Excerpt from the Proceedings of the COMSOL Conference 2008 Hannover

Calculating Iron Losses Taking into Account Effects of

Manufacturing Processes
P. Goes*1, E.Hoferlin1 and M. De Wulf1
1
OCAS N.V. ArcelorMittal Research Gent
*Corresponding author: John Kennedylaan 3 B9060 Zelzate, Belgium, Patrick.goes@arcelormittal.com

Abstract: Finite element modelling of electrical

machines usually assumes uniform magnetic 2. Energy losses in electrical steels
properties and loss behaviour throughout the
steel laminations. It is well known that 2.1. Total loss Ptot
manufacturing processes, like the assembly of
the machine, deteriorate the properties of the The methods for determining the loss properties
core material. Generally, the measured iron are standardized. To measure core loss the
losses in the machine are higher than the losses material is mounted in an Epstein frame with a
obtained under standard conditions, the ratio primary and a secondary winding around a test
being the so-called building factor. Taking sample. An alternating current i(t) is generated in
advantage of the multiphysics capabilities the primary winding, and the resultant secondary
offered by Comsol Multiphysics, a model was voltage v(t) is measured. The total energy loss
developed with magnetic properties and losses per cycle is then obtained by integrating v(t) i(t)
being dependent on the mechanical stresses over one period T of the alternating current:
T
introduced during assembly of the machine. 1
Ptot = v(t )i (t )dt
Keywords: iron losses, electrical machines, T 0
nonlinear material properties. It can be shown that this is equal to the area
inside the BH-curve, which has units of tesla-
1. Introduction amperes per meter, which is identical to joule per
cubic meter, or the total loss per volume unit.
These days, energy saving is the holy grail of 2.2. Hysteresis loss Physt
engineering. Electrical machines transform
mechanical energy into electrical energy or vice The hysteresis loss is the energy needed to make
versa. Ferromagnetic materials are essential for the material run through a magnetization cycle,
concentrating the magnetic flux into the areas when all induced eddy currents in the material,
where it must generate forces. Electrical steels caused by the dynamics of the exciting field, are
are developed to optimize their properties for use restricted to an absolute minimum. The
in laminated cores of electrical machines. corresponding loss per cycle is thus independent
The ideal electrical steel has: of the frequency f. The loss per time unit can be
High permeability: high B with low H, expressed as:
i.e. low currents and low copper losses. Physt = K h Bmax
2
f
High saturation polarization, for the
same reasons. 2.3. Dynamic loss Pdyn
Low electrical conductivity, for lower
The dynamic loss is the difference of Ptot and
eddy currents in the core.
Physt. This can be further separated into the
Additional non-electromagnetical properties,
classical Foucault loss (eddy current loss) and
such as stampability, are required.
excess loss. The equation for classical Foucault
Manufacturing constraints make this ideal
loss is:
unattainable: trade-offs have to be made. Large
efforts are being made in ArcelorMittal Research Pc = 2 d 2( Bmax f ) 2 / 6
Gent to develop optimal electrical steel grades With the electrical conductivity and d the
for specific applications. To assist customers in lamination thickness
the selection and application of the ArcelorMittal
electrical steels, finite element modelling studies
are conducted as well.
3. Effect of mechanical stresses on the
magnetic properties

The dependence of magnetic properties on

mechanical stress during elastic and plastic
deformation can be evaluated by means of
special constructed equipment, ([Permiakov
2002]) consisting of a single sheet tester in
which unidirectional tensile and compressive
loads can be applied to the specimen.
3.1. Effect on the magnetization loop
In general, the shape of the magnetization loops
is changing during the application of stress, see
Figure 1. The coercive field is increasing while
the remanent induction and the permeability are
decreasing.
Figure 2. Variation of hysteresis loss as a
function of mechanical stress.

4. Application in the modelling of

electrical machines

The stator of electrical machines is often

assembled with the housing by means of press-
or interference fitting. The housing inner
diameter is slightly smaller than the stators
outer diameter. The housing can then be
mounted by heating it, sliding it over the stator,
and allowing it to cool. The shrinking housing
compresses the stator core, and, as mentioned in
the preceding paragraph, these mechanical
stresses deteriorate the magnetic properties of the
electrical steel.
We developed a model in Comsol multiphysics
Figure 1. Variation of magnetization loops under to quantify these effects.
mechanical stress.
5. Solving for the interference fitting
Effect on the hysteresis losses stresses
Figure 2 depicts the hysteresis losses as a
The stresses occurring in the stator were
function of the tensile stress for the
calculated using the plane strain (smpn)
magnetization levels of 0.7, 1.0 and 1.2 Tesla
application mode of the structural mechanics
and 50Hz sinusoidal magnetic flux.
module, by including a thermal expansion load
on the housings subdomain and letting it cool
down from a reference strain temperature down
to the ambient strain temperature.
 2
s=0MPa
s=20MPa
s=40MPa
1.5

