The Second Iranian Conference on

Engineering Electromagnetics
1928
K N Toosi University of Technology
(ICEEM 2014),
Faculty of Electrical Engineering
Center of Excellence in Computation
Jan. 8-9, 2014
and Characterization of Devices and
Subsystems

A Comparison between Two Dimensional Finite

Downloaded from isseem.ir at 22:10 +0430 on Monday August 6th 2018

Element Method Analysis of Eddy Currents and

Analytical Results
Vahid Badeli*, Esmael Fallah

Department of Electrical Engineering, University of Guilan, Rasht, Iran

*
Corresponding author: Badeli.vahid@yahoo.com

ABSTRACT—Eddy currents are one of the main
loss components in the core of electrical
machines. While calculating eddy current using
conventional methods, flux density is assumed
to be distributed uniformly. Considering a
uniform flux distribution is not a valid
assumption. According to Lenz’s law, eddy
currents oppose to their causes and
consequently by producing flux skin effect,
cause a non-uniform distribution of flux
density. In this paper a new finite element
method has been proposed to analyze the effect
of eddy currents in a more accurate manner in
which the diffusion equation is solved in a
circular cross section and results in different
frequencies are compared.
KEYWORDS: eddy currents, finite elements,
skin effect, two dimensional.
I. INTRODUCTION
According to Faraday’s law, when
conductors are placed within a time-variant
magnetic field, an electro-motive-force is
induced in them that make the electrical
current flow in circular paths. These currents,
which are called eddy currents, are the origin
of the Ohmic loss in the magnetic cores. In
conventional method of calculating eddy
current loss, flux density is assumed to be
distributed uniformly on the core cross section
area and by this assumption, following, well
known formula is resulted for calculating eddy
current loss [1, 2]:
π 2 σ d2 2 2
Pe = f B max (1)
6
In which, Pe is the eddy currents loss in the
unit volume, σ is special electrical
conductivity, d is the thickness of the
lamination, f is the frequency of the magnetic
flux variation and Bmaxis the amplitude of the
magnetic flux density. Due to eddy currents,
assuming a uniform distribution of flux is not
accurate because according to Lenz’s law,
eddy currents oppose to their causes and force
the flux to flow along the skin of the
conductor. This effect is known as flux skin
effect [3]. In low frequencies this effect is
significantly weakened by laminating the core.
But if the flux variation frequency increases,
for instance in a case with a voltage
containing harmonics which results into the
flux having harmonics as well, the flux skin
effect could not be neglected [4]. In such cases
estimating eddy current losses that were
calculated using Eq. (1) is not exact and in
order to have a more accurate analysis, the
effect of eddy current shall be taken into
account. On the other hand, modeling eddy

611
This paper is authentic if it can be found in www.isseem.ir.
 currents play a main role in detecting cracks in following:
metal equipment. In this method a time variant
magnetic field isapplied to the metal ∂D (3)
∇× H = J +
equipment and by observing the eddy current ∂t
patterns, possible cracks are located [5].
Inorder to achieve the flux distribution in ∂D
In which could be negligible at Power
presence of eddy currents, the diffusion ∂t
equation is to be solved at the cross-section of frequencies. So this equation could be written
Downloaded from isseem.ir at 22:10 +0430 on Monday August 6th 2018

the core [6]-[8]. In this paper a new finite as:

element method has been proposed to analyze
the effect of eddy currents in a more accurate ∇× H = J (4)
manner in which the diffusion equation is
solved at a circular cross section and results in Ohm’s Law can be described as follows:
low frequencies are compared with results in
high frequencies.A comparison between J =σE (5)
results reached by the new method and the
results reached by analytical solution is done Faraday’s Law is describes as follows:
and it shows the accuracy of the new method.
∂B (6)
∇× E = −
∂t
II. FIELD EQUATION IN PRESENCE OF
EDDY CURRENTS Where,H is the magnetic field intensity,Jis
the electrical current density,E is the
A. Defining the observed problem electrical field intensity andB is the magnetic
For simplicity, a winding toroid with circular flux density. By Combining Eq. (4) to Eq. (6),
cross sections examined. The winding is the equation describing the field in presence of
connected to a time-variable voltage source eddy current will be as follows:
(for example sine function). The boundary
field intensity is assumed to be as follows: ∂B (7)
∇ × ∇ × H = −σ
∂t
H o (t ) = H max sin(ωt ) (2)
In Cartesian coordinate system, Eq. (7) can be
In which, ω isthe angular frequency of the field described as:
variation. Induction of eddy currents in the
∂2 H z ∂2 H z ∂B z (8)
spots within the conductor core, which are + =σ
2 2
affected by the time-varying magnetic field, ∂x ∂y ∂t
changes the field within the conductor. From
the Ampère’s Law, it can simply be inferred Where Hz and Bz are the z components of
that the eddy currents only influence the inner theH andB.
parts of the conductor and do not affect outer
spots of thecore. As a result,Ho can be used as By considering Eq. (9) the diffusion equation
a boundary value for the eddy current problem. is like Eq. (10).
By considering the existing symmetries, the
B = µH (9)
under investigation space is limited to cross
section of a core.
∂2 H z ∂2 H z ∂H z
2
+ 2
= µσ
∂x ∂y ∂t
B. Equations correspond to eddy currents (10)
The discrete form of Ampere’s law is as

612
 C. Analysis of magnetic field in presence of B ( x, y ) = a B + bB x + c B y
eddy current at a circular cross section 3
(18)
= ∑ N i ( x, y )Bi
By considering cylindrical coordinate system, i=1
the diffusion equation would be like Eq.(11).

∂2 H z 1 ∂H z ∂H z (11)
2
+ = µσ
∂r r ∂r ∂t
Downloaded from isseem.ir at 22:10 +0430 on Monday August 6th 2018

If the variation of Hz in time domain

considered as a sine function, Hz can be
represented as follows:

H z ( x, t ) = Re[ H z ( x )e jωt ] (12) Fig. 1 First order triangle element

So Eq. (11) could be written as below:

Where, Hiand Bi are respectively field intensity
∂2 H z 1 ∂H z and flux density of the vertices. Ni(x,y) is the
+ − jωµσ H z = 0 (13)
∂r 2 r ∂r first order shape function as follows:

The answer of Eq. (13) is: ai + bi x + ci y

Ni ( x, y) = (19)
2D
H z ( r ) = a0 Bessel0 ( − jωµσ r ) (14)
In Eq. (19) the coefficients a1, b1, c1 and D are
Where, Bessel0 is the Bessel function of the calculated using the following formulas,
first kind, with the order of zero. Now
boundary condition must be considered: a1 = x 2 y 3 − x 3 y 2 (20)

H z ( R) = H 0 (15) b1 = y 2 − y 3 (21)

At last, by considering the boundary condition, c1 = x3 − x2 (22)

the answer would be like Eq. (16).
ai + bi xi + ci yi
D= (23)
H 0 Bessel0 ( − jωµσ r ) 2
H z (r ) = (16)
Bessel0 ( − jωµσ R)
The rest of the Coefficients (for i = 2, 3)
reached by cyclic transformation. By applying
D. Two dimensional finite element method to the Galerkin method on Eq. (10) and using the
solve the diffusion equation discreteform of the time derivative phrase, the
By approximating B and H, as first order following statement can be written [9]:
functions of x and y, on a triangular element
∂2H ∂2H
like Fig.1, It can be written as: ∫∫ N i (
S ∂x 2
+
∂y 2
) dxdy
(24)
H ( x, y ) = a H + bH x + c H y σ
= ∫∫ N i (B (t ) − B (t − ∆t )) dr
3
(17) S
∆t
= ∑ N i ( x, y )H i
i=1
After simplifying above equation, the
following algebraic equation is reached which
works for an element and is called local
equation:

613
  b1 b1 + c1 c1 b1 b2 + c 1c2 b1 b3 + c1 c3  H1(t) 
1 
b2 b1 + c2 c1 b2 b2 + c2 c2 b2 b3 + c2 c3 H2 (t)
4D 
b3 b1 + c3 c1 b3 b2 + c3 c2 b3 b3 + c3 c3 H3 (t)
2 1 1 B1(t) B1(t − ∆t)  (25)
σD 
=− 1 2 1B2 (t) − B2 (t − ∆t)
12 ∆t 
1 1 2B3 (t) B3 (t − ∆t)
Downloaded from isseem.ir at 22:10 +0430 on Monday August 6th 2018

