation
H thickness of the pavement
H( ) generalized function
i square root of minus one
k1 subgrade elastic constant
k2 subgrade elastic constant
m ratio of subgrade elastic constants
M1 constant
M2 constant
M3 constant
M4 constant
n real number
N1 constant
N2 constant
N3 constant
N4 constant
p foundation reaction
P dimensionless quantity
P1 constant
P2 constant
P3 constant
P4 constant
q surface loading
Q1 constant
vii
Q2 constant
Q3 constant
Q4 constant
r moving Cartesian coordinate of the plate
R dimensionless distance
s transformed complex plane
complex root
complex root
complex root
complex root
generalized function
time
v velocity of the load
critical velocity of the load
w midplane deflection of the slab
deflection of the slab under a static load
w dimensionless deflection of the slab
X fixed Cartesian coordinate
y fixed Cartesian coordinate
~
4/lllf.-
Dirac delta function
dimensionless damping constant of the subgrade
damping constant of the subgrade
damping constant of the subgrade
velocity ratio
dimensionless damping constant of the subgrade
viii
Poisson's ratio
mass density of the slab
ix
CHAPTER I
INTRODUCTION
1. i . 1 Genera 1
1
2
Since highway pavement materials have both elastic and viscous prop-
erties (77), the stress-strain relations for the materials are not con-
stant but vary with time. Inasmuch as the load conditions which may be
imposed upon such pavements cover a wide range in time--from the essenti-
ally static condition associated with vehicle parking areas to the rapid-
ly applied repeated loads occurring on heavy-duty highways and airfield
taxiways--it would seem appropriate to apply the principles of visco-
elastic theory in preparing a more rigorous approach to the analysis of
these structures. The many unsolved problems posed by increasing traffic
demands, both in magnitude and frequency of loading, fortify the argument
for such an approach.
Recent researches (12} (13) (52) (67} (93) (98) indicate that physi-
cal behavior of soils can be expressed in terms of viscoelastic para-
meters. Essentially, viscoelastic models are composed of two basic
elements--purely elastic springs and purely viscqus dashpots. These
elements are combined into varipus parallel and series configurations to
produce mathematical expressions for stress-strain-time relations which
may suit a given material under study. The most familiar models are the
Voight or Kelvin and the Maxwell. Unfortunately, neither of these simple
3
1.3. Assumptions
REVIEW OF LITERATURE
2.1 Introduction
ed in Wayne County, M·ichigan in· 1909. These pavements were not designed
in the usual sense of the word beca.use. no rational theory of pavement
design existed at that time. During World War II, there was a signifi-
cant increase in volume and weight of truck traffic on highways and a
large increase in wheel load ·and tire pressure of military aircraft on
airport pavements. In addition, adequate maintenance during this period
was almost impossible, leaving many of these pavements in serious need
6
7
of repair and upgrading. Added to this was the need for many miles of
additional pavements to satisfy the tremendous growth of rural and metro-
politan areas. Consequently, research programs in all phases of pavement
technology were intensified to meet the ever increasing demands of post
war traffic. This research ultimately led to procedures which form the
core of pavement technology today.
of view, Vlasov (106) and Reissner (87) suggested other models. Relaxing
the assumption of equal properties in tension and compression in a
Winkler model, Tsai and Westman (104) treated the foundation as a bilin-
ear material, the tensionless found~tion being examined as a special
case. Later, Lin, Swartz and Williams (59) extended the previous work
9
Unlike the static load case, the probl~m bf moving load on rigid
pavement has received very 1ittl e attention until recently. Previous
works by Raleigh (84) and Lamb (53) in the later nineteenth century con-
cerning the vibration of bars, membranes and plates were primarily stud-
ies in mathematics. Later Ritz (89) elaborated on this work and made a
significant contribution towards the study of vibrating rectangular
plates.
Of late, engineers have felt the necessity to study the dynamic
response of pavements, and several soltuions have evolved. Pioneering
these solutions was the work of Timoshenko (103), Hovey (42),and Ludwig
(61) in their studies of the dynamics of rails subjected to moving loads.
In 1943~ Dorr (19), using Fourier integrals, extended the idea to a beam.
When a plate is subjected to transverse load of constant intensity which
moves parallel to the surface of the plate, the stresses induced in the
plate depend not only on the magnitude of the loads, but also strongly on
their speed of propagation. This phenomenon has been investigated for
simply supported, rectangular plates by Schmidt (95) and for the case of
a simply supported rectangular plate resting on an elastic foundation by
Hall (38}. Livesley (60) considered the response of a finite plate on
an elastic foundation to a traveling load. In all of these cases, criti-
cal speed of propagation of the load is shown to exist and the effect of
damping is neglected. Thus, deflections become unbounded when the load
propagates with a speed equal to a critical speed. In addition, most of
the solutions are restricted to sub-critical speed.
