Professional Documents
Culture Documents
~
⎡ A1 ⎤
⎢~ ⎥
⎡ ∂N1 ∂N 2 ∂N3 ∂N 4 ∂N5 ∂N6 ⎤⎢ A2 ⎥ (7)
~
⎢ ∂y
=⎢ ∂y ∂y ∂y ∂y ∂y ⎥⎢ A3 ⎥ = ∇ × NT A
[ ][ ]~
⎥⎢ ~ ⎥
⎢− ∂N1 − ∂N 2 ∂N ∂N ∂N ∂N6 ⎥⎢ A4 ⎥
− 3 − 4 − 5 − ~⎥
⎣ ∂x ∂x ∂x ∂x ∂x ∂x ⎦⎢ A
⎢ ~5 ⎥
⎣⎢ A6 ⎦⎥
In the above two equations,
Fig. 2. When the operating point moves below the H axis.
[A~] = [A~ 1
~
A2
~
A3
~
A4
~
A5 ]
~ T
A6 (8)
When the operating point moves below the H axis (B < 0), is the A’s values on each node as an example. Here second
the original B-H curve is extended so that the simulation can order 6-node element is used. For the xy solver:
continue (Fig. 2). However a warning will be given to the user ⎡ i j k⎤ ⎡ ∂A ⎤
because the B-H curve in the second quadrant, as specified by ⎢∂ ∂ ∂ ⎥ ⎛ ∂A ∂A ⎞ ⎢ ∂y ⎥ (9)
∇× A = ⎢ ⎥ = ⎜⎜ i − j ⎟⎟ = ⎢ ⎥
most commercial manufacturers, is not enough to express the
⎢ ∂ x ∂ y ∂ z ⎥ ⎝ ∂ y ∂ x ⎠ ⎢− ⎥ ∂ A
properties of PM fully. Hitherto such situation is not allowed in ⎣⎢ ∂x ⎦⎥
⎣⎢ 0 0 Az ⎦⎥
practical applications and users have to revise their designs.
In the Galerkin method, the weighting function is:
W e = [N1 N 2 N3 N 4 N5 N 6 ]
T
III. IMPLEMENTATION IN TIME-STEPPING FEM (10)
and
A. Basic FEM Formulation ⎡ ∂N i ⎤
⎢ ⎥ ⎡ (∇N i )y ⎤ ⎡(∇ × N i )x ⎤
∇ × N i = ⎢ ∂y ⎥ = ⎢
In order to deduce the FEM formulation, in this paper, a (11)
space vector is expressed in a matrix. For example, the ∂N − (∇ ) ⎥ = ⎢(∇ × N ) ⎥
⎢− i ⎥ ⎣ N i x⎦ ⎣ i y⎦
magnetic flux density B is expressed in matrix form as: ⎣⎢ ∂x ⎦⎥
B = Bx iˆ + By ˆj = [Bx By ] (1)
T
where N i is a vector in the z direction.
The basic field equation in the regions of air, iron cores, solid
conductors and PMs is: B. Nonlinear Iteration Formulation
∂A σ ˆ
∇ × (ν∇ × A) + σ − Vb k = J + ∇ × H c (2) PM is characterized for its two distinct directions, the easy
∂t l axis u and the hard axis v, and the magnetic anisotropy
where, A is the magnetic vector potential, ν is the reluctivity of nonlinear iteration formulation needs to be derived. To the
material and σ is its conductivity, l is the depth of the model in FEM problem, the algebraic equation obtained after
the z-direction, Vb is the voltage of the conductor. By applying discretization is [6]
the Galerkin method and using the shape function N as the {}
[S ] A~ = {P } (12)
weighting function, Its nonlinear iterative formula of the Newton-Raphson
⎛ ∂A σ ˆ ⎞
∫∫ (∇ × A) ⋅ν∇ × NdΩ + ∫∫ ⎜ σ − Vb k ⎟ ⋅ NdΩ method is
T
⎝ ∂t l ⎠
{ } {}
Ω Ω
[J ]n A~ n +1 − A~ n = {P}n − [S ] A~
n
(13)
= ∫∫ J ⋅ NdΩ + ∫∫ (∇ × H c ) ⋅ NdΩ (3)
Ω Ω Defining
here N is a vector. In two-dimensional (2-D) FEM, supposing
the solution domain is on the x-y plane, A and N only have the
f = [S ] A
~
