Professional Documents
Culture Documents
Induction Heating Basics PDF
Induction Heating Basics PDF
source. I.e. an electromagnetic field (or electromagnetic radiation) interacts with the part to
be heated (otherwise known as the work-piece), and causes it to be heated. Generally, the
frequency of the electromagnetic field interacting with the work piece characterises the
heating process. The frequency of the electromagnetic radiation can stretch from DC to
beyond daylight.
DC Through
Resistance (DC) through
I2R Heating
Mains-frequency Induction
(50Hz) Deep
Medium-frequency
Induction
(1-20kHz) shallow
Heat Penetration
Radio-frequency
Frequency
Induction
surface
(50kHz-10MHz)
surface
Infrared heating
Daylight Surface
This figure shows most of the different categories of electro-heat, arraigned in an ascending
order of frequency. At DC (or low frequency AC), direct resistance heating heats
conductive work pieces by attaching an electrode to each end of the material to be heated. A
large current is then passed through the work piece. This method features high power
densities and relatively high efficiencies. A typical use for this heating method is for
resistance welding of car bodies, etc.
Direct resistance welding is limited to relatively few materials and heating geometries,
however, as the current will flow through the shortest path between the electrodes, and the
highest resistance part of the circuit will get hottest, so if the work piece is a lower resistivity
than the electrodes, the electrodes will be heated more!
Billet heating by direct resistance heating is rather wasteful for this reason, as the part of the
material under and beyond the electrodes is not heated, and would be scrapped in a forging
application.
Page 1 of 1
A further limitation on the application of direct resistance heating is the contact pressure
required between the electrode and the work piece. This can be considerable, and can
damage some more fragile materials.
It can be seen that induction heating is split into three categories: Mains frequency, medium
frequency and radio frequency. This is for two reasons. The first is that the equipment for
generating the excitation current for the coil (or work head) that generates the time-varying
magnetic field has historically been very different.
MAINS FREQUENCY induction heating systems are excited by direct connection to the
utility supply. The current penetrates deeply into the work-piece and they tend to be used in
very high power systems.
The advent of the THYRISTOR in the 1960s saw most MG sets being replaced by various
different topology inverters. Thyristor-based inverters are still in production, and modern
thyristors can carry several thousand amps and can block several thousand volts, whilst
offering a very low forward conducting voltage drop (1.4V). This capability means that
medium-frequency induction heaters with powers in excess of 500kW are often best
manufactured with IGBT-based inverters. Across the power range 50kW upwards, thyristor
based inverters are still being manufactured, and are characterised by their reliability and
simplicity, and their manufacturing expense and their inflexibility.
Recently (within the last 5 or so years), design emphasis has moved from thyristor bases
inverters to IGBT based inverters. IGBTs offer three principal advantages over thyristors:
The first advantage is that, unlike thyristors, IGBTs turn off when their gate signal is
removed. Once turned on, a thyristor will remain conducting until it is reverse biased. This
Page 2 of 2
means that the power-electronic topology of an IGBT-based inverter can be considerably
simpler. The second major advantage of an IGBT over a thyristor is the packaging
technology is considerably easier to use. These two improvements make the manufacturing
costs of the inverter considerably lower. The final improvement that IGBTs offer is that
they can have considerably faster switching speeds, so the operating frequency range of an
IGBT based system will be extended.
The operation of these inverters will be discussed in more detail at the end of this course
section.
Finally, we have RADIO-FREQUENCY induction heating, which covers the range 50kHz
10MHz +, although it gets difficult to make work-head coils that work at the highest
frequencies. From the 1920s, the excitation has been provided by thermionic valve-based
oscillators, originally made from modified radio transmitters. Themionic valves are
characterised by very fast switching times (ns) and very large forward conduction voltage
drops (100V). To make a workable induction heating system, a very high voltage DC power
supply is used (up to 12kV). Even with this high supply voltage, the efficiencies of these
generators is only in the region of 60%, so, with powers of 2kW to 2MW, it can be seen
that a considerable amount of power is wasted.
This inefficiency, coupled with the safety and reliability concerns associated with a high
voltage DC link, compounded with the high unit cost for a power triode, have led, since the
1980s to the increasing use of MOSFET based inverters. Currently, commercially available
inverters can deliver 1MW at 800kHz, and the frequency range is expanding.
So, to summarise, for medium frequency systems, the THYRISTOR is being replaced by the
IGBT, and for radio frequency systems, the TRIODE is almost completely superseded by
the MOSFET.
Having briefly explored how the different frequencies are generated, we should now look at
why the different frequencies are generated. This is all related to current penetration. At
low frequencies, the magnetic field penetrates deep into the work-piece, which means that
the induced current will flow deep into the work-piece. If the field can penetrate further
than half way through the material, then the field coming from the opposite direction will
cancel it out, as it will be acting in the opposite direction. Therefore, for a given piece of
material, there is a minimum frequency that can be used to heat it. It is this effect that
allows the use of a laminated steel magnetic core on a transformer.
Page 3 of 3
Conversely, if the frequency is too high, only the surface of the material is heated, which
may mean that the applied power has to be limited such that the outside is not overheated
(melted or burnt) before the inside is heated. It may be, of course, that you only want to heat
the outside!
Moving further up the electromagnetic spectrum, we get to possibly the best-known use of
electro-heat: the microwave. Microwaves are used in industrial heating as well as domestic,
for example drying paper as it is produced. As a far-field (i.e. wave) electromagnetic source
is used, care has to be taken when using it, as if any leakage occurs, operators can become
injured.
Microwaves are a shallow heating mechanism, with most of the incident energy being
absorbed within the first ~4cm of the target thickness (dependent on the material and the
frequency).
Page 4 of 4
(
P = A s4 T4 )
where epsilon is a material constant called the emissivity.
Having defined electro heat in terms of operation, its use must be justified. The most
common criticism of electro heat in most forms is its cost, both operational and in terms of
capital plant cost. An example of this can be seen in the small company of R.S. Hall
Engineering Ltd, the small company that I worked in between finishing my degree and
starting my PhD. Seven years ago, they invested in an induction heater to replace a gas
furnace, for heat-treating steel. The induction heater heats steel bar from ambient to 1200C.
