4.

2
Advanced Thermal Simulation of a Circuit Breaker
Peter U. Frei and Hans O. Weichert
Rockwell Automation AG
5001 Aarau, Switzerland
Email: {pfrei, hweichert}@ra.rockwell.com

Abstract— The quality of thermal simulation results depends

on the accuracy of both the modelled heat sources and heat
sinks. Starting from an existing simplified thermal simulation
model, this paper describes further steps towards a more accurate
representation of the heat sinks in a circuit breaker.
Heat generated in switchgear can leave the system either along
the metallic conducting path to the connecting wires or through
the case to the surrounding air. The former simulation model
accurately represented the heat generation and the thermal path
along the metallic conductor. However, the path through the case
was only summarized by convection coefficients on the metallic
parts’ surfaces.
Due to the high performance requirements of modern breakers
it is interesting to know not only the temperatures of the
conducting parts but also the thermal impact on the plastic
components supporting the conductors or providing essential
functions. Therefore, the non-metallic parts of the circuit breaker
(including the air within the device) are now part of the model.
Consequently, the heat flux through the housing is a result of the
simulation and not an estimated boundary condition anymore.
Moreover, conclusions can be drawn about the thermal stress of Fig. 1. Previous simulation model, consisting of the circuit breaker’s metallic
conducting parts only. Restrictions on the model size did not allow a more
the plastic parts of the breaker. detailed model then.

I. I NTRODUCTION
In the past, the size of switchgear has decreased whereas
A circuit breaker contains several heat sources that cause
the electrical capabilities have not. Accordingly, the heat
its temperature to be higher than the temperature of its
generation per unit volume has increased, causing the analysis
environment. These include (cf. [1])
of the thermal behavior of a device to become a main issue in
• Joule losses in the conducting parts development. Despite this fact, only a few papers deal with the
• Thermal bimetal used for the thermal release and thermal analysis of switchgear during normal operation [1]–
• Contact resistances. [6]. Compared to [2], [4], the model described here features a
As this study is focused on the steady-state thermal analysis more detailed representation of the device’s parts.
of normal operation (as opposed to short circuit condition), In addition, the extension of the simulation tool to practical
the heating effect of the switching arc is not considered here. applications of a group of devices in a panel board is described.
Computer simulation often starts with considerations about Simplifications play an important role in this step.
complexity of the problem and effort required to solve it. The
famous Albert Einstein once stated that “Everything should be II. P REVIOUS THERMAL SIMULATION MODEL
made as simple as possible, but no simpler.” In the following The simulation model presented in [2] and [3] consists of
it is described that this is exactly what is required here. the metallic conducting parts of an Allen-Bradley 140M-C2E-
When we started the thermal simulation of circuit breakers C25 circuit breaker (see fig. 1). Physical properties (thermal
the software package had been limited to 32 000 elements conduction coefficient, electrical resistivity, both temperature
and 32 000 nodes because of cost. This required making some dependent) are assigned to these parts. Electrical and thermal
simplifications in modeling the device which finally have been resistances at the main contact locations are represented by
made responsible for some deviations of the results from the insertion of small additional parts in the conducting path
measurements [2]. having the appropriate electrical and thermal resistance.
In the following, the more detailed thermal simulation of the
internal processes of the same single device presented in [2] A. Electrical resistances
is described, but now the number of nodes available in the In order to find the electrical resistance at the main contacts,
software package is theoretically unlimited. the radius ae of the elastically deformed contact area is