B[T]
1

0.5

0
0 500 1000 1500
H[A/m]
Figure 4. B-H curve as function of stress
Figure 3. von mises stress field

1.5

B[T]
1

0.5

0
0
1000
-20
800
600
-40 400
200
S[MPa] -60 0
H[A/m]

As the emqa equation system needs H as a

Figure 4. von mises stress along a radius function of B, the above function has to be
inversed.
The line plot shows compressive stresses inside
the stator and tensile stresses in the housing, as
expected. 10000

This stress state is subsequently used to modify 8000

the magnetic properties of the stator material. 6000
H[A/m]

4000
6. Modelling the magnetic properties
2000

0
For simulation purposes, the magnetization 0

curve of electrical steel can be expressed as a -20

1.5
2

function of just two parameters: the saturation -40

0.5
1

polarization Js and the relative permeability r: S[MPa] -60 0

B[T]

2 Js ( r 1) 0 H
B( H ) = 0 H + arctg This H(B,S) function is implemented in the
2 Js Comsol model as a 2D interpolated table-lookup
The effect of the stress can be introduced by a function.
correction factor:
B ( H , S ) = B ( H ,0) (1 + K S B S ) 7. Solving for the magnetic flux density
With the factor KS = 0.0035 determined by 2D models of electrical machines are well suited
fitting to experimental data, stresses being for the perpendicular induction currents, vector
expressed in MPa. potential (emqa) application mode: the currents
are perpendicular to the modelling plane. This
implies that the magnetic field is present only in 10. Table 1: variables defined for electrical steels
the modelling plane. Hence the magnetic vector
potential has only one nonzero component,
perpendicular to the modelling plane. Maxwells variable Font Description
equations simplify to a second-order scalar PDE
Js 2.1 [T] Saturation
with this magnetic potential component Az as
magnetic
dependent variable.
polarization
The constitutive relation to be used is mur0 8000 Relative
H = f(|B|).eB, with eB the unit vector pointing in permeability at
the direction of the B-field. H=0
The AC/DC modules application mode supports Kh 420 Hysteresis loss
nonlinear relationships between H and B. [W*s*T^- coefficient
Here however, we need H = f(|B|,S).eB, and this 3*m^-3]
requires some modification of the equation Ke 0.96 Excess loss
system: coefficient
d 0.2[mm] Lamination
By default, the expressions for Hx_emqa and thickness
Hy_emqa refer to the HB interpolation function
of the chosen material, e.g. mat1_HB: sigma 3.2e6[S/m] Electric
if(normB_emqa==0,nojac(pdiff(mat1_HB(norm conductivity
B_emqa[1/T])[A/m],normB_emqa))*Bx_emqa,ma
t1_HB(normB_emqa[1/T])[A/m]*Bx_emqa/normB
_emqa)
11. Conclusions
These expressions have to be modified in order
to refer to the 2D interpolation function that Performing these calculations using typical
describes the H = f(B,S), e.g. myHB2D: electromagnetic FEA products requires splitting
if(normB_emqa==0,nojac(pdiff(myHB2D(normB the geometry into discrete affected and non-
_emqa[1/T],mises_smpn[1/Pa])[A/m],normB_e affected zones and assigning different material
mqa))*Bx_emqa,myHB2D(normB_emqa[1/T],mise
s_smpn[1/Pa])[A/m]*Bx_emqa/normB_emqa) properties to each zone. As shown here, Comsol
allows for continuous adaptation of the magnetic
By first solving the smpn-physics for the stress material properties, based on previous geometric
and subsequently solving emqa, one finds the B- and/or structural mechanical results.
field degraded by the effects of the mechanical
stresses.
12. References
8. Postprocessing for the losses
1. G.Bertotti, General properties of power losses
It is then a very simple matter to define the in soft ferromagnetic materials, IEEE Trans. On
above expressions for the loss components as Magn., 24, p.621-630 (1988)
subdomain expressions, to generate loss maps,
and to integrate over complete cycles of the 2. V.Permiakov, A.Pulnikov, L.Dupr, M.De
machine to determine the overall core loss. Wulf, J.Melkebeek, Magnetic properties of Fe-Si
steel depending on compressive and tensile
stresses under sinusoidal and distorted
9. Electrical steels material library
excitations, J.Appl.Phys., 93, p.6689-6691
(2003)
A material library was developed, containing the
ArcelorMittal electrical steels properties.