As the bar that they were heating could be a minimum of 9.5mm in diameter, they had to use
a 30kHz induction heater (well derive this frequency later), and, to get the throughput they
needed, they chose a 30kW power unit. This cost them 22000. It was estimated that, if
they had rebuilt their gas-powered furnace from scratch, it would have only cost them
8000. Clearly, for them, the initial outlay was offset by other factors.
One major factor is the energy cost. This will be shown (on the next slide) to be reduced.
From a process point of view, one of the most attractive features of electro heat in general is
the quality of the heat control. Most electro heat processes allow an instantaneous power
control with continual variability. Compare this with a coal-powered furnace that can take
several hours to warm up. Further, the power put into the process can be accurately
controlled with an electro heat system, whereas most combustion-based methods are subject
to fuel quality variations.
A further advantage of electro heat is that the power density within the work piece can be
very high. This is because the heat is developed within the work piece and therefore no
conduction process is required. The fact that the heat is developed in the work piece means
that it, rather than a surrounding furnace, is the hottest part in the system. These factors
combine to reduce the heating time, which reduces the material losses due to oxidisation and
chemical (metallurgical) changes such as decarburisation, which traditionally account for a
high proportion of the lost material in combustion-based heating processes.
As well as being able to modulate the power in the time-domain, the special control of the
heat available with induction, dielectric, and, to a lesser extent direct conduction heating, is
considerably better than any other means (such as an oxy-acetylene flame). This has its
advantages: the profiles of heated parts can be accurately controlled, such as the seam of a
plastic weld, or the part of a pop-rivet that bends.
Page 5 of 5
A further tangible improvement that electro heat offers over traditional heating processes it
the elimination of soot and smoke from the workplace. Most electro heat methods are also
much quieter than combustion processes, and therefore the replacement of combustion
heating processes with electro-heat processes results in a workplace that is considerably
more pleasant.
Again, the main criticism of electro heat is the cost. To make a true comparison of each
energy source, it is important to consider its calorific value and to compare the cost on an
equal plane.
Oil=20p/40MJ=0.5p/MJ
Gas=34p/106MJ=0.32p/MJ
Coal=4800p/27500MJ=0.17p/MJ
Electricity=4p/3.6MJ=1.1p/MJ
It is therefore obvious that coal is the cheapest form of energy calorifically, and electricity is
the most expensive, by a factor of about 6.
All we have done so far is to heat a flame the flame then has to heat up a billet, generally
an inefficient process, with half of the heat going up the chimney. Finally, the equipment
utilisation has to be taken account of. With a gas furnace, a warm-up time of over an hour is
unexceptional, which means that the furnace will be left on over coffee breaks etc, wasting
energy.
60/100*50/100*60/100=18%
and the cost of the gas per useful MJ can then be found
0.32/0.18=1.77p/MJ
Page 6 of 6
A similar analysis can be carried out on the same process if it were to be done using direct
resistance heating. It can be seen that the system is not 100% efficient either. The burner
(i.e. energy conversion) efficiency is not 100%. This is due to I2R heating of the contact
electrodes etc, and is a function of the relative resistivity of the work piece and the
electrodes.
The heat transfer, though, is 100% efficient, as the heat is generated in the work piece.
The heat utilisation is not 100% because of the contact area. The metal under the contact
area is not heated uniformly, and may have to be scrapped, so the energy that has gone into
that part of the work-piece is wasted.
This gives an overall efficiency of 72%, and a cost per useful MJ of 1.52p, considerably
cheaper than the cost of gas! The electro-heat process suddenly seem more economically
viable!
Unfortunately, with direct resistance heating, the ends of the work-piece dont get hot. In a
forging application, this can mean that 10% of the material can be wasted.
For the forging case, the unheated part of the billet is generally scrap metal, which
introduces further costs into the equation. However, metal loss as scale (oxides), which can
be significant in a gas-fired system, is almost eliminated.
With induction heating, the heated length can be as long as the work-piece. The inverter can
be up to 98% efficient, and the coil can be over 90% efficient. The near-instantaneous
power means that the heat utilisation can be 100%, but there is loss due to radiation etc.
from the work piece. It is therefore possible to get a total energy efficiency of 80% from an
induction heating system with only minimal material wastage.
Both combustion furnaces and direct resistance heaters can give rise to a significant degree
of material wastage: 3% due to oxidation and metallurgical change in the case of
combustion furnaces, and up to 10% due to poor current distribution in the case of resistance
heaters. This increased material wastage can mean that the direct resistance heater is not an
economical choice for a process heating for billet forging applications. With its lack of
contact and increased power density, coupled with its ability to heat 100% of the billet
length with good heat distribution, induction heating can be a more economical choice.
Page 7 of 7
This picture shows a work head connected to a 7 kW 150kHz (i.e. RF) induction heater,
heating an M10 bolt to well over 1000oC. As we shall see over the next few weeks, the
efficiency is a function of the material being heated, along with the heating profile, and, as
such, is variable. With billet heating, a heating efficiency of 70% is reasonable.
With induction heating, as with direct resistance heating, the heat is directly generated
within the material, so the heat transfer efficiency is 100%.
The heat utilisation, like the efficiency, is a function of the required heating pattern. It is
easy to heat 100% of the length of a billet; you simply put it in a long work-head coil. It is
more difficult to heat a narrow band, as focussing the magnetic flux can only be done to a
certain degree, with fringing over the (required) air-gap between the work-head and work-
piece spreading the heating somewhat. This means that although it is possible to make a hot
spot, it is difficult to avoid a warm area around it.
As induction heaters tend to give instantaneous heating control and can draw no quiescent
power, the time utilisation of energy can be 100% (i.e. when its not heating, its not
drawing power)
Therefore, for billet heating, we can get a 70% total energy efficiency, coupled with a sub-
1% total material wastage, if scale/decarburisation wastage is accounted for.
