Downloaded from SAE International by Harbin Engineering University, Saturday, July 22, 2023

2023-01-1074 Published 08 May 2023

Simulation and Validation of Stator Modes of a

Hairpin Motor
Fulong Zhang AAM Investment Management Co.

Jing Ba and Zhaohui Sun American Axle & Manufacturing

Citation: Zhang, F., Ba, J., and Sun, Z., “Simulation and Validation of Stator Modes of a Hairpin Motor,” SAE Technical Paper
2023-01-1074, 2023, doi:10.4271/2023-01-1074.

Received: 20 Nov 2022 Revised: 12 Feb 2023 Accepted: 07 Mar 2023

Abstract motor noise prediction. This paper presents a simulation

W
ith the trend of electric drive unit gradually method based on the finite element model and defined ortho-
replacing ICE powertrain, in additional to gear tropic material properties of the stator. The material property
noise, the motor noise has become a new major parameters of the stator core and hairpin windings are
NVH challenge. These tonal noises are easier to be detected reverse-engineered through iterative correlations to test data.
in the pure electric vehicle that has no masking effect of ICE High accuracy FEA model is achieved that can determine the
powertrain. Therefore, how to accurately predict and reduce stator mode shapes and frequencies of this hairpin motor
the motor noise is a key to solve the problem. The accuracy of accurately, which provides a reliable and effective approach
calculated motor stator modes determines the accuracy of for the motor noise prediction and optimization studies.

Introduction
range of operating frequency that can be  higher than

A
ccording to the overall forecast of China’s future new 10,000Hz. Meanwhile, the number of motor noise orders is
energy vehicle market and carbon emission reduction large and their root causes are complex, which increases the
given by the team of academician Ouyang Minggao difficulty of the electric drive unit modal arrangement and
of Tsinghua University, the number of new energy vehicles in NVH system integration. Motor noise is mainly caused by
China will reach 90 million in 2030; It will reach 200 million the electromagnetic concentrated radial force, tangential
by 2035 and 300 million by 2040; By 2050, the number will force and torque ripple applying on the top of stator teeth
be  around 360 million [1]. Facing the rapidly expanding and amplified by the resonances of stator assembly. In the
market of new energy vehicle, domestic and foreign OEMs three factors of noise (source, path and receiver), the accuracy
and auto parts enterprises are constantly increasing the capital of capturing path characteristics, e.g., the modes of motor
and technology investments into new energy vehicle stator, plays a key role to determine the accuracy of motor
researching and development for creating reliable, efficient, noise prediction. However, the complicated laminated struc-
energy-saving and environmental protection products to meet ture of stator core and the elongated structure of winding
the government regulation and end-customers’ demands. lines determine that their material properties are not the
NVH performance has become one of the major concerns in same as the standard isotropic material properties, which
the development of electric vehicles [2]. This is because strin- greatly increases the difficulty of calculating and predicting
gent vehicle lightweight targets and the disappearance of ICE the stator modes accurately.
powertrain masking effect have made the vibration and noise, Based on finite element method, a stator assembly model
which once were imperceptible to most users, now become of a hairpin motor that includes stator core and hairpin
apparent even to the most unfocused driver [3]. windings is presented in this paper. A reverse-engineering
Motor noise is one of the major NVH challenges among approach was taken, in which the stator core and hairpin
all noise sources. As the designed motor power density is winding material property parameters were adjusted itera-
higher and higher, the top speed of motor is increasing corre- tively based on the understanding of lamination features of
spondingly. The top speed of some drive motors has reached the elements to match the test results. High accuracy modal
or even exceeded 20,000rpm, which leads to an increasing results are obtained and presented.
 Downloaded from SAE International by Harbin Engineering University, Saturday, July 22, 2023

2 SIMULATION AND VALIDATION OF STATOR MODES OF A HAIRPIN MOTOR

Stator Structure and Theoretical Analysis of

Simulation Model Orthotropic Material
Compared with the traditional round wire motor, the slot The orthotropic material has three mutually perpendicular
filling rate of hairpin motor is higher, the contact area of flat symmetry planes, and the direction which is perpendicular
wire is larger, and the utilization rate of copper winding wire to the symmetry plane is defined as elastic principal direction.
is better. Therefore it has obvious advantages of higher power In the elastic principal direction, the elastic properties of
density, better thermal conductivity and lower cost. Although material are consistent. The axis paralleled to the elastic prin-
there are still some barriers in terms of equipment, materials cipal direction is defined as elastic or material principal axis,
and process complexity, it has become a general trend to denoted by 1,2 and 3 as shown in Figure 2.
replace the traditional round wire motor in the current era of The elastomeric governing equation of orthotropic
increasingly high requirements to the power density of drive material can be determined by generalized Hooke’s law, and
motor of new energy vehicle. the stress and strain components or their increments are
The research object of this paper is a hairpin permanent represented in the elastic principal axis coordinate system:
magnet synchronous motor with a peak power of 160KW and
a top speed of 14,000rpm. The stator core is composed of many     D   (1)
silicon steel sheets with a thickness of a fraction of a millimeter
pressed and welded together. So it is a difficult and complex work  11 
to create a laminated multi-sheet FEA model as how the physical  
part is constructed. A practical simplified method proposed in  22 
this paper is that the stator core can be represented by a FEA  33 
solid model with equivalent material property as Figure 1 shown.      (2)
  12 
The hairpin windings made of copper, are composed of two   23 
parts. One is the winding lines fixed in the slots with insulating  
resin, and the other one is the winding coils suspended at the   31 
two ends of stator core. As to the modelling approach of hairpin
windings, some scholars only considered the mass influence of
winding lines and coils but ignored the influence of their stiffness  11 
 
