Reduced Order Modeling of Rectangular Wires in a

Magneto-quasi-static Field for Integral Formulation

Shingo Hiruma Luca Di Rienzo Carlo de Falco

Graduate School of Engineering Dipartimento di Elettronica, MOX—Modeling and Scientific
Kyoto University Informazione e Bioingegneria Computing, Dipartimento di Matematica
2024 IEEE 21st Biennial Conference on Electromagnetic Field Computation (CEFC) | 979-8-3503-4895-8/24/$31.00 ©2024 IEEE | DOI: 10.1109/CEFC61729.2024.10585815

Kyoto, Japan Politecnico di Milano Politecnico di Milano

[email protected] Milano, Italy Milano, Italy
[email protected] [email protected]

Abstract—This paper presents a new approach based on the occupied by the wire, and air Ω0 . Given an external, known
Cauer ladder network method to calculate the complex magnetic flux density 𝐁0 , magnetization 𝐌 and magnetic flux
permeability of a rectangular wire, which is a key ingredient of the density 𝐁M in the wire are governed by the following IE
integral equation formulation. The proposed method has
equivalent accuracy but lower computational cost and memory 𝜇𝑟 − 1
𝐌(𝐱) = (𝐁0 (𝐱) + 𝐁M (𝐱)), 𝐱 ∈ ΩM (1)
requirements with respect to the standard finite element method, 𝜇0 𝜇𝑟
which is the current method of choice for the same computation.
where 𝜇𝑟 is the complex relative permeability of the wire.
Keywords—integral equation formulation, magneto-quasi- Replacing 𝜇𝑟 with an effective complex relative permeability 𝜇̇ 𝑟
statics, eddy current losses. of the wire, we can extend the IE formulation to the time-
harmonic regime.
I. INTRODUCTION
To discretize the IE, the domain Ω𝑀 is divided into 𝑛
Rectangular wires are widely used in many electrical cuboids with thickness 𝑑 , width 𝑤 , and height ℎ . The
equipment components that play a fundamental role in the magnetization is assumed to be uniform in each cuboid.
infrastructures of electrical systems, such as transformers and
Applying the collocation method at the geometric center 𝐱 (𝑖) of
mechanical armors of submarine power cables. Such wires can
each cuboid, we obtain the following discretized IE
be purely conductive or both conductive and ferromagnetic
depending on the application. The electromagnetic behavior of (𝜇̇ 𝑟 − 1)𝑆 (𝜇̇ 𝑟 − 1)𝑆
these wires must be accurately simulated numerically to (I − A) 𝐦 = 𝐛 (2)
𝜇0 𝜇̇ 𝑟 𝜇0 𝜇̇ 𝑟
calculate the power losses due to eddy currents and magnetic
hysteresis. A(1,1) ⋯ A(1,n) 𝐦(1) 𝐛 (1)
A=[ ⋮ ⋱ ⋮ ] , 𝐦 = [ ⋮ ] , 𝐛 = [ ⋮ ] (3)
The Finite Element Method (FEM) usually results in a large
computational burden because it requires the discretization of A(n,1) ⋯ A (n,n)
𝐦 (𝑛)
𝐛 (𝑛)

the surrounding volume and the wires; furthermore, a very fine where 𝐦(𝑖) = 𝐌 (𝑖) 𝑆 denotes the magnetic moment in the i-th
mesh is required in strong skin effect conditions. cuboid, and 𝑆 denotes the area of the cross-section of the
Integral Equation (IE) formulations are considered as valid cuboids, being 𝑆 = 𝑤 ⋅ 𝑑. Submatrix A(𝑖,𝑗) ∈ ℝ3×3 defines the
alternatives [1], [2]. By restricting the computational geometry relation between the i-th and j-th cuboids, whose k-th column
only to the wires and by introducing the assumption of the (k=1,2,3) is given by
uniformity of the magnetization in the discretized elements, the (𝑗)
computational burden can be significantly reduced by requiring (𝑖,𝑗) 𝜇0 𝐞 𝐱 (𝑖) − 𝐱 ′ 2 ′
𝒂𝑘 =− ∫ (𝐧(𝐱 ′ ) × 𝑘 ) × (𝑖) d 𝐱 (4)
fewer degrees of freedom. 4𝜋 Ω(𝑖,𝑗) 𝑆 |𝐱 − 𝐱 ′ |3
𝑀

This paper introduces a new method utilizing a reduced order which can be evaluated analytically or via numerical integration.
modeling technique, the Cauer ladder network (CLN) method
[3], to calculate the complex permeability of a rectangular wire, B. Calculation of the complex permeability
which is required for the IE formulation [2]. The present method To obtain the complex permeability of a rectangular wire,
provides complex permeability that is reliable for a wide range we have to numerically solve a boundary value problem over a
of frequencies at a lower computational cost than using FEM. wide range of the frequencies because no analytical solution is
available. Instead of using FEM, we aim at reducing the
II. FORMULATION computational cost to obtain the frequency profile of the
A. Magneto-quasi-static integral equation formulation complex permeability by the CLN method [3].
The IE formulation considered here is introduced in [2] and To evaluate the complex permeability using the CLN
focuses on wires of rectangular cross section. The three- method, we first consider a cross-section of the rectangular wire
dimensional space is partitioned into ΩM , which is the domain immersed in the time-harmonic uniform external flux density

Authorized licensed use limited to: PRYSMIAN SpA. Downloaded on October 02,2024 at 08:57:32 UTC from IEEE Xplore. Restrictions apply.
B0 𝒆𝑘 , 𝑘 = 1,2. The electromagnetic fields are then reduced to a
Cauer circuit using the CLN method for each axis 𝒆𝑘 . Solving
the Cauer circuit at a certain frequency, we can reconstruct the
electromagnetic fields as follows:
𝑚 𝑚