12
ANALYSIS
4 4 4 2
o• (~ + 2 a w +~)+pH~= q(x,y,t) - p{x,y,t) {3.1)
ax 4 ax 2 ay 2 ay 4 at 2
where
o· = flexural rigidity of the slab;
w =mid-plane deflection of the slab (positive downward);
x,y = fixed coordinates;
p = density of the slab;
H = slab thickness;
q = surface loading;
p = foundation reaction; and
t = time.
It is assumed that the slab is supported by a standard solid model.
The relationship between the deflection and the foundation pressure can
then be written as:
nl a . kl + k2
P + - E.£. = k2w + nl ( ) aw (3.2}
k1 at k1 at
where
16
17
4 2
D ~ + pH a w = F(x,t) - p(x,t) (3.3)
ax 4 at 2
6(x) = oo when x = 0
6(x) = 0 when x ; 0
00
J 6(x)dx =1
-oo
r = vt·- x (3.5)
18
(3.6)
(3.7)
where
and
19
f3 =
Y§
40
R= sr
6 =-
v
vcr
where
4 2
d W(R} + 4s 2 d W(R} + 4P(R} = 8o(R} (3.8}
dR 4 dR 2
p + e(l - m}z; dP(R} = W(R} + ~ dW(R} (3.9)
m dR m dR
20
The problem now is to determine W(R) from Equations (3.8) and (3.9)
under the condition that
If the complex Fourier transforms of W(R) and P(R) are W(s) and
P(s), respectively, Equations (3.8) and (3.9) after the transforms be-
come
Equations (3.12) and (3.13) are obtained under the following bound-
ary conditions:
W' -+ 0
in which
(3.17)
With the Budan-Fourier theorem, it can be shown that the number of
real roots of Equation (3.17) does not exceed three. Because the coeffi-
cients of Equation (3.17) are all real, the remaining roots are all con-
jugate complex roots. Returning to Equation (3.16), the function F(s}
may then assume any of the following forms:
{3.18)
22
in which
sl ,2 = ~ d - i c
1. f
s3,4 = ~ 9 -
(3.21)
in which A, B, C, D, and E are all constants. Therefore~
8[ 1 - e(l-m)z;.]
m 1s = A(s .- s 1 )( s - s 2 )(. s - s 3 )( s - s 4 )
(3.22)
23
8{Ml - iN }{Pl + iQ }
1 1
B = --'----.i:-----'--;:---'--
2d(Pf + Qf)
8{M 1 + iN 1 HP 1 - iQ 1}
c= --'----.i;-----'-..,.---'--
2d(P2 + Q2)
1 1
8{M 2 - iN 2}{P 2 + iQ 2}
D= 2 2
2g(P2 + Q2)
8{M 2 + iN 2}{P 2 - iQ 2}
E = 2 2
2g(P2 + Q2)
where
a(l-m)r,;
M
1 =1 - m
c
M =1-a(l-m)r,;f
2 m
N = a(l- m)r,; d
1 m
N =a(l-m)r,; 9
2 m
With Equation (3.21) and the related Equations (3.23) and (3.24),
it is found, after some arithmetic, that
W= A1[sgn(a)]e-aR H[sgn(a)R]
in which
25
8 M1Q1 - N1P1
B2 = . e(l-m)r;
m
d(P2 + Q2)
1 1
83 =
8 Ml2 + N2Q2
2 2
.e(l-m)r;
m
g{P2 + Q2)
B4 =
8 M2Q2 - Nl2
2 2
. e(l-m)r;
m
g{P2 + Q2)
M =-
M* (3.27)
Mo .