[4]. Some scholars only considered the mass and stiffness influ-  22 
ences of winding lines but ignored the influence of winding coils  33 
on the two sides [5]. The author believes that the mass and stiff-      (3)
ness of copper winding lines and winding coils both contribute  12 
to the modes of stator assembly and should not be ignored.  23 
 
Therefore, in the simulation environment, the FEA models of  31 
the winding lines and winding coils are established separately
to comprehensively consider the contributions of their mass and
stiffness. As shown in Figure 1, the winding coils suspended at
the two ends are also represented in the FEA solid model.

 FIGURE 1   Stator structure and simulation model  FIGURE 2   Elastic principal axis of orthotropic material
 Downloaded from SAE International by Harbin Engineering University, Saturday, July 22, 2023

SIMULATION AND VALIDATION OF STATOR MODES OF A HAIRPIN MOTOR 3

 D11 D12 D13 0 0 0   FIGURE 3   Material coordinate system of stator core

D D22 D23 0 0 0 
 12
 D13 D23 D33 0 0 0 
D   0 0 0 D44 0 0 
 (4)

 0 0 0 0 D55 0 
 
 0 0 0 0 0 D66 

Where, {σ} is the stress matrix; {ε} is the strain matrix;

[D] is the stiffness matrix and Dij (i,j=1,2……,6) is the stiffness
coefficient, that can be expressed as following equation:

 1  23 32  21   31 23  31   21 32 

 EE 0 0 0 
E 2E 3  E 2E 3 
 2 3

 12  32 13
  1  13 31  32   12 31
0 0 0 
 E1E 3 E1E 3 E1E 3
 
 13   12 23  23   21 13 1  12 21
0 0 0 
 E1E 2  E1E 2  E1E 2  
 
 0 0 0 G 23 0 0 
 0 0 0 0 G13 0 
 
 0 0 0 0 0 G12 

(5)
Where,
and welded together along with Y-axis, the elastic modulus
along with Y-axis is different with X-axis or Z-axis, however,
1  12 21  23 32  31 13  2 21 32 13 the elastic modulus along with X-axis is consistent with the
 (6) one along with Z-axis. The shear modulus on XZ-plane is
E1E2 E3
different with XY-plane or YZ-plane, however, the shear
modulus on XY-plane is consistent with the one on YZ-plane.
Where, Ei, νij and Gij (i,j=1,2,3) are the elastic modulus, The change of Poisson’s ratio in these three perpendicular
Poisson’s ratio and shear modulus respectively. planes is not obvious, so its effect can be ignored. Therefore,
At this point, the relationship between the stiffness in all the relationship among the elastic property parameters of
directions of any point in the orthotropic material, and the stator core can be redefined as following:
material property parameters is defined clearly. This is critical
for the next session of the study on the material property 1. Elastic modulus: Ex = Ez, Ey
parameters of stator core. 2. Shear modulus: Gxy = Gyz, Gxz
3. Poisson’s ratio: νxy = νxz = νyz

In this way, the original 9 elastic property parameters are

Simulation and Validation reduced to 5. Theoretically, Ex (Ez), Gxy (Gyz) and Gxz of stator
of Stator Core Modes core will have an obvious decrease comparing with the
isotropic material, which is determined by the interactions of
the laminations. With the standard material of silicon steel
Definition of Coordinate as baseline, the influence of above parameters to the modes
of stator core is studied.
System of Material Property
Parameters
The material coordinate system of stator core is defined in
FRF Testing on Stator Core
Figure 3, where the X-axis is corresponding to material prin- As shown in Figure 4, the stator core is suspended with bungee
cipal axis 1, the Y-axis is corresponding to the material prin- cable, and an accelerometer is attached at 1/4 length of stator
cipal axis 2 and the Z-axis is corresponding to the material core and at 12 o ‘clock direction. Driving point FRF was
principal axis 3. collected by using modal hammer impacting at the
Based on the theory of orthotropic material, the vibration same location.
characteristic is controlled by the following 9 elastic property Table 1 shows the baseline material properties used at the
parameters: Ex, Ey, Ez, νxy, νxz, νyz, Gxy, Gxz and Gyz. Since the starting point of model tuning. The mass density is based on
stator core is composed of many silicon steel sheets pressed the measured mass of physical part. The values of Ex, Ey, Ez,
 Downloaded from SAE International by Harbin Engineering University, Saturday, July 22, 2023

