1228 IEEE TRANSACTIONS ON MAGNETICS, VOL. 34, NO.

4, JULY 1998

ON ROT~TIONALEDDY CURRENT LOSSES IN STEEL LAMINATIONS

I.D. Mayergoyz
Electrical Engineering Department
University of Maryland, College Park, MD 20742

Abstract-It is analytically shown that rota-

y current losses are exactly equal to the Z ( t ) = ZxBm sinwt + ZYB, COSwt, (1)
sum of eddy current losses for two unidirectiond where e‘, and ZYare unit vectors along x- and y-axes. By
and orthogonal components of magnetic flu den- invoking the Maxwell equation:
sity acting separately under the assumption that
the peak value of the magnetic flux density is uni-
form over a lamination cross-section. By using ex-
.+
curl E = --.
as
act analytical solutions, the rotational eddy current
at
losses are also evaluated when the peak value of the and using 1D approximation (thin lamination), we find:
magnetic flux density is nonuniform.
Index terms-eddy currents, steel lamination -dE,
=-- dB, -%E,
=- OB,
(3)
lQSS@S. az at az at .
Rotating magnetic fields occur in various types of elec- From formulas (1) and (3), we obtain:
tric machinery and actuators. It has long been realized
that eddy current losses in steel laminations caused by
those fields are appreciably higher than eddy current losses
due to unidirectional alternating magnetic fields of compa-
rable magnitude. The rotational eddy current losses were By assuming that B, is uniform over lamination cross-
first studied experimentally (see [l]and [ 2 ] )and then they section and using the odd symmetry of E, and E, with
were evaluated numerically in [3] by solving the appro- respect to z , we integrate the equations in formula (4):
priate nonlinear diffusion equations. It was observed in
papers [l], [a],and [3] that sometimes the rotational eddy E x ( z , t ) = wzBmsinwt, (5)
current losses are approximately equal t o the sum of eddy
current losses associated with two unidirectional and or- E Y ( z , t )= wzB,coswt. (6)
thogonal components of the magnetic flux density acting Now, we can compute the local power density loss by using
separately. the generic formula:
In this paper, it is first shown, by using simple an-
alytical arguments, that the above equality is exact un-
der the assumption that the peak value of the magnetic (7)
flux density is uniform over a lamination cross-section. By substituting expressions ( 5 ) and (6) into formula ( 7 ) ,
It is remarkable that under the “uniformity” assumption we derive:
the above exact equality holds for magnetically nonlinear
isotropic a well as anisotropic laminations with (and with-
out) hysteresis. Next, by using analytical results from pa-
per [4], the rotational eddy current losses are also exactly It is apparent from the last equation that in the case of cir-
evaluated for the case when the peak value of the mag- cular polarization of the magnetic flux density the “eddy
netic flux density is nonuniform over the lamination cross- current” energy dissipation occurs at constant rate in time.
section. The results obtained in this direction complement As a result, instantaneous power loss pcir(z,t ) is equal to
the formulas for rotational eddy current losses derived un- the &erage power loss Fir.This clearly reveals why rota-
der the “uniformity” assumption and they also clearly re-
tional eddy current losses are higher than those for unidi-
veal the limits of that assumption. rectional magnetic fields.
To start the discussion, consider a magnetic conduct-
The eddy current power losses per unit surface area
ing lamination with height h, width w, and thickness A. of lamination can be obtained by integrating jF”(z) with
It is assumed in the sequel that A << h and A << w, which respect to z . This yields:
means that we deal with a thin lamination. It is also as-
sumed that the magnetic flux density is circularly polarized --cir A3
and has only x- and y-components: P = a w 2 B 2- (9)
12

0018-9464/’98$10.00 0 1998 IEEE

 1229

By taking into account that @, = BmA,the last formula By comparing formulas (13) and (17), we conclude that:
can be written in the form: -1 -1in -1in
P =P, +P,.
It is important to stress that in the previous derivations
we have not used any assumptions concerning magnetic
Non., we shall evaluate eddy current losses for the case properties of the conducting lamination. This means that,
when the magnetic flux den_sity is linearly polarized along under the “uniformity” assumption, exact equalities (14)
the y-axis, that is when B ( t ) = e‘,B,coswt. For this and (18) are valid for magnetically nonlinear isotropic as
case. the electric field has only z-component, which is given well as anisotropic laminations with (and without) hystere-
by formula ( 5 ) . By using this formula, we arrive at the sis.
following expression for the instantaneous power losses: Next, we shall depart from the “uniformity” assump-
tion and shall try t o evaluate rotating eddy current losses
plin(z,t ) = aw2B%z2cos2 w t , (11) for nonuniform distributions of the magnetic flux density
over the lamination cross-section. We shall base our anal-
which leads to the following expression for average eddy ysis on exact analytical solutions obtained in paper [4] for
current power losses: the case of circular polarizations of the magnetic field and
magnetically nonlinear isotropic media with the constitu-
1 tive relation H = ( f ) n ,where n is usually larger than ten.
jj””(z) = -aw2Biz2. (12)
2 For the case of magnetic conducting laminations. these so-
lutions can be written as follows:
By integrating the last expression with respect t o z , we
-+ + zo.
i
easily derive the following formula for the power losses per H , (1 - + ) a , if-+ 5 z I
5 5+
unit surface area:
&(z) = 0, if-+ + zo z - zo,
-1in
P =a&--
w2aA
(13)
H, (A;”)a
1-1- , if+-zo<zi+,
24 .
(191

i
By comparing formulas (10) and (13), we find: -jH, (1 - * “0
)a, if-$ 5 z 5 -$ + Z O ,
I;Ty(z)= 0, i f - + + z o < z < TA - z o