104
0-7803-8460-1/04/$20.00 (C) IEEE
expressed by

  −1
3 3 1 − µ21 1 − µ22 1 1
ae = FK + + (1)
4 E1 E2 r1 r2
where FK is the contact force, Ei and µi are the moduli of
elasticity and the Poisson’s ratios, respectively, and ri are radii
of the spherical members in contact. The radius ae is then used
to calculate the contact resistance
ρ̄
RE = (2)
2ae
according to Holm’s ellipsoid model [7]. The average specific
electrical resistivity of the two different materials in contact
is denoted by ρ̄. The resultant contact resistance then is
2 · 153 µΩ at ambient temperature for the two main contacts
per pole [2].
At the terminals, the electrical contact resistance is calcu-
lated by
Cρ Fig. 2. New, improved simulation model, containing not only most of the
RE = m . (3) metal parts but also most of the plastic parts of the circuit breaker and the air
FK within the housing. This allows determining the influence of the plastic parts
on the heat transportation paths more accurately.
The product Cρ is a constant dependent on the materials in
contact and m is an exponent dependent on the shape of
the contacts [8]. Given the contact force and the materials,
Due to the formerly limited model size, the housing has
the resistances of 120 µΩ (copper–brass) and 35 µΩ (copper–
a very simplified representation in this model. It is modelled
copper) result at the line side and load side terminals, respec-
by the two small boxes visible above the coil (fig. 1). Heat
tively. These resistances are then added to the resistances of
leaving the model by means of convection and radiation on
the wire ends in the terminals.
the housing’s surface is modelled in two steps: The first
B. Thermal resistances step represents the heat transfer from the metallic parts to
The thermal contact resistances are deduced from the elec- the surrounding plastic housing, the second step models the
trical contact resistances according to heat transfer from the housing’s surface to the environment.
Both steps are realized by convection boundary conditions, the
Rel coefficient of convection being based on literature [9], [10],
Rth = . (4)
ρ̄λ̄ measurements and experience.
Similar to ρ̄, λ̄ denotes the average thermal conductivity of The model is limited to a single pole in order to keep the
two different materials in contact. calculation time short.
According to the results of [2], the thermal conductivity of The results of [2] show that the simulated temperature data
the contacts has to be increased by a factor of 2.1 to account of this model matches measured temperature data well.
for the thermal conductivity of the air in the voids of the III. I MPROVED THERMAL SIMULATION MODEL
contact area. The resultant thermal contact resistances amount
to 2 · 10.4 K/W (two main contacts), 7.1 K/W (terminal line Even though the previous simulation model produces accu-
side) and 2.4 K/W (terminal load side). rate results for the current carrying parts, there are some points
that should be improved:
C. Thermo-electrical coupling • It is important to know the temperature distribution not
An applied current flowing in the electrically conducting only in the conducting parts, but in the plastic parts as
path causes Joule heat P = RI 2 to be generated. This heat well. This would help in the optimal material selection
increases the temperature of the conducting parts. Due to for the plastic parts.
the resulting temperature differences, heat is transported by • Due to the formerly limited number of nodes and el-
means of conduction, convection and radiation. Conduction ements available, the simplified representation of the
in the metallic parts is represented by assigning these parts housing in the previous model required the calibration
the appropriate thermal conductivity value in the simulation of the heat transport through the housing using results of
model. A temperature-dependent thermal resistance links the measurements. Design changes on the housing demand
conducting path at its boundary (terminals) to the environment, for the re-calibration of the model and thereby for new
representing the cooling effect of the connecting wires. The measurements. It is desirable to have a model that is
characteristics of the thermal resistance are determined with a independent of calibration measurements in order to get
separate simulation model. simulation results quickly.

105
 CV R R

R
CD2
CD1
R

CD3 Fig. 4. A typical arrangement of switchgear in a control cabinet. From left

CV to right: Starter combinations consisting of a circuit breaker and one, two and
three contactors.

R and air (0.025 . . . 0.035 Wm−1 K−1 ).

CD1
Comparative measurements of the contact resistance have
shown that the actual values are somewhat higher than cal-
culated theoretically for elastic deformation (2). This can be
R understood with the observation that already a single opera-
tion (calibration shot) can cause up to twice the calculated
CV R
resistance. Graphite particles are released from the metallic
R
compound and contribute an additional film resistance.