So, how does induction heating actually work? As briefly discussed in the last lecture,
induction heating works by inducing a current into a conducting object. It does it like this:
First, a copper coil (often a solenoid, but not exclusively), has a large, time varying current
set up in it by the imposition of a time-varying voltage across it (generally in the form of a
sine wave).
Page 8 of 8
NI
This current then creates a time-varying magnetic field (for solenoid H = ), which will
l
cause a time varying flux ( B = H ).
If a conducting object is placed in the field, then a voltage will be induced around it
d
(E = , = BA ).
dt
If the conducting object is a closed ring, then, like a shorted turn in a transformer, then the
voltage will cause a current to flow around the outside of the object.
V
V = I (R + jX ) . I =
R + jX
The allowance for an impedance has to be made as this is an AC system:- If it were Dc, then
d
the rate of change of flux with time ( ) would be zero, so no current would be induced.
dt
Finally, this induced current causes I2R losses in the work piece, which makes this method
of heating effectively a resistance heating method, albeit with the current flowing at right
angles to that of direct resistance heating (i.e. around the billet rather than along it).
Having shown in some more detail how the basic principle of induction heating works by
basically considering the current flow in a thin tubular foil, we shall now look at what
happens to the induced current when induction-heating a solid work piece.
The answer to this question is fairly mathematically heavy, and to go too deeply into it
would be a bit of a waste of time. Therefore, I will simply give you a hand-waving
account of how the field, and therefore the current, is distributed throughout the materials,
followed by the analytical solution. This way we avoid vector integration, Bessel functions
etc.
Page 9 of 9
Ho
H1 z
x
H2
H3
y
To avoid having to discuss flux return paths and end effects, we consider a semi-infinite slab
of material being heated by an infinite 2-diamentional sheet of current just above it. This
diagram shows a finite part of a cross-section of the infinite set-up. The sheet of current that
represents the work head stretches infinitely to the right and left (in the x direction), and
infinitely forward and backward, or into and out of the page (the z direction). It occupies no
space a all in the y direction.
The semi-infinite slab that represents the work-piece also extends infinitely into and out of
the page, and infinitely to the left and right, but it goes from y=0 to y=-.
To see where the current goes, we can turn the homogeneous slab into a series of thin slices.
Consider the top slice. It has a time varying magnetic field, H 0 cos(t ) acting on it. In the
same way as the last slide, this will have a current density induced in it, J 0 cos(t + ) . The
phase shift (lag) is due to the inductance of the slab causing a lag between the EMF induced
in the slice and the current flowing through it.
This current density in the slab creates an opposing magnetic field in the slab, marked as H1.
The resistivity and the inductance of the slab reduce the magnitude of the current, and hence
the field, so H1 is smaller than H0.
Now consider the strip below. It sees a field that is equal to the vector sum of H0+H1 so,
this strip sees an attenuated field, as H0 opposes H1. This strip will therefore have an
attenuated current density induced in it J 1 cos(t + 1 ) . This attenuated current density
creates a magnetic field H2.
The third strip down will see a field made up of the vector sum of H0,H1 and H2, i.e. a
further attenuated field, which will induce a smaller still current density, making the
resultant total field smaller and smaller as you go down through the y axis.
This effect, known as the skin effect, means that the field, and hence the loss, or heating
effect, is concentrated on the surface of the work-piece.
Page 10 of 10
It can therefore be shown, by letting the thickness of the slices tend to zero, and solving the
resultant differential equation, that the field in the x direction, the current in the z direction,
and the flux in the x direction all follow this form:
H x ( y ) = H x (0 )
y
J z ( y ) = J z (0 ) e
cos t y
x ( y ) = x (0 )
which is an exponential decay multiplied by an oscillating term with a variable phase shift.
1
y
0.8
Magnitude attenuation = e
0.6
|J/J0|
0.2
0
0 1 2 3 4 5
y/
This assumes that the current flowing through the coil is sinusoidal. Generally, this is the
case, with only a small distortion. The reason for this will be made more clear when we
have discussed how the coil is connected to the excitation circuit.
These terms are only true for a semi-infinite slab, so they have no direct use. However, they
are simple, and most of induction heating theory is based on them.
The most important part of the equations is the term . This is the skin depth, or the
penetration depth, and is the depth into the slab at which the current etc. has fallen to 1/e
(i.e. 1/natral number) of its surface value.
2
=
On inspection of the formula for the penetration depth, it can be seen that the heating depth
is a function of the resistivity, the permeability and the frequency. As the resistivity and the
Page 11 of 11
permeability of the work-piece are fixed by its material, the only way to actively control the
penetration depth of the current into the material is to alter the frequency. This is why
induction heating systems are divided into the three different frequency bands:
Mains used for through heating large pieces of metal (gas cylinders etc)
MF used for through heating smaller billets and strips down to 15mm
R.F used for surface heating or for heating very small pieces.
Although the current (and therefore the heat) is induced into the very surface of the material
under an RF magnetic field, RF induction heaters can be used for through heating by
allowing the heat to be conducted in through the material. This limits the rate at which the
material can be heated, as too high a power will result in the material melting on the surface
before the inside is even warm!
One other thing to note about the equations is the phase shift in the oscillating term. As the
y position decreases, the current, H-field and flux become more retarded.
The total current (per unit length) in the slab can be found by integrating the current density
from the surface to infinity, with respect to the semi-infinite axis, y.
y
J z ( y ) = J z (0 )e
cos t y
JZ(0) is the surface current density. The phase shift on the current with depth has an effect
on the integration, which evaluates to:
J z ( 0)
I= cos t
2 4
This total current I can be considered to be flowing in 1 skin depth . Therefore, as far as
our semi-infinite slab goes, we have made the y-direction into a sheet of thickness that
carries all of the current in a uniform manner.
The most important bit is that it allows the surface power density to be defined, and from
that, an equivalent circuit.