in which
F
M = __Q_ {3.28)
o 4e
M= - l2 w· (3.29)
(3.31)
(3.32)
(3.33)
In addition to the dimensionless quantities already introduced for
the Standard Solid model, another dimensionless quantity needs to be
introduced, defined by
28
A. = (3.34)
(3.35)
2 2. . 2 2
p + ~ (t..m + r;) dP + ~ (l _ m) .9__i_ = ~ d w + et.. dw
m dR m dR2 m dR2 dR
(3.36)
Equations (3.35) and (3.36) can now be combined separately with the
I
2
8 [ 1 _ e ( t..m + z;) is _ e A. z;( 1 - m) s2]
W(s) = m m (3.38)
F2 (s)
where
(3.39)
29
(3.40)
Equations (3.39) and (3.40) are sixth-order equations and for the range
of values considered here assume the following form:
(3.41)
in which
s3,4 = ~ g - if
Following what has been done in section 3.3 and noting that the inverse
transform of
~ is - t sgn(R) (3.42)
A2 R
W= 2 [sgn(R) - B[sgn(a)]e-a H[sgn(a)R]
2
8[ 1 _ e (Am + r;) a + e Ar; ( 1 - m) a2 ]
m · m m
B= - a{d 2 + (c- a) 2Hg 2 + (f- a) 2} • e2Adl- m)
30
2
B[l _ e(A + mz;;) a + e Al;(l- m) a2]
m m m
e2AI;;( 1 - m)
Other constants in Equation (3.43) are the same for Burger•s and the
m
2
e Al;{l-m)
2
M3 =1 _ e(Amm+ d c _ e Al;~-m) (d2 _ c2)
2
N =e Adl - m) 2cd _ e (Am + d d
3 m m
2
M =1 _ e (Am + ~;; ) f _ e A~;; ( 1 - m) ( g 2 _ f2)
4 m m
P3 = {d 2 2 2 2
- c(c-a)H(d - g ) - (c-f )}
- (c - f) 2}
- (f - c) 2}.
CHAPTER IV
It can be seen from Equation (3.43) that although W', W11 , and W111
are zero at R =co, Wis not and, therefore, it does not satisfy the pri-
mary condition that the deflection Wmust be zer:o at an infinite distance.
I
The constant term~ [sgn(R)] in Equation (3.43) jis the direct contribu-
tion of the real root zero in Equation (3.41). 1This real root zero
appears whenever there is a dashpot connected in series with the Standard
Solid, VanDerPoel, or Kelvin model. It is found that the value of the
constant A in Equation (3.43) is directly dependent on the value of n2 ,
and therefore on A. For values of n2 much greater than n1 or, in other
words, when A becomes much greater than s, the numerical value of A be-
comes negligible and therefore the deflection giv~n by Equation (3.43)
satisfies all the boundary conditions given by Equation (3.10). Inter-
estingly enough, values of n1 and n2 , as suggested by Secor and Monismith
(97) from their creep test results on asphaltic concrete, show that A is
about 4000 times large than s; consequently, the numerical value of A
becomes almost zero an Equation (3.43) becomes a valid solution. It
should be noted that a A+co, Burger's model becomes VanDerPoel's model
and the Four Element m del becomes a Standard Solid model. In fact, for
A > 50, the deflection val~es of the two viscoelastic models with four
32
33
Both the Standard Solid and Van Der Poel models exhibit an initial
elastic response and delayed elasticity. Two elastic responses are thus
always associated with each model. The elastic responses are limit cases
of the viscoelastic responses because they correspond to ~ = oo and ~ = 0.
i
For a Standard Solid model it can easily be seen from the relation be-
tween the foundation pressure and the plate deflection Equation (3.2)
that the elastic constants, k1 + k2 (initial elasticity) and k2 (delayed
elasticity), correspond to~= oo and~ = 0, respectively. For the Van
Der Poel model, as can be seen from Equation (3.31), the elastic con-
stants are k1 (initial elasticity) and mk 2 (delayed elasticity). The
complex Fourier transform of the deflections for the four elastic cases
follow from Equation (3.15).
Standard Solid:
Elastic Case I: ( 4. 1 )
VanDerPoel:
= 8
Elastic Case III: ~ --.-----,..-~..::,__---~ (4.3)
s4 - 482s 2 + 4m ( 1 - r
m 1
8
Elastic Case IV: ~ = 0, W(s) 4 = -4.,._-----.2----=2-- (4.4)
s - 48 ·s + 4m
34
The inverse transform of a function \'lith poles on the real axis can
be determined by making the convention that the integration is not exact-
ly along the real axis but is along a line an infinitesimal amount above
the real axis. An additional requirement is that the function should
i
have simple poles only on the real axis. It is 'noted from Equations
(4.1) through (4.4) that for e = (1- m)- 1/ 4 , 1, [m(l- m)- 1] 114, and m1/ 4 ,
respectively, W(s) 1 , W(s) 2 , W(s) 3 , and W(s) 4 have poles of the second
order on the real axis, and as a consequenc~, inverse transforms do not
exist. For Elastic Cases I and II the poles are in the complex plane,
if e < (1- mf l/ 4 and e < 1, respectively, and the inverse transforms
are then readily evaluated. For Elastic Cases III and IV the poles are
in the complex plane, if e < [m(l-m)- 1] 1/ 4 and e < m1/ 4 .