A. Model setup
Fig. 3. Thermal effects included in the model: conduction through metallic
parts (CD1 ), conduction through plastic parts (CD2 ), conduction through the
The thermal effects included in the model are shown in
air inside the circuit breaker (CD3 ), convection on the outer surfaces (CV) fig. 3. Barcikowski [4], [5] performed a comprehensive analy-
and radiation on the outer surfaces (R). sis on heat transportation effects in switchgear. For switchgear
of the dimensions discussed here, he came to the conclusion
that both convection and radiation inside the device has a
• Circuit breakers are arranged either side-by-side in con-
negligible effect on the resulting temperature distribution. In
trol cabinets or standalone. This has a noticeable impact
the interior part of the model, it is thus sufficient to consider
on the temperature distribution inside the breaker as
thermal conduction only. This is a significant reduction of the
the cooling conditions are totally different in the two
model’s complexity. Thermal conduction, however, has to be
cases. The previous simulation was optimized for the
considered in all materials: metals, plastics and air.
side-by-side arrangement. For some applications, it is
useful to have the possibility to simulate the standalone B. Boundary conditions
arrangement, too.
At the model boundary (the surface of the circuit breaker’s
In order to get the temperature distribution in the plastic
housing), boundary conditions represent the convection and
parts, they have to be included in the model. As a consequence,
radiation heat transfer to the environment indicated in fig. 3.
the boundary of the model was moved from the surface of
1) Convection boundary conditions: In theory, the convec-
the metallic parts to the surface of the breaker’s housing.
tion heat transfer rate is expressed by
Everything inside the housing is now part of the model. This
also includes the air in the void spaces inside the circuit q̇ = αA∆T (5)
breaker. Fig. 2 shows a cut though a single pole of the extended
model. where α denotes the convection coefficient, A stands for the
All main parts of the circuit breaker are modelled hav- area and ∆T is the temperature difference between the surface
ing their actual shape material. However, the mechanism and the environment. In the model it is assumed that the
of the device, consisting of more than two dozen small temperature of the environment is constant and given. In the
parts each having a complicated shape, was simplified by finite element model of the circuit breaker, the parameter α
a single volume in order to keep the modelling effort rea- has to be specified for the convection boundary condition.
sonable. In reality, the mechanism’s are made of metals The value of the convection coefficient α depends on the
and plastics and it also contains a fair amount of air. characteristics of the fluid motion around the circuit breaker.
In the model, its thermal conductivity was given an aver- Often, circuit breakers are arranged side by side in control
age value of 0.1 Wm−1 K−1 . This value is the estimated cabinets as shown in fig. 4. If many of them are arranged this
weighted average of the thermal conductivities of metal way, this can be interpreted as convection around a long bar
(12 . . . 386 Wm−1 K−1 ), plastics (0.09 . . . 0.33 Wm−1 K−1 ) as shown in fig. 5. The coefficient α of a long bar can be

106
 of 1.3 · 10−5 . . . 2.9 · 10−5 m2 s−1 in the relevant temperature
range).
Substituting all these numbers into the formulae, the convec-
tion coefficient α (a function of both the surface temperature
and the environment temperature) takes values in the range of
0.093 . . . 7 Wm−2 K−1 .
2) Radiation boundary conditions: The radiation heat
transfer from the circuit breaker’s surface to the environment
is given by
q̇ = σA(TS4 − TE4 ). (10)

The emissivity , the area A, the Stefan-Boltzmann constant

σ = 5.67 · 10−8 Wm−2 K−4 and the temperatures of the
surface (TS ) and the environment (TE ) determine the value of
the radiation heat transfer rate q̇. In the finite element model,
the temperature of the environment and the emissivity of the
surface have to be given. Based on data of similar plastics [12],
[13], it is assumed that the emissivity of this circuit breaker’s
housing is  = 0.95.
3) Connecting wires: As indicated in fig. 3, part of the heat
produced in the device is dissipated by the connecting wires.
As these are not part of the model (cf. fig. 2), their effect is
modelled by boundary conditions, too. A separate model that
Fig. 5. Convection across a row of devices (front) can be interpreted as contains an insulated copper wire is used to find the amount
convection across a long horizontal bar (back). Figure taken from [11]. of heat dissipated by convection and radiation on the wire
surface.
The convection coefficient α of the insulation surface in
expressed in terms of its dimensions and the fluid parameters
the separate wire model are calculated according to [14,
and is relatively easy to calculate.
Section 4.2]:
For free convection, [9] states that Nuλ
Nul λ α= . (11)
α= (6) dI
l
In the above formula, dI denotes the outer diameter of the
where Nul is the Nusselt number, λ is the thermal conductivity wire’s insulation and λ is again the thermal conductivity of
of the fluid and l denotes the length that a fluid particle can the air. For the calculation of the Nusselt number Nu, the
travel along the surface. In our case, l = 0.157 m and λ varies following approximation is given in [14]:
between 0.025 Wm−1 K−1 and 0.035 Wm−1 K−1 in the rel-
evant temperature range. Due to the temperature dependency Nu = 0.8(Ra0.05 ) + 0.35(Ra0.27 ) (12)
of many of the parameters included, the convection coefficient
α is a function of the temperature, as well. with the Rayleigh number
The Nusselt number is defined as Ra = 3.912 · 1019 (273.15 + TE )−4.69 d3I (TS − TE ). (13)
 2
Nul = 0.752 + 0.387(Ral · f3 (P r))1/6 (7) As before, TS and TE are the temperatures of the surface and
  