So far, we have got as far as finding the total current in the work piece, and stating (although
proof can be found in the induction heating handbook by John Davies and Peter Simpson),
that the total current
Page 12 of 12
J z ( 0)
45 0
2
can be considered to be uniformly concentrated in the outer skin depth of the material
2
=
If this is true, then the surface power density can be found, by the use of
l
R=
A
1m
1m
A=1m, l=1m
=I2
in Wm-2
Now, its all very well knowing the power density in the work piece as a function of the
surface current in the work piece, but this has no real connection with induction heating: the
power is provided by a magnetic field, not by direct connection.
The total current density in the coil is equal to the coil current multiplied by the number of
turns divided by the length.
Ic N
J0 = Am-1
l
We can substitute this value for J0 into the surface power density to get
Page 13 of 13
2 2
Ic N / l I N
P= = c Wm-2
2 l 2
From this, we can find the power density due to the coil current in a long solenoid. If we
now consider the field due to a solenoid
NI
H0 = Am-1
l
this is equal to the surface current density. Therefore substituting H0 into the surface power
density equation,
P = H RMS
2
The RMS symbol is inserted as a reminder that the rms value of the field must be used =
H
2
To show the effect of the work piece properties on the surface power density, three
examples are worked through, each with the same field magnitude (100kAm-1) and
frequency (50Hz), but with varying permeabilities and resistivities.
Page 14 of 14
The first case is for mild steel at room temperature. The steel is below the Curie point
(~720oC), at which the ferromagnetic properties of the steel stop, and therefore a magnet
will not stick to it. This means that the relative permeability of the steel is relatively high.
Electrical engineers amongst you will notice that the permeability of the steel is
considerably lower than you would expect, with r being well over 1000 normally.
This discrepancy can be explained by considering the given r as a large signal quantity,
and the expected r being a small signal quantity. If the expected r of 1000 is substituted
into
B = H
B=1000410-7100103=124T
At the surface of the steel, but steel saturates at about 2T. The effect of the saturation is to
reduce the incremental permeability.
The effective relative permeability is therefore reduced to take this saturation into account
i.e. the large signal incremental permeability is used. Effective values of 20 to 50 are often
used for steel under lower frequency induction heating fields.
Anyway, to find the surface power density, we first need to find the current penetration
depth
2 2 200 10 9
= = =0.0045m=4.5mm
50 4 10 7 50 2
Substituting the skin depth into the surface power density equation
(100 10 3 ) 200 10 9
2
P=H 2
rms = =222kWm-2
2 2 0.0045
Note, then, that the hot steel has a much deeper current penetration than the cold, due both
the (relatively slow) increase in the resistivity with temperature, and to the rapid reduction in
the permeability at curie.
Although the resistivity of the hot steel is over 5 times greater than the resistivity of the cold
steel, the losses in the steel are lower, as the current flows in an area that is over 16 times
Page 15 of 15
greater. Remember that the total current in the material is only a function of the H-field, so
the current is the same. The resistance is equal to the resistivity multiplied by the length
divided by the area, so if the penetration depth is smaller, the resistance, and hence the loss,
is higher.
From this, we can see that the steel is heated much faster below the Curie point than above it
for a fixed field strength.
Moving on now to the copper load, we can see that the penetration depth is just over twice
that for the cold steel load. Although the conducting area is similar, the per-unit surface
area effective resistance of the copper is much smaller that that of the steel. It can be seen
that the losses in copper, under the same field strength as the steel, are considerably smaller.
This can cause problems when induction heating copper!
On the other hand, if the H-field source (i.e. the work head) is also considered to be a semi-
infinite slab, starting where the work piece semi-infinite slab ends, then the analysis is valid
for the work head as well as the work piece. This allows the work head losses to be
estimated.
As the H-field at the very surface of the load slab is equal to the H-field at the very surface
of the source slab, all of the previous calculations can be used.
H0(workhead)=H0(workpiece)
Almost invariably, the work-head is made out of copper, although above 10MHz, the copper
is silver plaited to increase the surface conductivity.
As all of the losses worked out on the previous slide were for the same field intensity, the
relative efficiencies can be found for heating steel above and below Curie, and also for
heating copper.
222
E= 100% =96%
222 + 9.16
Page 16 of 16
similarly,
Hot steel=89%
Cold copper=50%
This calculation makes the assumption that the length and the width of the work piece and
the work head are equal. Whilst the width of a practical work piece may be equal to the
width of a practical work head, the length of the work head has to be longer, as it is wrapped
around the work piece. The efficiency is therefore always worse than this calculated value.
By substituting the skin depth equation into the surface power density equation, and
substituting this into the efficiency equation, both in terms of the work-head and the work
piece, then the best-case efficiency can be shown to be
1
Eff =
c
1+
w rw
It is easy to draw the conclusion that for efficient heating, you need a highly conductive coil
and a highly resistive work piece, and if the work piece has a high permeability, then the
efficiency is improved.
Of course, many assumptions have been made here, not least the semi-infinite slab model:
this result suggests that you could induction heat air (resistivity infinite) with 100%
efficiency. This is clearly not the case.
Also, the analysis assumes that the work head is an homogeneous sheet of current, rather
that a series of discrete turns separated by layers of insulation. This means that the area that
the current flows in the work head is reduced, meaning that the local current density is
higher, increasing the losses. This is compounded by the fact that the length of the current
path around a practical work head is longer than the length of the current path around most
practical work pieces.
To make a better model of the work head / work piece pairing; which includes the reactive
as well as the resistive losses, we can create an equivalent circuit by considering the
magnetic flux.
We are first going to derive the EMF induced in the work piece by considering the magnetic
flux in the work piece.
If we start with the flux density. This follows the general form:
Page 17 of 17
y
y
B = B0 e
cos t
where we have the surface value modified in amplitude and phase as you pass through the
material.
B = H
holds, then we can substitute H0 for B0 and relate the flux density to the applied H-field.
We can then find the flux in the work piece by noting that the flux density is the rate of
change of flux with respect to space, so flux is the integral of flux density, with respect to
space.
Therefore, the total flux in the work piece can be found by integrating the flux density over
whole depth of the semi-infinite slab.