{} (14)
the Jacobian matrix is
components in the z direction,
⎡ ∂ ( f − P ) ⎤ ⎡ ∂f ⎤
A = Akˆ (4) J =⎢ = ~ (15)
⎣ ∂A ⎥⎦ ⎢⎣ ∂A ⎥⎦
~
N = Nkˆ (5)
The scalar variables A and N are the components in the z From the field equation in the PM region, one has
fi ( A) = ∫∫ ν (B 2 )(∇ × A) ⋅ ∇ × N i dΩ = ∫∫ H T ⋅ ∇ × N i dΩ
T
direction, respectively. Ω Ω
6
A = ∑ N k ( x, y )Ak = [N1 N6 ]
~
∑W
N2 N3 N4 N5 (16)
k =1
= k ⋅ H T ⋅ ∇ × Ni
[ ]
A6 = [N ] A []
~ ~ ~ ~ ~ ~ T k (Gauss point)
T ~
⋅ A1 A2 A3 A4 A5 (6)
where Wk is the weighting value at the kth Gauss point. The
T ~ . Moreover,
⎡ ∂A ∂A ⎤ magnetic vector on node j is denoted as A
B = ∇× A = ⎢ − ⎥
j
⎣ ∂y ∂x ⎦ ⎡ ∂H T ⎤
∂fi ( A)
∂Aj
~ = ∑
Wk ⋅ ⎢ ~ ⋅ ∇ × N i ⎥ (17)
k (Gauss point) ⎣⎢ ∂Aj ⎦⎥
To perform element assembly on PM materials, the
coordinates are rotated to the u-v coordinates and one has
ANSYS 2011中国用户大会优秀论文
∂H T ∂
~ = ~ Hu
∂A j ∂A j
[ Hv ] cross-sectional area of the region occupied by the winding in
the solution domain; Nw is the total conductor number of this
winding; a is the number of parallel branches in the winding; Rw
T ⎡ν ′ ν vu′ ⎤ ′ ν vu
⎡ν uu ′ ⎤
= (∇ × N j ) ⎢ uu
ν ′ ν ′
[ ]
⎥ = (∇N j )v − (∇N j )u ⎢ν ′ ν ′ ⎥
(18) is the d.c. resistance of the winding; iw and uw are the branch
⎣ uv vv ⎦ ⎣ uv vv ⎦ current and voltage of the winding, respectively; p is the
Therefore, symmetry multiplier which is defined as the ratio of the original
⎡ν ′ ν vu
′ ⎤
[ ⎡
] ⎤
∂f i ( A) full cross-sectional area to the solution area. The additional
~ = ∑ Wk ⋅ ⎢ (∇N j )v − (∇N j )u ⎢ uu ⎥ ⋅ ∇ × Ni ⎥ current iad are introduced in regions of solid conductors to
∂Aj k (Gauss point) ⎣ ′ ν vv′ ⎦
⎣ν uv ⎦ ensure the last coefficient matrix of the field - circuit coupled
⎡ ′ ν vu
⎡ν uu ′ ⎤ ⎡(∇ × N i )u ⎤ ⎤ (19)
= [ ]
∑ Wk ⋅ ⎢ (∇N j )v − (∇N j )u ⎢ν ′ ν ′ ⎥ ⋅ ⎢(∇ × N i ) ⎥ ⎥
equations is symmetrical [6]. Using the Galerkin method to
discretize the field equation and circuit equations in the
k (Gauss point) ⎣ ⎣ uv vv ⎦ ⎣ v ⎦⎦
∂H u ∂ν ∂ν ∂B 2 1 ⎛ ∂H ⎢ M 31 0 ΔtS33 ⎥ ⎪⎩ iad
k ⎪ ⎪ lp ⎪ ⎪
⎭ ⎪⎩ 0 ⎪⎭ ⎪ M 31 A
k −1 ⎪
⎞ ⎪
= Bu + ν = 2 Bu + ν = 2 ⎜ − ν ⎟ Bu2 + ν (21) ⎣⎢ ⎦⎥ ⎩ ⎭
∂Bu ∂Bu ∂B ∂Bu B ⎝ ∂B ⎠ (30)
∂H v ∂ν ∂ν ∂B 2
1 ⎛ ∂H ⎞ The branch equation of the external circuits is:
= Bv + ν = 2 Bv + ν = 2 ⎜ − ν ⎟ Bv2 + ν (22)
∂Bv ∂Bv ∂B ∂Bv B ⎝ ∂B ⎠ [Re ]{ie } = {ue } + {Qe } (31)
∂H u ∂ν ∂ν ∂B 2 1 ⎛ ∂H ⎞ where Re is the matrix of the resistance and Qe is the column
= Bu = 2 Bu = 2 ⎜ −ν ⎟ Bu Bv (23)
∂Bv ∂Bv ∂B ∂Bv B ⎝ ∂B ⎠ matrix associated with sources. Multiplying -Δt/lp to the two
∂H v ∂ν ∂ν ∂B 2
1 ⎛ ∂H ⎞ sides of (31), one has:
= Bv = 2 Bv = 2 ⎜ −ν ⎟ Bu Bv (24)
⎡ Δt ⎤ ⎧ Δt ⎫ ⎧ Δt ⎫
∂Bu ∂Bu ∂B ∂Bu B ⎝ ∂B ⎠ − R {i } = − u + − Q