9/16 −16/9 the environment, respectively.
0.559
f3 (P r) = 1+ (8) In the end, the wire model returns the characteristics of
Pr two thermal resistances representing the heat dissipated by
with the Rayleigh number being the product Ral = Grl P r. convection and radiation, respectively, on the wire’s surface.
The Prandtl number P r is a material property of the fluid, the These nonlinear thermal resistor properties are then used in
Grashof number Grl can be expressed as the circuit breaker model to account for the connecting wires.
4) Periodic boundary conditions: The case of side-by-side
gl3 β∆θ
Grl = . (9) arrangement of the circuit breakers can be interpreted as an
ν2 infinitely repeated linear arrangement of a single pole. It is
Here, g = 9.81 ms−2 is the gravitational acceleration, β = thus sufficient to simulate a single pole only and represent its
1
T is the volume coefficient of expansion (assuming that neighbouring poles with periodic boundary conditions. This
air behaves like an ideal gas), ∆θ denotes the temperature is realized by coupling the temperature degree of freedom of
difference between the surface and the environment and ν every node on one side of a pole to the corresponding node
is the kinematic viscosity of air (taking values in the range on the other side of the pole.

107
 140M-C2E-C25, Ie=25A, 4mm2

15%

39%
140
conduction rad. from wire surface: 0.46W
total joule heat:
3.10W (100%)
to wires: 130
convect. from wire surface: 1.22W
1.68W (54%)
120
conduction to convect. from housing surface: 0.62W

Temperature rise ∆T [K]

housing surf.: 110
1.42W (46%) rad. from housing surface: 0.80W
100

20%
26%
90
Fig. 6. Simulated distribution of the heat dissipation per pole from the 80 Lab measurement
(b)
circuit breaker to the environment. Abbreviations: rad.: radiation, convect.:
convection, surf.: surface. Simulation
70
(previous model) (c)
60 (a)
Alternatively, the standalone arrangement can also be sim- Simulation
ulated. To do so, all three poles of the circuit breaker are 50 (current model)
modelled (unfortunately increasing the model size by a factor 40

terminal L2
line side

coil in

coil center

coil out

bimetal base

bimetal tip

stationary contact
line side

contact bridge

stationary contact
load side
terminal T2
load side
of 3) and convection/radiation boundary conditions are defined
on all of its surfaces, including the two lateral surfaces.
C. Solution type
Once a circuit breaker is turned on, its temperature rises
from the initial ambient temperature to the final steady state
temperature. Obviously, the highest temperatures occur after Fig. 7. Comparison of measurement and simulation results (previous and
the transient phase is over and the steady state (thermal improved simulation model).
equilibrium) has been reached. To find the most critical state
of the device, it is thus sufficient to calculate the steady state
solution. Lab measurements show that the temperature rise a short summary on the electrical and thermal contact re-
•

during the transient phase is approximately exponential with sistances at the main contacts and the terminals resulting
a time constant of τ ≈ 15 min. from the given parameters.
Note that the same model can be used for transient cal- • the name and the revision number of the input file for