0 r
x = H0 cos t
2 4
From the basic electromagnetic definitions, the EMF in a coil is equal to the number of turns
multiplied by the rate of change of the flux contained in the coil with respect to time.
d
E=N
dt
Of course, with the work piece, N=1, but we can refer this voltage back to the coil by
substituting the number of turns into N.
So, differentiating the total flux with respect to time, and multiplying by the number of coil
turns, we get the coil EMF due to the flux in the work piece.
H 0
E = N sin t + Vm-1
2 4
H 0
E = N cos t + Vm-1
2 4
i.e. The voltage leads the H-field by 45o. If we consider that in the case of a long solenoid
Page 18 of 18
NI
H0 = Am-1
l
I 0N 2
E= cos t + Vm-1.
2l 4
I.e. the voltage leads the current by 45o, or, more conventionally, the current lags the voltage
by 45o.
P=VI cos()
where cos() is the power factor, it can be seen that the power factor of the work piece is
cos(45o), or 1/2
N 2 1 1
E=I + j Vm-1
2l 2 2
If we consider this to be in the form of Ohms law, then we can see that the impedance of
the work piece, referred to the coil terminals, is
N 2 1 1
Z= + j m-1
2l 2 2
N 2
allowing equivalent values for the resistance and the reactance to be found: R =
2l
N 2
m-1, X = m-1
2l
XL=L
N 2
L= Hm-1
2l
Page 19 of 19
This per-unit equivalent circuit can be used to analyse a very big (d>20) cylinder with a
solenoid wrapped tightly around it.
N 2
Rload = D
2l
Similarly
N 2
X load = D
2l
Using the concept that the coil can be approximated to a semi-infinite slab, providing that
the conductor is over 10 times thicker than the skin depth of the current, it is possible to
define the equivalent circuit components in a similar manner.
N 2 N 2
Rcoil = k c D , X coil = k c D
2l 2l
Inspection of the equation shows that a coil factor, kc, has been included. This is to account
for the fact that the coil is made up of discrete turns rather than a homogeneous mass.
kc is normally of the range 1.1 to 2, depending on the shape and spacing of the coil
conductors. The lowest end of the range, being the most efficient, is for rectangular
conductors with minimal turn-on-turn insulation.
If round tube is used, then the efficiency is compromised by the reduction in the conducting
area, where only the bottom of the tube caries the full skin-depth of current, so a fair bit of
the area facing the work piece does not carry any current.
Page 20 of 20
Once again, this analysis assumes that the coil is touching the work piece, and that the
diameter is much larger than the penetration depth.
If the coil is not touching the work piece, then there will be some stray flux in the air gap
between the coil and the work piece. This stray flux will cause an EMF to be induced in the
coil, just like the flux in the work piece and the flux in the coil.
We can consider the field H0 to be constant across the air-gap by considering the penetration
depth.
2
=
the resistivity of air is infinite when it is not ionised (i.e. a spark or plasma). Therefore the
penetration depth of magnetic field into air is infinite, which means the rate of change of the
H-field with space is zero across the airgap. This means that the identity
H0(coil)=H0(workpiece)
Remains true in the presence of an air gap. Further, the equations for Rc, Xc, Rw and Xw also
remain correct. However, we now have to consider the flux in the air gap, which will
contribute to the coil terminal voltage.
gap = 0 H 0 Ag
Where Ag, the area of the air-gap, is equal to the area enclosed by the coil minus the area of
the work piece.
Ag =
4
(d 2
coil d wokrpiece
2
)
The coil voltage due to the air-gap can be found by finding the rate of change of the flux in
the air-gap with respect to time.
d gap d NI
Vg = N =N cos(t ) 0 Ag
dt dt l
N 2I
Vg = sin (t ) 0 Ag
l
Page 21 of 21
N 2 I 0 Ag
Vg = j
l
i.e. the voltage in the coil due to the air-gap is totally reactive, which makes sense as the air
does not get hot! The equivalent reactance of the air-gap can be found by dividing the
voltage by the current.
N 2 0 Ag
Xg =
l
Although the equivalent component for the air-gap is totally reactive, it does indirectly
increase the losses in the coil, as it increases the length of the current path in the work head.
Therefore, induction heating systems with a large air-gap are less efficient than those with a
small air-gap.
As the total flux within a coil gives rise to the EMF induced in the coil, it appears as if all of
the EMFs, and therefore all of the equivalent components, are in series. Therefore we have
derived a full equivalent circuit for a coil heating an electrically large load.
Xc Rc Xg
Lw
Rw
Example
This example is for a long thick bar of steel (300mm*2m) being heated from room
temperature to 1200oC under a field being excited at 50Hz. The coil is 500mm in diameter
and has 100 turns. It is the same length as the steel bar. Similar bars of steel are heated at
chesterfield cylinders at a power of 1MW prior to being forged to make gas cylinders.
Initially, we will find the equivalent circuit of the system with the steel at room temperature.
If we compare the skin depth (4.5mm at room temp), with the diameter (300mm), then it can
be seen that the condition given for the work piece to be electrically large is true (20<d).
Page 22 of 22
Following the formula, Rcoil can be found:
c 0 N 2
Rc = d c k c
2l
=0.0216
Xc=0.0216 also
Lc=Xc/=69H.
We have now found the equivalent component values for the coil. Moving on to the work
piece:
w w N 2
Rw = d w
2l
=0.209
Xw=0.209 also
Lw=Xw/=666H.
Finally, the inductance due to the air gap. This is due to the area inside the coil that is not
accounted for by the work piece.
X g = N 2 0Ag
Ag = Ac Aw =
4
(d 2
c d w2 )
Ag =
4
(0.5 2
)
0.3 2 =0.126m
Lg=Xg/=790H
So, if we were to apply 1000V to the terminals of the coil, what power would we get?
Page 23 of 23
First find the coil current
V
I=
Z
Z = R2 + X 2 = (Rw + Rc )2 + (X w + X c + X g )2
1000V
I = =1874A
0.53
The efficiency can be found by finding the useful power divided by the total power.