The previous discussion can be easily generalized to the

- j ~ ,(1 - e)*, if+ - zo
(20)
5 z 5 +,

case of elliptical golarizations of the magnetic flux density, where exponent a and pentration depth zo are given by
that is when: B ( t ) = Z,,B,,sinwt +
Z,,B,,coswt. By the expressions:
using the same line of reasoning as before, it can be shown
that in the case of elliptical polarizations formulas ( 5 ) and
(6) are still valid with the replacement of B, by B,, and
B,,, respectively. This means that we have the following
zo =
+
[2n(n 1)(3n l)’]+ +
expressions for the peak values of E,(z, t ) and Ey(z, t): J=iG(n - 1) ’
and
1-1
p,=kH; . (23)
By using the last two expressions, the local eddy current Formulas (19) and (20) are valid for zo 5 First, let us e.
losses can be evaluated as follows: consider the case when we have strict inequality 20 <
To compute the eddy current losses per unit surface area
e.
1 1
p”’(z)= ;a(E;, + EL,) = -ow2(B;,
2
+ B;,)z2. (16) --sir
P , we shall use the notion of surface impedance 7 . which
leads t o the formula:
By integrating the last expression with respect t o z and -cir
by using the same transformations as before, we arrive at P = 2HkRe(q). (24)
the following formula for the eddy current losses per unit
From analytical solutions (19) and (20), we can easily de-
surface area in the case of elliptical polarizations:
rive the following expression for the surface impedance:
a
q=-,
020
1230

By substituting (25) into (24) and taking into account for- By substituting formulas (25) and (29) into (24), we arrive
mulas (21) and (22), we arrive at the following result: at:
(33)
By taking into account expressions (32) and (21), we de-
rive:
In the above formula, rotational eddy current losses axe n-1
expressed in terms of the magnetic field peak value at the (34)
lamination boundary. In applications, it is desirable to ex-
press eddy current losses in terms of the total flux through
For sufficiently large n we have ei M and the last
formula assumes the form:
the lamination. This can be done by using the following
relation between the peak value @m of the total flux and
H, :
(35)

The last equation coincides with formula (lo), derived by

using the “uniformity” assumption and a different line of
which can be derived from exact solutions (19) and (20) reasoning. This coincidence is natural. This is because
By substituting (25) into (27), we find: under condition (32) the distribution of B, is almost uni-
form (see Figure 1). Thus, we can conclude that for
zo 5 we can use f o r m u l a (30) (or (31)) for the
evaluation of rotational e d d y c u r r e n t losses, while
for to >. $ formula (10) is a p p r o p r i a t e , and there
Since zo depends on p m (see formula (22)), which, in turn, is a sufEciently s m o o t h t r a n s i t i o n f r o m the values
depends on H , (see formula (23)), the last expression can of eddy c u r r e n t losses p r e d i c t e d by formula (30)
be construed as a nonlinear equation for Hm. By solving (or (31)) t o the value of those losses predicted by
this equation, we can find H , and p m as functions of G
,. formula (10 ) .
In other words, for any given 4p, we can find the appro-
priate values of p m and 20. By using these values in (28),
we obtain the following expression for H , in terms of a
:,

By substituting (29) into (as),after simple transformations Figure 1

we arrive at: It is also important to point out that in the cases
zo 5 and zo > $ there are different frequency depen-
--cir wb [2n(n + 1)(3n + 1)’]$
2(3n+1) ‘ (30)
those losses grow as w P -
dencies for rotational edd current losses. In the first case,
2 , while in the second case those

losses increase as N w2.Another important point is that

Since n is typically large ( n > lo), the last formula can be the above two cases are determined not only by the fre-
approximately simplified as follows: quency of magnetic flux but by medium saturation as well
This is because the penetration depth is field dependent.
Thus, the level of s a t u r a t i o n may affect the fre-
q u e n c y dependence of the r o t a t i o n a l e d d y c u r r e n t
losses.
Formulas (30) and (31) can be used for the calculation of REFERENCES
rotational eddy current losses in the case when the peak
1. R.D. Strattan and F.J. Young, J. Appl. Phys.. 33.
value of the magnetic flux density is nonuniform over the (1962), 1285-1286.
lamination cross-section. In this sense, the above formulas 2. F.J. Young and H.L. Schenk, J. Appl. Phys., 37,
complement the expression (10) derived under the “unifor- (1966), 1210-1211.
mity” assumption. 3. R.M. DelVecchio, IEEE Transactions on Magnetics,
Next. we consider the case when: 18, (1982), 1707-1709.
4. I.D. Mayergoyz, F.M. Abdel-Kader and F.P.
zo = -.A2 Emad, J. Appl. Phys., 55, (1984), 618-629.