culations, as well. However, this requires additional material an unambiguous mapping of the parameter values to the
properties (density and specific heat) to be known for all the calculated results
materials used. Reliable specific heat values are not always By summing up the electrical power losses and comparing
readily available, especially for plastic materials. them to the heat dissipated to the environment, a simple check
of the correctness of the result can be performed.
IV. R ESULTS Fig. 6 shows the heat dissipation results of the simulation
The solution of the finite element model described above of the above circuit breaker. Almost two thirds of the total
produces a huge amount of data. Both electrical (current heat generated (3.1 W per pole) is dissipated by convection,
density, voltage) and thermal (temperature, heat flux density) the rest by radiation. Slightly more than half of the heat is
results are available for every single element of the model. dissipated by the connecting wires whereas the remainder is
With a model size of 266 446 elements and 46 610 nodes (one dissipated by the housing’s surface. Barcikowski [4] found in
single pole only) some postprocessing is required to extract his simulation of a 100A circuit breaker that 42% of the heat is
the key results. dissipated at the connecting wires versus 58% at the housing.
The postprocessing procedure automatically generates a Fig. 7 shows the graphical representation of the temperature
report containing distribution along the current path (cf. fig. 1) from the line side
• a list of the Joule heat generation in the individual to the load side. Of course, the temperature distribution can
sections of the current path be evaluated at any point desired, as is indicated by fig. 8.
• a list of the heat dissipation to the environment by means Compared to the results of the previous simulation model,
of convection and radiation (for both the housing and the the improved model shows a better agreement with mea-
connecting wires). surement data at the line side terminal (a). This is due to
• a list of the temperatures at the same locations at which the fact that the previous model could not account for the
our lab staff usually performs measurements spatial proximity of the line side terminal and the hot bimetal
• the electrical voltage at these points (cf. fig. 1). The improved model now includes the heat flux
• the electrical resistance of the individual parts of the from the bimetal to the terminal through the air and thus
current path in the heated state yields a higher (more accurate) temperature at the line side

108
 electrical
I Rel

thermal
Pth=Rel I2
1/2R 1/2R
th1 th1

Rth2

Fig. 9. The network representation of three segments of the current carrying

path of a circuit breaker.

model as it just becomes too large and complex to handle. So

the characteristics of its constituting parts (circuit breaker and
Fig. 8. Simulated temperature distribution in the steady state. Temperatures
contactors) have to be determined individually and the overall
are expressed in degrees Centigrade. model has to be one that integrates their key properties only.
One way of accomplishing this is to interpret both the
circuit breaker and the contactor as a network of electrical
terminal. Accordingly, the temperatures at the bimetal are and thermal resistances that correspond with each other. The
lower compared to the previous model. current carrying path of the circuit breaker is naturally divided
The temperature of the contact bridge, formerly also a part into different segments, namely the individual metallic parts.
with differences between measurement and simulation, is now From the electrical point of view, every one of these segments
modelled with better agreement, too (b). is a resistor. All of them are arranged in series. The top part
Unfortunately, the improved model yields a temperature at of fig. 9 shows one segment and its two adjacent ones.
the load side terminal that is significantly higher than the The electrical network notation can also be used for the
measurement (c). This may be due to the approximation of thermal part of the problem. The electrical terms “current”,
the coefficient of convection based on the simple bar geometry “voltage” and “electrical resistance” correspond to the ther-
shown in fig. 5. In reality, the actual value of α(TS ) depends mal items “heat flux (power)”, “temperature” and “thermal
not only on the temperature of the surface but also on the resistance,” respectively. Ohm’s law U = Rel I for electrical
location on the circuit breaker’s surface [4]. As the cool air is circuits also has its direct thermal counterpart T = Rth Pth .
approaching the device from below, the cooling effect at the
The Joule heat generation Pth = Rel I 2 couples the
bottom of the circuit breaker (load side) is larger than at its
electrical (top) and thermal (bottom) part of the network
top (line side), where the air has already collected some heat.
representation. The thermal flux Pth generated by the Joule
This effect is not considered in the model, possibly causing
heat is denoted by the heat flux source symbol in the thermal
the modelled temperature to be too high at the bottom of the
part. This heat may leave the system either along the current
circuit breaker. A detailed calculation of the air flow across
carrying path (thermal resistance Rth1 ) or through the housing
the housing using computational fluid dynamics is expected to
(thermal resistance Rth2 ). Neglecting the fact that the Joule
give more insight here.
heat generation is distributed along the whole length of the
Overall, the improved model is better adaptable to design
segment, it is modelled entirely in the middle of the segment.
changes and even shows a better agreement with measured
Accordingly, the thermal resistance of the segment’s current
data.
carrying path Rth1 is divided in two.
V. O UTLOOK Connecting these segments in series, a simple network
The results of the improved model are promising, but the model of the whole circuit breaker can be built. The finite
model has one downside: its increased complexity makes it element simulation model described above is necessary to
more difficult and time-consuming to manage and evaluate. obtain the thermal resistances Rth2 as they are very difficult
One would like to finally have the advantages of the new three- to find otherwise.
dimensional model in combination with a less complex model. Once these network models for both circuit breaker and
Real applications often consist of starter combinations that contactor are known, they can be connected to form the
are usually composed of a circuit breaker and up to three thermal model of a whole starter combination as desired. Addi-
contactors (cf. fig. 4). One of our goals is having an accurate tional elements of interest (connecting parts, increased contact
thermal model that allows us to simulate these real applica- resistance due to contact wear, etc.) could be introduced easily
tions. There is no way of doing this with a single finite element into this form of the model.