Pw
E= 100% =734/(734+77)100%=90.5%
Pw + Pc
The semi-infinite slab example, which does not account for the air gap, gives an efficiency
of 96%. It can be seen therefore that the air gap does indeed increase the losses in the coil,
despite dissipating no power itself.
It is all well and good being able to calculate the equivalent circuit values for an electrically
large work-piece in a long solenoid: this covers about 35% of all induction heating cases. A
further 35% of induction heating cases are for systems that are electrically small, and the
rest are for cases where the work head is short it can be down to 1 turn.
One of the best conditions for rapid through heating is to have the skin-depth similar in
magnitude to the diameter, such that the majority of the load is being directly induction
heated, rather than being indirectly heated by thermal conduction from the heated surface.
Having an increased heating penetration gives a smaller temperature gradient through the
material for a given surface power density, allowing the middle to be heated at a much
higher rate than would be the case with an electrically large work piece, without the surface
melting or becoming degraded. As well as increasing the throughput, this is useful, as a
rapidly heated product will spend less time at an elevated temperature, so less material will
be lost through oxidation or metallurgical change.
Page 24 of 24
To enable analysis of magnetically smaller parts, the equations for the resistance and
reactance due to the load are modified to include a correction variable.
N 2 pAw N 2 qAw
Rw = , Xw = ,
lc lc
Aw = d w2 .
4
Substituting Aw = d w2 into the new equations for Rw and Xw, and then comparing these
4
dw
equations with those for electrically large cylinders, it can be seen that p has replaced
2
dw
as the effective width of the current flow in terms of the losses, and q has replaced as
2
the effective width of the current flow in terms of the magnetic stored energy.
These modifiers (p & q) are due to the work-piece not being adequately approximated by a
semi-infinite slab. Their values have been derived from the solution of the Poncare integrals
used to analyse the magnetic field distribution in a finite geometric shape. As such, they are
a function of component shape and relative electrical size.
For cylinders with d w being below 8, a graphical lookup of the p and q values is often
2 2
used. For values of d/>8, the solution approximates to q = , p= , and for
d 1.23 + d
Page 25 of 25
1
0.9
q
0.8
0.7
0.5
0.4
p
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7 8
d/
The induction heating handbook, amongst other books, has the curves in. These curves can
also b found in some of the more advanced electromagnetic field textbooks, although in
some cases, p and q are reversed in meaning.
Example
This example is the same set up as before, except that the steel is at 800oC. The steel will be
past the Curie point, so the current penetration will be much deeper.
First, the relative current penetration depth must be found. The current penetration depth,
given by
2
= ,
has already been found to be 0.0747m or 74.7mm. The diameter of the steel bar is 300mm,
so the steel is not electrically large in diameter.
The ratio of the skin depth to the diameter has to be found: 300/75=4. This is lower than the
simple approximation, so the graphical lookup approach is used. First, we shall find the
equivalent resistance, Rw.
Page 26 of 26
1
0.9
q
0.8
0.7
0.5
0.4
p
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7 8
d/
Using the graphical lookup curve, we can find a value for p=0.38. This can be substituted
into the equivalent resistance formula
N 2 pAw
Rw =
lc
Aw = d w2 =0.0707m2.
4
Remember that N is the number of turns in the work head coil, as the voltage is referred to
the work head coil terminals
Both d/ and Aw are common, but q is different from p. Therefore, for an electrically small
work piece, the reactance is different from the resistance. Inspection of the p and q curves
reveals that the q curve is always higher than the p curve, with the two asymptotically
converging. This means that the work-piece is always more reactive than resistive,
especially with small electrical diameter parts.
Page 27 of 27
Substituting all of the relevant data into the Xw formula
=0.075/(250)=240H
As the frequency and the coil and work-piece dimensions are the same as in the previous
example, then the equivalent circuit values for the coil and the air-gap will be the same.
Therefore the power and the efficiency can be found, again with a 1000Vrms excitation
voltage.
I=V/Z
Z = R2 + X 2 = (Rw + Rc )2 + (X w + X c + X g )2
Using P=I2R, the losses in the work-piece and in the coil can be found:
Pw=421kW
Pc=172kW
The efficiency can be found by dividing the work-head power by the total power
Pw
E= =421/(421+172)=71% - compare with 89% for the semi-infinite slab
Pw + Pc
This reduction in efficiency is due to the airgap, although the efficiency actually peaks when
the work piece is not electrically large, by concentrating the current. The peak efficiency
and therefore the best heating occur around d/=3.5, at the peak in the p-curve. This is very
important it does not occur at d/=2, which is predicted by assuming that all of the current
flows in the first skin depth.
Page 28 of 28
The reason for the lack of power for skin-depths of less than 1/3 of the diameter is a
phenomenon known as flux cancellation, where flux from one side cancels out flux from
the other side, causing the current flow around the inner circuits of the work-piece to be
reduced. This makes the work-piece a net reactive element in the equivalent circuit.
If the equations for the coil and the air-gap are compared to the new equations for the work-
dw
piece, bearing in mind that p can replace as the effective width of the current flow in
2
terms of the losses it can be seen that they all share a number of common terms. They are
all in the same coil, so the number of turns and the coil length are common, and they are
excited at the same frequency. Further, the permeability of free space can be taken out.
This can massively simplify the calculation by pre-calculating a common constant:
2f 0 N c2
K=
lc
R w = K r pAw , X w = K r qAw
1
Rc = X c = Kk r d c c
2
X g = KAg
So far, although we have been calculating values for the equivalent reactance of the work-
head system, we have not really considered its impact, which is to increase the apparent
power (VrmsIrms), without increasing the actual power (I2rmsR), for a given excitation
current. This means that the power factor is reduced by the reactance (p.f.=P/(VI)).
Rtot
Power factor=
Z tot
Example
Find the power factor of the example coil equivalent circuit before and after curie.
Before:
Z tot = Rtot
2
+ X tot
2
Rtot=0.209+0.0216=0.231
Page 29 of 29
Xtot=0.209+0.0216+0.25=0.481
p.f.=Rtot/Ztot=0.231/0.533=0.433
After:
Rtot=0.053+0.022=0.075
p.f.=0.21
So, after curie, 1/5 of the terminal VA of the coil is used in the heating, so the coil would
need five times more VA to heat the work-piece than direct resistance heating would,
despite comparable efficiencies.