By combining local equations corresponding to

all of the elements, the following general
equation is resulted:

[S ][H (t )] = −[G][B(t )] + [G][B(t − ∆t )] (26)

By using Eq. (9) we have:

([SS])[H (t )] = µ[G][H (t − ∆t )] (27)

In which:

[SS] = [S ] + µ[G] (28)

That[S] and [G] are attained using special

methods of adding up the matrixes of Fig. 2 Flowchart of solution process
coefficients in Eq. (25) for all the elements [9].

According to the above equation, the

procedure of solving the field equation can III. CALCULATION OF EDDY
carry out. So in every instant of t, the value of CURRENTS AND THEIR LOSSES
[H(t-∆t)] is known sothe right side of the Eq. For calculating eddy currents and losses that
(27) will be known and the equation can be they cause at the core cross section, we could
solved to yield the [H(t)]. use of Eq. (4). To do so, at first, the magnetic
field intensity of each element is like Eq. (29).
It should be mentioned that in every instant of a1 + b1 x + c1 y a + b x + c2 y
t, the Eq. (27) is solved by applying the H(x, y) = H1 + 2 2 H2 +
2D 2D
boundary conditions. (29)
a3 + b3 x + c3 y
H3
2D
Fig.2 shows flowchart of the solution process. Now by Eq. (4), x and y components of eddy
currents are respectively like Eq. (30) and Eq.
(31).
 H 1 (t ) 
∂H ( x, y) b3 
] H 2 (t )
b1 b2
Jy =− = [− − − (30)
∂x 2D 2D 2D 
 H 3 (t )
 H1 (t ) 
∂H ( x, y) c3 
] H 2 (t )
c1 c2
Jx = − = [− − − (31)
∂y 2D 2D 2D 
 H3 (t )
Now to calculate the Eddy currents cross
through each element, we could use ofEq. (32).

614
 J = ( J x2 + J y2 ) (32)
To calculate the eddy currents losses of each
element we could use of Eq. (33).
J2 D
Pe = (33)
σ
In which J is the eddy currents crossing every
element, Dcomes from Eq. (23) and σ is
Downloaded from isseem.ir at 22:10 +0430 on Monday August 6th 2018

special conductance of the core.

At the end, to calculate losses of all elements
of the circular cross section, we should sum
losses of each elementof cross section of the
core, reached by Eq. (33),together.

IV. RESULTS OF ANALYZING A PRACTICAL

SYSTEM
Fig. 3 Dimensions and meshes of a typical thin sheet.
The method described before, used for atypical
system which its parameters have been shown
in Table 1.

Table 1 parameters of typical system

Boundary magnetic field
Ho=100sin(2πfot)
intensity
Permeability of the core µr=1000
Radius of the circular
R=1m
crosssection area of the core
σ =2.3×106m/s
Special conductance of core

The meshes and dimensions of the cross

section areaare shown in Fig. 3.
Boundary elements are shown by yellow and
boundary nodes are shown by red circles.
A gray path as could have been seen in Fig. 3
is shown.
The magnetic field distribution at cross section Fig. 4 Magnetic field distribution at cross section of the
core at fo=2150hz
area, at fo=2150Hz, is shown in Fig. 4.
The skin effect is obvious in this figure.
For elements with the gray color, a comparison
of magnetic field intensity obtained by the
numerical and the analytical method is shown
in Fig. 5. This is done for two different
frequencies.
A comparison between eddy currents density
reached by the numerical and analyticalmethod
is shown in Fig. 6. Figures show the acceptable
accuracy of the proposed method.

(a). fo= 50Hz

615
Downloaded from isseem.ir at 22:10 +0430 on Monday August 6th 2018

(b). fo=2150Hz (b). fo=2150Hz

Fig.5 Comparison ofthe magnetic field intensity, Fig.6 Comparison of Eddy current density, reached by
reached by 2D FEM and analytical method, for two 2D FEM and analytical method, for two different
different frequencies. frequencies.