- 8
W(s) = 4 2 2 (4.5)
s - 4e s - 4ies + 4
where
- 8Q5
Dl - d(P2 + Q2)
5 5
8P 5
D = - --..---'--..--
2 d(P2 + Q2)
5 5
8Q6
D = --..-----'-....,.--
3 g(P~ + Q~)
8P 6
D = - ---=------'--=--
4 g(P2 + Q2)
6 6
increases (i.e., 6 > O),the point of maximum deflection falls behind the
load; then the load appears to be imposed on the inclined side of the
trough. All profiles computed in this work show the trend of maximum
deflection falling behind the load as speed increases. To the extent of
the present calculations, however, the point of load application never
rises above the level of the undeflected pavement surface in 11 Climbing
out" of the trough with increasing speed.
Beginning with the symmetrical condition at zero speed, ahead of
the load the wave length becomes shorter with increasing speed; behind
the load the wave length becomes extended to where, at supercritical
velocity, no oscillatory waveform will ever be obtained. This pattern
appears to be a kind of 11 Doppler 11 effect. A complete discussion to de-
fine this matter becomes increasingly complex because in elastic solids
there are two conditions of sonic speed corresponding to the propagation
velocities of longitudinal, compression and tension waves, and of trans-
verse shear waves. Wave interactions then tend to obscure a simple
phenomenological picture.
In comparing the displacement curves for 6 = 0.5, Figures 2, 3, and
4, it is noticed that for the small value of t. the deflections are only
slightly different from the deflections for t = 0 (Elastic Case II). As
t increases, the maximum deflection tends to fall behind the load more
" '
~
'\.
'I ' / ,/
L'
/
.
~' //
-
-------Elastic Case I
/
2
- ~-
"
----
~
...... ~ .
L/ Elastic Case II
4 3 2 1 0 -1 -2 -3 -4
R
w
o:>
-1
Standard Solid
--- ---
l~
4 3 2 0 -1 -2 -3 -4
R
w
<.0
-1
Standard S Dlid
-
_....
~~
0 t--- --- =-.: .::::-::::-:;::::;::
~ ....
... ...
/:)
...,..-_.0
vr= m· = 0 . 5 , = 0. 5
~~--
~:;
...-~
.....
l .. - ~ / - - - - - m=0.5, ~:;=1.0
~
0
-1 ~--~--~-------r------~--------+-------~--~~-4--~~--~----~
01- __ _
£-
..... ~--
Standard So 1i d
H
\.~ I I '' ' ' I
" .. ".. I
I ~ f I I '-< I
11 I ',
...
...
2~------~------~~------+--------+--r-----+--------r------~~----~
5 ~ 3 2 0 -1 -2 -3
R
..j::o
Figure 5. Deflection Was a Function of R fore= 1.0 --'
-1
0 low .::": - :J: w """" - ...:: :3o-1L- - ...-.;;: T •......,. '' I I / J "/ /' I '~
~~
Standard Solid
I I I I '
, ,. ', , . , +11
.... - ;.,
'-............... "· ....
I m=o.5z:;=o.5
-------m=0.5z:;= 1.0
- · - · - · m = 0.5 z:; = 2.0
2l I -·· -··- m = 0. 5 z:; = a
5 4 3 2 1 0 -1 -2 -3
R
~
N
-1
0
Ke 1vi,.
w
1 -------m=0.5 Z:=0.5
rd Solid m= 1. 0 1;; = 0. 5
Van Der
---·-m=0.5 1;; = 0. 5
2 ,_
5 4 3 2 0 -1 -2 -3
R
.j::=.
w
44
St ndard Solid
I'; = 0.10
I'; = 0.25
~ 3
= 0.40
E
:3: I'; = 0.55
s::::
0 I'; = 0.85
.j.J
tO
u
I'; = l. 30
•r-
4-
•r-
I'; = 2.00
.....-
c..
E
c::r.:
s::::
0
2
•r-
.j.J
u
QJ
.....-
4-
QJ
a
E
:::J
E
><
tO
:::::
0~------------~-------------L------------~
2 3
0
Velocity Ratio e
Figure 8. Stability Diagram for Surface Displacement
Behind the Load
46
X
ttl
E
::: = 0.10
c:
0
...... = 0.25
+.>
ttl
u 2 = 0.40
......
1+-
...... = 0.55
......
c..
E
= 0.70
c:c
c:
= 0.85
0
......
+-'
= l. 30
u
Q)
...... = 3.00
1+-
Q)
0
E
::::::1
E
......