As the air-gap adds only to the reactance, the power factor of a real coil is always less than
1/2, and decreases as the air-gap grows. If we examine the resistance and reactance due to
a large cylindrical work piece (or coil), substituting the skin depth into the equations
2 N 2 d N 2 d
R, X = . =
2l l 2
we can see that the resistance and reactance due to the work piece and the coil are
proportional to the square-root of the frequency, whereas examination of the equation for the
reactance due to the air-gap shows that Xg is proportional to the frequency. This suggests
that as the frequency increases, the power factor will fall. It is not uncommon for RF
induction heating coils to have power factors of less than 1/15.
At the time of writing these notes, I was designing a 100kW induction heater to heat chain
from 720oC to over 1200oC in a coil ~1m long with 20 turns. This machine was to harden
chain for use in anchoring for ships. Due to the non-uniform shape of the coil, and its high
temperature, the area of the air-gap between the steel and the coil was considerably larger
than the cross-sectional area of the chain segments. I measured the power factor to be 0.063.
If this coil was to be directly connected to the 100kVA inverter, a power of only 6.3kW
would be developed. This would be an incredibly expensive heating method, in terms of
plant cost (4000/kW).
This poor power factor was corrected by the standard method: A capacitor was inserted into
the circuit to supply negative kVARs to cancel out the positive reactive power provided by
the coil. The capacitor and inductor form a resonant circuit, and if this circuit is excited at
Page 30 of 30
its resonant frequency, the power factor at its terminals will be 1. The actual form of the
circuit is dictated by the nature of the inverter used to supply it. There are two main types of
inverter: voltage-source and current-source. Voltage-source inverters produce a voltage
square-wave across their output terminals and the nature of the coil circuit determines the
current. Current-source inverters try to output a square-wave of current through their output
terminals, and the output circuit determines their output voltage.
If you were to connect a capacitor across the output terminals of a voltage-source inverter,
then the rapid rate of change of voltage with respect to time at the edge of each square-wave
dv
half cycle would cause a large current spike through the capacitor ( I = C ). This spike
dt
would create large losses in the switching devices within the inverter, and the transistors
would probably rapidly fail. Therefore, for a voltage source inverter, the power factor
correction capacitor is often connected in series with the induction-heating coil. The work-
head circuit and the capacitor form a resonant circuit, and tend to filter out all but the
fundamental of the excitation voltage. This means that although the output voltage of the
inverter is a square wave, the output current is a fairly pure sinusoid. The disadvantage of
this connection is that the inverter has to provide all of the coil current, which can be very
high if few turns are used. In this case, a transformer is used to scale the output voltage and
current appropriately.
Q1 SW 1 Q2 SW 4 100
D1 D4
Cfilt 50
Current(A)
0
Tank current
-50
Link current
Q1 SW 2 Q2 SW 3
D2 D3 -100
0 60 120 180 240 300 360
normalised time (deg.)
400
200
Voltage (V)
-200
-400
0 60 120 180 240 300 360
TANK WORK normalised time (deg.)
CAPACITOR HEAD
CT L2+RL
Page 31 of 31
square-wave would cause large voltage spikes across the inductance of the work-head
circuit, which would be carried through the capacitor and the resistance to the inverter
terminals, and would probably rapidly destroy the switching elements in the inverter. A
parallel compensation circuit is therefore used for current-source induction heating inverters.
Once again, this circuit forms a resonant circuit that filters out all but the fundamental of the
excitation waveform, making the voltage across the inverter output terminals a fairly pure
sine-wave.
LSupply D1 D4 1000
Voltage (V)
500
Q SW 1 Q SW 4 0
-500 Tank voltage
Link voltage
-1000
0 60 120 180 240 300 360
D2 D3 Normalised time (deg.)
SW 2 SW 3 100
Q Q
Current (A)
50
0
TANK WORK -50
CAPACITOR HEAD
-100
0 60 120 180 240 300 360
Normalised time (deg.)
In both cases, for MF and RF systems, the reactive elements in the circuit dominate the
resistive. This makes the resultant behaviour of both circuits highly resonant. As the phase
difference between the fundamental of the square-wave quantity (i.e. voltage for a voltage
source) and the sine wave (i.e. current for a voltage source) is zero when the switching
frequency of the inverter is at the resonant frequency of the network, the power factor is at a
maximum.
As the waveforms of the current and the voltage are different, the power factor cannot be 1.
In fact, for a sinusoidal current and a square wave voltage, the in-phase power factor is
4/(2) = 0.9. The square-wave component of the output is often scaled by this value to
give an equivalent sinusoidal voltage or current to make the maximum power factor be equal
to 1 when the current is in phase with the voltage.
Often, the system is run at a slightly inductive power factor to allow the inverter to
commutate correctly, but this is beyond the scope of this course. We will analyse the circuit
at its resonant frequency, as it is much simpler and yields sufficiently accurate results.
Consider the series resonant (voltage source) case. At resonance, the voltage across the
capacitor is equal in magnitude, but 180o out of phase, to the voltage across the total
inductive component of the equivalent work-head circuit. This means that the voltage
Page 32 of 32
across Ccomp and Ltot cancel out, which means that the voltage across the resonant circuit
terminals is equal to the voltage across the equivalent series resistance. Therefore, the
resonant impedance of the circuit is equal to Rtot.
At resonance:
VC=-VL
VR=Vin
VR V
I in = I R = = in
Rtot Rtot
ZT=Rtot
The resonant frequency can be found by equating the reactance of the capacitor with the
negative reactance of the inductance and rearranging. It can be seen that this frequency is
the undamped resonant frequency no matter what the loading is:
from VC=-VL
Solve for
1
0 =
LC
The parallel case is a little more complex. Due to the circuit topology, the voltage across the
capacitor is equal to the voltage across the work-head.