The eddy loss reached by the numerical and It is obvious in this table that in low
analytical method is shown in Table 2. frequencies, eddy current loss is much lower
than high frequencies. This is in agreement
Table 2 eddy loss reached by the numerical and with the results represented in [10]-[11].
analytical method By former explanation, the difference between
Frequency Eddy current loss (w/m3) eddy currentlosses reached by two different
(Hz) 2D FEM Analytical methods at higher frequencies is expected.
50 189.3033 193.3857 This divergence is because of derivative inEq.
100 702.4913 734.2284 (30) and Eq. (31).
150 1.4000×103 1.5237×103 Also it should be mention that the calculated
550 6.5590×103 7.8640×103 eddy current density for boundary elements by
4
1750 1.1367×10 1.5928×104 this method is not accurate because the
2150 1.2063×104 1.7935×104 divergence of derivatives in that area is not
negligible. So the calculated results for these
elements are neglected.
As a solution to these problems, the eddy
current density can be reached directly from
the diffusion equation.By using this method
the error of the derivation would be
eliminated.
This subject is the idea of the future works for
authors.

V. CONCLUSION
(a).fo= 50Hz
A new finite element method is proposed in
this paper to solve the two dimensional
diffusion equation. The analysis is done for a
core with a circular cross section. For different
frequencies,the results reached by the new
methodare comparedwith the results of
analytical solution, which show the accuracy
of the new method.

616
 REFERENCES
[1] G. R. Slemon, A. Straughen, Electric
Machines, Addison-Wesley, 1980.
[2] I. D. Mayergoyz and C. serpico, “Eddy-
Current Losses in magnetic Conductors
with Abrupt Magnetic Transitions,”
IEEE Trans. Magnetics, VOL. 36,
pp.1962–1969, July 2000.
Downloaded from isseem.ir at 22:10 +0430 on Monday August 6th 2018
[3] R. jafari-shapoorabadi, A. Konard , and
A. N. Sinclair, "Comparison of Three
formulations for Eddy-current and Skin
effect Problems," IEEE Trans.
Magnetics,VOL. 38, pp. 617-620,
March 2002.
[4] H.G. Brachtendorf and R. Laur,
"Simulation of Skin Effects and
Hysteresis Phenomrna in the Time
Domain," IEEE Trans. Magnetics,
VOL.37,pp. 3781 – 3789, September
2001.
[5] G. Yang, A. Tamburrino, L. Udpa, S.S.
Udpa, Z. Zeng, Y. Deng, P. Que,
“Pulsed Eddy-Current Based Giant
Magneto Resistive System for the
Inspection of Aircraft Structures,”
IEEE Trans. Magnetics, VOL. 46, pp.
910-917, March 2010.
[6] S. R. Naidu, “Time domain model for
the ferromagnetic core, including the
617
effects of hysteresis and eddy currents,”
IEEE Proc., VOL. 138, pp. 44-50, Jan.
1991.
[7] Bastos A, Sadowski N. Electromagnetic
Modeling by Finite Eement Methods.
Brasil: Marcel-Dekker, 2003.
[8] L.Rouve, F.Ossart, T.Waeckerle,
A.K.Lebouc, “Magnetic flux and losses
computation in electrical laminations,”
IEEE Trans. Magnetics., VOL. 32, pp.
4219-4221, Sept 1996.
[9] Kurt Preis, OszkárBíró, “FEM Analysis
of Eddy Current Losses in Nonlinear
Laminated Iron Cores,” IEEE Trans.
Magnetics, VOL. 41, pp 1412-1415,
May 2005.
[10] Biorci G, Pasquale M. “Physical
interpretation of Induction and
Frequency Dependence of power
Losses in Soft Magnetic Materials”.
IEEE Trans. Magnetics VOL. 28, pp
2787-2789, Sept 1992.
[11] T. Chevalier, G. Meunier, A. Kedous-
Lebouc, and B. Cornut, “Numerical
Computation of the Dynamic Behavior
of Magnetic Material Considering
Magnetic Diffusion and Hysteresis,”
IEEE Trans. Magnetics, VOL. 36,
pp.1218-1221, July 2000.