⎢ lp e ⎥ e ⎨ e
(32)
⎬ ⎨ e ⎬
Here the following relationship is used: ⎣ ⎦ ⎩ lp ⎭ ⎩ lp ⎭
⎛H⎞
Adding the external circuit equations (32) into (30) gives:
∂⎜ ⎟ ⎡ T11 ⎤
∂ν B ∂B 1 ⎛ ∂H H ⎞ 1 1 ⎛ ∂H ⎞ ⎢ S11 + Δt S13 ⎥ ⎧ A k ⎫ ⎧ 0 ⎫ ⎧Q A + T11 A k −1 ⎫
= ⎝ ⎠ 2 = ⎜
S12 0
− ⎟ = ⎜ −ν ⎟ (25)
⎢ M ⎪ ⎪ ⎪ Δt k ⎪
⎪ Δt ⎪
∂B 2 ∂B ∂B B ⎝ ∂B B ⎠ 2 B 2 B 2 ⎝ ∂B ⎠ ⎢ 21 ΔtS 22 + M 22 0 0 ⎥⎥ ⎪ i wk ⎪ ⎪⎪− lp uw ⎪⎪ ⎪⎪ M 21 A k −1 ⎪⎪
Δt ⎨ k ⎬=⎨ ⎬+⎨ ⎬
⎢ 0 0 − Re 0 ⎥ ⎪ ie ⎪ ⎪− Δt uek ⎪ ⎪ − Δt Qe ⎪
C. Circuit Equation Coupling Using Loop Method ⎢ lp ⎥ ⎪ k ⎪ ⎪ lp ⎪ ⎪ lp ⎪
⎢ M ⎩ i ad ⎭ ⎪
The electric circuit equations of the windings in the ⎣ 31 0 0 ΔtS 33 ⎥⎦ ⎩ 0 ⎪⎭ ⎪⎩ M 31 A
k −1
⎪⎭
magnetic field domain are coupled with FEM equations. The (33)
overall equations are [6]: Using the loop method, the relationship between the branch
∂A d w N w
∇ ⋅ (ν∇A) − σ
d N current ib and the loop current il is:
+ iad + w w iw = 0 (field equation) (26)
∂t Swap S wap {ib } = [BlbT ]{il } (34)
dwNw ∂A Rw 1
S wap ∫∫Ω ∂t
− dΩ − iw = − uw (branch equation) (27) where Blb is the loop-to-branch incidence matrix. The
lp lp Kirchhoff’s voltage law can be expressed as,
d N
− w w ∫∫
∂A R
dΩ + w iad = 0 (additional equation) (28) [Blb ]{ub } = 0 (35)
S wap Ω ∂t lp Substituting these relationships into the system equations,
where; dw is the polarity (+1 or –1) to represent, respectively, one obtains the final global equations:
the forward paths or return paths of the windings; Sw is the total
ANSYS 2011中国用户大会优秀论文
⎡ M11 ⎛ S12 0 ⎞ T ⎤ has surface mounted ceramic magnet. The transient process is
⎢ S11 + ⎜⎜ ⎟⎟ Blb S13 ⎥ ⎧ M 11 k −1 ⎫
⎢ Δ t ⎝ 0 0⎠ ⎥ k ⎪Q A + Δt A ⎪ simulated when one phase is suddenly short-circuited when
⎢ ⎛ ⎛ ΔtS 22 + M 22 0 ⎞ ⎥ ⎧ A ⎫ ⎪⎪ ⎛ M 21 A k −1 ⎞ ⎪⎪ time is 0.25 s. Fig. 3 shows a typical flux plot. Fig. 4 shows the
⎢ B ⎜ M 21 0 ⎞⎟ B ⎜ Δt ⎟⎟ BlbT
⎪ ⎪ ⎜
0 ⎥ ⎨ ilk ⎬ = ⎪⎨ Blb ⎜ Δt
⎟⎪
lb ⎜
⎢ ⎝ 0 ⎟ lb ⎜
⎜ 0 − Re ⎥ ⎪ k ⎪ ⎪ ⎜ − Qe ⎟⎟ ⎬⎪ stator phase current. The short-circuit current is about 5.5 times
⎢
0⎠
⎝ lp ⎟⎠ ⎥ ⎩ ad ⎭ ⎪ ⎝
i lp ⎠⎪ of the rated current. The armature reaction during the short
⎢ M 31 0 ΔtS 33 ⎥ ⎪ M 31 A
k −1
⎪
⎢ ⎥ ⎪⎩ ⎪⎭ circuit demagnetizes the PM. Fig. 5 shows the change of the
⎣ ⎦ average HcEq in the PM. Before the short circuit, its value is
(36) about 31500 A/m; after the short circuit, its value is reduced to
where the coefficient matrix is symmetrical. about 30600 A/m. It can be observed that irreversible
D. The Implementation of Nonlinear Problem demagnetization will happen for this design. Fig. 6 shows that
the distribution of Hc of PM is still uniform after the short
The derivative ∂H u ∂B u is dependent on the history of the circuit current reaches steady-state.