109
 VI. C ONCLUSION R EFERENCES
• The result of the simulation of the single device confirms [1] H. Stammberger, C. Dehning, M. Anheuser, and A. Zacharias, “Tem-
that the problem has to be described in more detail right peraturberechnungen für Schaltgeräte,” in Kontaktverhalten und Schalten
(16. Fachtagung Albert-Keil-Kontaktseminar), Karlsruhe, Sept. 2001, pp.
from the beginning to get more accurate results and to 195–204.
illustrate the physical dependencies within the device for [2] J. Paulke, P. Steinhaeuser, and H. Weichert, “Simulation of contact
better understanding. spots,” in Proceedings of the 21st International Conference on Electrical
Contacts, Zürich, 2002, pp. 388–393.
• Once the device is well understood, simplification can be [3] J. Paulke, P. Steinhaeuser, and H. Weichert, “Thermal simulation of
made to enable the calculation of a group of devices in ar- switchgear,” in Proceedings of the 47th Holm Conference, Montreal,
rangements of practical applications within the generally 2001, pp. 6–11.
[4] F. Barcikowski, “Numerische Berechnungen zur Wärme- und Antriebs-
limited resources. auslegung von Schaltgeräten,” Ph.D. dissertation, Technische Universität
• One result of the detailed three-dimensional simulation Braunschweig, Göttingen: Cuvillier, 2003.
is the possibility to describe the device by certain char- [5] F. Barcikowski and M. Lindmayer, “Simulations of the heat balance in
low-voltage switchgear,” in Proc. 20th Int. Conf. on Electrical Contacts,
acteristic numbers. Further investigation of changed pa- Stockholm, June 2000.
rameters may be made e.g. with simpler and more cost [6] F. Völker, “Computer-aided calculation of the steady-state thermal
effective tools like network analysis. performance characteristics of low-voltage circuit breakers,” European
Transactions on Electrical Power Engineering (ETEP), vol. 3, no. 5, pp.
This step frees the expensive three-dimensional tool and 347–352, 1993.
it may again advantageously be applied to new problems [7] R. Holm, Electric Contacts – Theory and Application, 4th ed. Berlin:
where its complexity is needed. Springer, 1967, third printing 2000.
[8] M. Minovic, Schaltgeräte: Theorie und Praxis. München: Hüthig und
• The more exact data of the materials used in the sim- Pflaum, 1977.
ulation are known, the better the results comply with [9] Verein Deutscher Ingenieure, Ed., VDI-Wärmeatlas: Berechnungsblätter
the real device (this is a truism). Manufacturers of any für den Wärmeübergang, 8th ed. Berlin: Springer, 1997.
[10] J. P. Holman, Heat Transfer, 9th ed. New York: McGraw-Hill, 2002.
material are encouraged to take care of their data sheets [11] P. U. Frei and H. O. Weichert, “Erwärmungssimulation des Magnet-
and provide reliable data. antriebs eines Schützes,” in 8. Schweizer CAD-FEM User’s Meeting,
Zürich, 2003.
ACKNOWLEDGMENT [12] H. Böhme, Mittelspannungstechnik. Berlin: Verlag Technik, 1992.
[13] G. G. Gubareff, J. E. Janssen, and R. H. Torborg, Eds., Thermal
The authors would like to thank Peter Steinhaeuser (VT- Radiation Properties Survey, 2nd ed. Minneapolis: Honeywell Research
GmbH, Schaffhausen, Switzerland) for his work on the de- Center, 1960.
velopment on the previous simulation model that served as a [14] “IEC 60943 – Guidance concerning the permissible temperature rise for
parts of electrical equipment, in particular for terminals,” Geneva, Jan.
basis for the model described here. 1998.

110