X
ttl
::E
0 2 3
Velocity Ratio e
Figure 9. Stability Diagram for Surface Displacement
Ahead of Load
47
4
Stand rd Solid
X
ttl 2
E
:::: e = 1.1
e = 1.0
e = 3.0
0~------~--------~---------L------~
0 0.5 1.0 1.5 2.0
1:;
~1
I I I
ndard Solid
I ............. "<:::
"
/:! ~ I
- - - - - - - m = 0.5
I I
z;; = l.O
0
---
-1
~------~------~--------~------~--------~--~--~--------~------~
4 3 2 0 -1 -2 -3 -4
R
+=>
1.0
21 I I I I Van Der roe1 =-", I I \ I I
\
\ \
.,.,.......-
/'
""
t·1 0 k -~ L I j.....C _, / I I' \\ \ I I , v ,
7 I \\ "~ I
\
~1 I I I I I'""\ '" L Jf/1
;rJf'
/ I \ I
\
4 3 2 0 -1 -2 -3 -4
R
Van Der
M0 I j I 1---- ~ +k \ I I 1\ ~
-1
4 3 2 1 0 -1 -2 -3 -4
R
U1
52
than Elastic Case III; therefore, the nature of the deflection profile
is self-explanatory.
In Figures 3, 5, 7, 10, 11, and 12, deflections and moments in the
slab are plotted against the position of the load for different values
of velocity ratio e for Kelvin, VanDerPoel, and Standard Solid sub-
grades. Although, in general, the nature of the deflection and moment
profiles is the same for the three models, the relative difference in
magnitude of maximum deflection and maximum moment between Van Der Poel
and Kelvin or Standard Solid is as great as 100 percent. The significant
difference in amplitude between ~lvin and Van Der Poel is due to initial
elasticity, and the difference between the Standard Solid and the Van Der
Poel is due to lesser initial elastic deformation, because Standard Solid
has stiffer elastic constants.
Although recent investigations established beyond doubt that the
subgrade has both elastic and viscous properties, it is not possible yet
to concl~de which model simulates the subgrade most realistically because
of a paucity of available data in the current literature. A comprehen-
sive experimental program is required to relate soil to a specific model.
Until more is known about the elastic and viscous parameters of soils,
the Standard Solid model seems to be the logical choice to idealize the
subgrade.
one obtains:
Eh 3 8 .
D• = _--=..:.;..----:::2~ = 1 . 7460 x 10 psi I in.
12(1-].l)
s 'w
= 4fT2 = 1.9452 x 10- 2 rad/in.
4K o· 114
Vcr = [(p~)2] = 290.96 ft/sec = 198.4 mph
The deflections and positions of the maximum deflection fore and aft of
the load can now be found as follows:
At e = 0.3 (59.5 mph) and s = 0.5,
(~) -2
= +1.02998 w+max = l. 02998 x l. 215 x 10 =0. 0125 in.
wo +max
(~) = -0.064 W
-rna
X= -0.064 X 1.215 X 10-2 = -0.000
·
8 1n.
.
wo -max
(s r) +max = 0. 045 r+max =0.045/0.019452 = 2.3 in.
54
0.3 59.5 0.5 +1.02998 +0.0125 -0.064 -0.0008 +0.045 +2.3 -3.00 -154.2
0.3 59.5 1.0 +0.98756 +0.0120 -0.064 -0.0008 +0.070 +3.6 -2.50 -128.5
0.3 59.5 2.0 +0.89789 +0.0109 -0.080 -0.0007 +0.090 +4.6 -2.50 -128.5
0.4 79.3 0.5 +1.05601 +0.0128 -0.080 -0.0010 +0.060 +3.1 -2.50 -128.5
0.4 79.3 1.0 +0.98733 +0.0120 -0.079 -0.0009 +0.080 +4.1 -2.50 -128.5
0.4 79.3 .2.0 +0.87560 +0.0106 -0.067 -0.0008 +0.090 +4.6 -2.50 -128.5
U1
U1
CHAPTER V
5.1 Summary
56
57
5.2 Conclusions
depending on the model. For heavy damping, the maximum deflection for
the Kelvin model is always less than the static value. It never gets
lower than the static value for the Van Der Poel model and is between
these two values for the Standard Solid Model.
7. Due to the paucity of experimental data available in literature,
it cannot be concluded which viscoelastic model gives the best approxi-
mation for the actual subgrade. However, it can be concluded that the
Standard Solid subgrade give deflection profiles which fall between those
of the Kelvin and Van Der Poel subgrades, and therefore is recommended by
the author as the model to idealize the subgrade until enough is known
about the actual viscoelastic behavior of soils.
60
61
( 105)