𝐇(𝐱) = ∑ 𝑖2𝑖+1 (𝜔)𝐇2𝑖+1 (𝐱) , 𝐄(𝐱) = ∑ 𝑒2𝑖 (𝜔)𝐄2𝑖 (𝐱) (5)

𝑖=−1 𝑖=1

As a result, we can evaluate the magnetization and the moments

via
1 − 𝜇𝑟 1
𝐦𝑀 = ∫ 𝐁(𝐱 ′ )d2 𝐱 ′ , 𝐦𝐽 = ∫𝐱 ′ × 𝐉(𝐱 ′ ) d2 𝐱 ′ (6),
S 𝜇0 𝜇𝑟 2 𝑆
Fig. 1 Frequency characteristics of the relative complex permeability of
and also the equivalent total magnetic moments and relative rectangular wires obtained by the CLN method and FEM.
complex permeability by
𝐦 = 𝐦M + 𝐦J , (7)
1
𝜇̇ 𝑟𝑘 (𝜔) = 𝜇0 m𝑘 , 𝑘 = 1,2 (8) Fig. 2 2D test case: two infinitely long straight rectangular conductors are
1− placed at distance Δ𝑥 and are immersed in a uniform time-harmonic
𝑆 ⟨B𝑘 ⟩
magnetic field.
where the averaged flux density is defined as ⟨B𝑘 ⟩ = 0.35 0.9
1 Hx 50Hz
∫ B (𝐱 ′ ) d2 𝐱 ′ . The complex permeability along the Hx 50Hz

Relative error (imaginary part)

0.3 0.8
Hy 50Hz
𝑆 𝑆 𝑘 Hy 50Hz

Relative error (real part)

Hx 1kHz 0.7 Hx 1kHz
longitudinal direction of the wire can be obtained in the same 0.25 Hy 1kHz 0.6 Hy 1kHz
manner. 0.2 0.5
0.15 0.4
The advantage of this method lies in that the frequency 0.3
0.1
characteristics can be obtained with lower computational cost 0.2
0.05
and memory than using FEM because the size of the Cauer 0.1
0
circuit is much smaller than that of the FEM model. 0.01 0.1 1 10
0
0.01 0.1 1 10
Normalized distance Δx/w Normalized distance Δx/w
III. NUMERICAL RESULT Fig. 3 Relative error on the real (left) and imaginary (right) parts of the x,
y component of the magnetization at 50 Hz and 1 kHz.
A. Validation of the CLN method
First, we calculated the complex permeability of a for both real and imaginary parts increases when the normalized
rectangular conductor with a cross-section of width 𝑤 = 12 distance are smaller than 1. This is due to the assumption of the
[mm] and thickness 𝑑 = 3 [mm], with conductivity 𝜎 = uniformity of the magnetization, which is actually not valid for
7.246 × 106 [S/m], and relative permeability 𝜇𝑟 = the rectangular cross-section case.
300 exp(−𝑗 𝜋⁄3). We applied the CLN method to obtain 10
basis functions for electric and magnetic fields. The frequency IV. CONCLUSION
characteristics of the complex permeability is shown in Fig. 1. It We have introduced the CLN method for the computation of
can be noticed that the complex permeability is in perfect the complex permeability, which is required for the integral
agreement with the solution obtained by FEM. The equation formulation introduced in [2]. The CLN method
computational cost to obtain the complex permeability for each enables us to obtain the frequency characteristics with lower
frequency in the CLN method is much smaller than that required computational cost and less memory than FEM. Although the
in FEM because the size of the Cauer circuit is just 10. obtained complex permeability is satisfactory accurate, the
B. Infinitely long straight rectangular conductors modeling error increases when the conductors are closer. To
reduce the modeling error, we plan to modify submatrices A(𝑖,𝑗)
The IE formulation (2) is implemented in the magneto-quasi-
for the adjacent cuboids by a solution of a two-dimensional
static approximation for a 2D test case where two infinitely long
variational problem, which can be also obtained by CLN.
straight rectangular conductors are immersed in a uniform time-
harmonic magnetic field oriented in the x and y-axis, as shown REFERENCES
in Fig. 2. Both conductors have 𝜇𝑟 = 300 exp(−𝑗 𝜋⁄3). In this
[1] L. Giussani, L. Di Rienzo, M. Bechis and C. de Falco, “Computation of
case the unknowns are 𝐦(1) for the first conductor and 𝐦(2) for armor losses in AC submarine cables," IEEE Transactions on Power
the second one and for symmetry 𝐦(1) = 𝐦(2) . The reference Delivery, vol. 36, no. 5, pp. 3014-3021, (2021).
solution is obtained by FEM. [2] Cambareri, Pasquale, et al. "An Integral Formulation for Rectangular
Wires in a 3D Magneto-Quasi-Static Field," IEEE Access (2023).
In Fig. 3 the relative error on the real and imaginary parts of [3] A. Kameari, H. Ebrahimi, K. Sugahara, Y. Shindo, and T. Matsuo, “Cauer
the x and y component of 𝐦 at 50 Hz and 1 kHz are plotted Ladder Network Representation of Eddy-current Fields for Model Order
against the normalized distance Δ𝑥 ⁄𝑤 . We can see that the error Reduction Using Finite Element Method,” IEEE Trans. Magn., vol. 54,
no. 3, Art no. 7201804 (2018).

Authorized licensed use limited to: PRYSMIAN SpA. Downloaded on October 02,2024 at 08:57:32 UTC from IEEE Xplore. Restrictions apply.