At resonance, the current through the capacitor is equal in magnitude and 180o out pf phase
with the reactive current flowing through the work-head equivalent circuit this means that
the reactive currents will cancel and the phase shift will be zero. Solving this equality for
yields the resonant frequency.
Vin (R jL )
I work =
R 2 + 2 L2
At resonance:
IC=VinjC=-Im(Iwork)
Vin jL
Vin jC =
R 2 + 2 L2
Solve for
Page 33 of 33
L CR 2
0 =
L2 C
If the resonant frequency (in radians s-1) is substituted into the equation for Iwork, then the
terminal current will be equal to the real component of Iwork, as the imaginary component
will be cancelled out by the parallel capacitors current at resonance.
CR
ZT =
L
In the case of the parallel-tuned circuit, the tuning capacitor and the inductance of the work-
head circuit affect the impedance.
The definition of resonant impedance brings us on to the last part of this course: Matching.
Although an inverter will have a nominal power specification, it will also have a maximum
voltage and a maximum current rating, and will only be able to deliver the rated power at the
maximum voltage and current. The maximum inverter voltage divided by the maximum
inverter current gives the inverters nominal output impedance. If the load circuits
impedance is above the inverters nominal impedance then the output current will not be
high enough to give the full power even when the voltage is turned up to the maximum
level.
Alternatively, if the load circuits impedance is lower than the inverters nominal
impedance, then the output current will equal the maximum inverter output current before
the voltage is turned up to full, again limiting the power.
If the output circuits impedance is wildly wrong, then the inverter may be damaged,
although most inverters have limiting circuits to prevent over currents.
Although it is sometimes possible to choose the number of coil turns etc to make the coil
impedance match the inverter, geometrical, mechanical or metallurgical constrains may
force the coil dimensions to be such that the required impedance is not directly attainable.
In this case, a transformer is often used. There are 2 circuit positions to insert the
transformer, the first being between the work-head and the capacitor (i.e. inside the resonant
circuit):
Page 34 of 34
Placing a transformer inside the resonant circuit is useful if the required work-head current
or voltage cant be matched to a standard capacitor. These are often used for single-turn
coils, where currents in excess of 20,000A are required at only 10-100V. MF induction
heating capacitors are generally limited by their terminal current, with 200A/stud being a
common limit, and 8 studs being used to terminate the capacitor. These capacitors often
cost in excess of 1000, and are rated in voltage at 600 to 1200V, depending on capacitance.
Generating the required compensation reactive current by the direct use of such capacitors
would clearly be expensive, as well as technically difficult, so a transformer is often used,
with a ratio of as much as 30:1, to scale the current down and the voltage up, such that the
capacitors and the inverter are better utilised. A transformer inserted between the capacitor
and the work head coil does have to carry the full kVA of the coil, and with power factors
being as low as 0.05, this can be up to 20 times the nominal heating power. This circuit
topology is not very efficient due to the high stress on the transformer, so the transformers
tend to be large and an efficiency of 60% or less can be expected from the resonant circuit
alone.
N 1
Where practical, therefore, a transformer is placed between the inverter terminals and the
resonant circuit. This topology is much more efficient, as the transformer only carries the
heating power, with the reactive power being contained within the resonant circuit.
The ratio of the transformer can be simply dimensioned. If ZT is the terminal impedance of
the resonant circuit, and ZI is the nominal impedance of the inverter, then the ratio can be
found:
ZI
N=
ZT
VI and II represent the voltage and the current at the inverter terminals
Page 35 of 35
VT and IT represent the voltage and current at the resonant circuit terminals
VI
VT =
N
VT V
IT = = I
ZT ZT N
The current into the transformer is the current into the resonant circuit divided by the ratio
IT VI
II = =
N ZT N 2
Re-arraigning
VI
ZT N 2 = = ZI
II
Solving for N
ZI
N=
ZT
So, we have found the equivalent circuit for a work-piece in a multi-turn coil, found out how
to chose a value of compensation capacitor to tune the coil to a specific frequency, and
chosen a transformer ratio to match the coil to the inverter. This is the end of the theoretical
part of the course.
Example:
Chain Heating
Page 36 of 36
Freq 30kHz
Chain Bar 10mm
Link length 40mm
Link width 20mm
Chain Material Mild Steel
Chain (start) 750oC
Chain (end) 950oC
Power to chain 40kW
Chain 1.110-6
Chain 1
d coil 70mm
l coil 1.2m
coil 0.01710-6
N coil 12
Solution:
First find the common constant:
2f 0 N c2
K= =28.42
lc
Assume chain can be represented by k long round bars of same diameter as the wire used to
make the links (k=2 i.e. 2 bars!)
find skin depth in bar
2 2 1.1e 6
= = = 3.0476mm.
2 30e3 4 1e 7
find the ratio diameter: skin-depth
d
= 3. 3
Therefore the graphical look-up table should be used
p=0.38, q=0.67
Resistance of chain = kresistance of bar
D 2
Aw = =7.9e-5m2
4
Page 37 of 37
Rw = K r pAw =28.4210.387.9e-5=0.848m
Rchain=1.69m
Reactance of chain = kreactance of bar
X w = K r qAw =28.4210.677.9e-5=1.50m
Xchain=2.99m
Airgap area= Acoil-k(Abar)
D 2
Aw = =7.9e-5m2
4
D 2
Acoil = =3.845e-3m2
4
Agap=3.69845e-3m2
X g = KA g =104.9m
Capacitor-
1
re-arrange 0 = to find C
LC
1
C=
L 02
L must be found
X tot
Ltot = =(2.99e-3+104.9e-3+1.42e-3)/(230e3)= 0.58H
Page 38 of 38
C=48F
Assume voltage-source inverter
Vcap=Icap1/(2fC)
Icap=Icoil=4.87kA
Vcap=532V
Inverter Matching-
at resonance, Xterm=Rtot=1.69m +1.42m=3.11m
Inverter required impedence=Xreq=V2/P
Xreq=6002/73600=4.9
to find N, use N2=Xterm/Xreq
N2=3.11e-3/4.9=6.35e-4
N=0.025 (or 40:1)
Page 39 of 39