operating point in each element. The data are stored element by
element for all nonlinear PM materials. The direction of
magnetization remains as those assigned before. Each finite
element of the PM materials uses its own recoil curve in the
respective transient simulation. Therefore the worst case of
demagnetization for the entire transient simulation is recorded.
When restarting the simulation, the mesh and geometry must be
identical with those of the previous design and target design.
In each object with the same material, the recorded data are Fig. 3. The flux plot of the PM motor.
Hc and νeq. In each element, the recorded data are:
Bmin: the minimum value of Bu (u is the direction of
magnetization) in the simulation history.
HcEq: the current equivalent Hc (which is dependent on the
operating point).
Bmin_temp: the temporary value of Bmin. It is needed in
nonlinear iteration.
HcEq_temp: the temporary value of HcEq. It is needed in
nonlinear iteration.
Status is a flag used to remember the location of the
operating point:
⎧1 : operating point on nonliner curve (37)
Status = ⎨ Fig. 4. The stator current when one phase is suddenly short-circuited.
⎩2 : operating point on recoil curve
The procedures in time stepping FEM are:
(1) Initialize the data: Bmin = Br, HcEq = νeqBr, Bmin_temp = Br,
HcEq_temp = νeqBr, Status = 1.
HcEq (A/m)
V. CONCLUSION
By using a matrix to express a space vector, the nonlinear
formulation of FEM can be conveniently deduced. It is
applicable to any type and any order of finite elements, and also
to anisotropy materials such as PMs studied in this paper. A
time-stepping field-circuit coupled FEM to address the
irreversible demagnetization process of PM is developed. Only
normal B-H curves of the PM materials are used. An effective
algorithm to implement the methodology with nonlinear
iteration is presented. The status of the operating point at each
element is recorded and the nonlinear iteration is directly
controlled by the true Jacobian matrix.
VI. REFERENCES
[1] J.-R. R. Ruiz, J. A. Rosero, A. G. Espinosa and L. Romeral, “Detection of
demagnetization faults in permanent-magnet synchronous motors under
nonstationary conditions,” IEEE Trans. Magn., vol. 45, no. 7, pp.
2961-2969, July 2009.
[2] W. N. Fu, S. L. Ho, H. L. Li and H. C. Wong, “A multislice coupled
finite-element method with uneven slice length division for the simulation
study of electric machines,” IEEE Trans. Magn., vol. 39, no. 3, pp.
1566-1569, May 2003.
[3] Ki-Chan Kim, Kwangsoo Kim, Hee Jun Kim and Ju Lee,
“Demagnetization analysis of permanent magnets according to rotor
types of interior permanent magnet synchronous motor,” IEEE Trans.
Magn., vol. 45, no. 6, pp. 2799-2802, June 2009.
[4] S. Ruoho, E. Dlala and A. Arkkio, “Comparison of demagnetization
models for finite-element analysis of permanent-magnet synchronous
machines,” IEEE Trans. Magn., vol. 43, no. 11, pp. 3964-3968, Nov.
2007.
[5] W. N. Fu and S. L. Ho, “Matrix analysis of 2-D eddy-current magnetic
fields,” IEEE Trans. Magn., vol. 45, no. 9, pp. 3343-3350, September
2009.
[6] W. N. Fu, P. Zhou, D. Lin, S. Stanton and Z. J. Cendes, “Modeling of
solid conductors in two-dimensional transient finite-element analysis and
its application to electric machines,” IEEE Trans. Magn., vol. 40, no. 2,
pp. 426-434, March 2004.