4 SIMULATION AND VALIDATION OF STATOR MODES OF A HAIRPIN MOTOR

 FIGURE 4   FRF testing on stator core  FIGURE 6   The predicted modes of stator core

model prediction and test data. The mode shapes of peaks

P1-P5 predicted by the model are shown in Figure 6.
In Figure 6, the mode shape of stator core can be described
by the number of N and M, where N is the number of pitch
circle diameter on the XZ plane and M is the number of pitch
circle along with the Y-axis [6]. Wherein the mode shapes of
M=1 are with symmetric mode shapes and those of M=2 are
with asymmetric mode shapes. In the next session, a DOE
method is utilized to study the relationship between the elastic
property parameters and these modal frequencies.

TABLE 1  Baseline material property parameters of stator core

Correlation Study between
Material property
parameter Silicon steel Stator core
FEA Model and Test Data
Mass density ρ (kg/mm3) 7.69E-06 7.48E-06 The influence of elastic modulus and shear modulus on the
Elastic modulus E (mPa) EB = 210,000 Ex/EB = 1 modal frequencies of stator core is studied and the results are
Ey/EB = 1 given in Figures 7–10.
Ez/EB = 1
Figure 7 shows that stator core modal frequency is sensi-
tive to Ex (Ez), particularly for the higher order modes. Lower
Poisson’s ratio ν νB = 0.3 νxy /νB = 1
Ex (Ez) produces lower modal frequencies.
νxz /νB = 1 Figure 8 shows that Ey variation has no significant influ-
νyz/νB = 1 ence on the stator core modal frequency.
Shear modulus G (mPa) GB = 80,000 Gxy/GB = 1 In Figure 9, the impact of Gxy (Gyz) is studied. The results
Gxz/GB = 1 show that it has no obvious influence on the frequencies of
Gyz/GB = 1 symmetric modes (M=1), but is proportional to the frequen-
cies of asymmetric modes (M=2), which means that with the
decreasing or increasing of Gxy (Gyz), the frequencies of asym-
νxy, νxz, νyz, Gxy, Gxz and Gyz of stator core are consistent with metric modes (M=2) decrease or increase accordingly. The
those of standard silicon steel. The stator core modes and reason is that the asymmetric mode shape has a 90° phase
frequency response functions are calculated by solving difference between the two pitch circles along Y-axis, and the
FEA model. direction of internal tensile stress is opposite to the direction
Figure 5 shows the comparison between the test data and of compressive stress as Figure 10 shown. The strain of asym-
simulation results based on the baseline material property metric mode (M=2) is contributed by shear stress in XY and
parameters. Significant difference is identified between the

 FIGURE 7   Ex (Ez) vs modal frequency

 FIGURE 5   Test data vs simulation data based on silicon

steel material
 Downloaded from SAE International by Harbin Engineering University, Saturday, July 22, 2023

SIMULATION AND VALIDATION OF STATOR MODES OF A HAIRPIN MOTOR 5

 FIGURE 8   Ey vs modal frequency  FIGURE 11   Gxz vs modal frequency

 FIGURE 9   Gxy (Gyz) vs modal frequency YZ planes partially, therefore the asymmetric mode (M=2) is
affected by shear modulus.
Figure 11 shows that Gxz has impact on modal frequency
as well. With decreasing or increasing of Gxz , the modal
frequencies decrease or increase accordingly. It also suggests
that both the symmetric and asymmetric modes will generate
shear stress in the XZ plane.
According to above study results, Ex (Ez), Ey, Gxy (Gyz) and
Gxz are considered as independent variables, through iterative
correlation study, a good correlation between model predic-
tion and test results is achieved, as Figure 12 shown. Where,
P1 is the mode of N=2, M=1, P2 is the mode of N=3, M=1 and
P3 is the mode of N=4, M=1. The modal frequency and peak
amplitude errors are given in Table 2, which shows that the
frequency errors from P1 to P3 are less than 2%, and the
amplitude errors are less than 3dB. The obvious gap between
the curve profile of FEA model and test data is caused by the
poor quality of test data. The coherence checking shows that
the strong correlation between hammer excitation and accel-
erometer response mainly happened at the three peaks.
 FIGURE 10   The stress of asymmetric mode shape
Although the curve profile of FEA model doesn’t match the
test data perfectly, it captured the main dynamics of the
stator core.
In the next step, based on these correlated material
parameters of stator core, the vibration characteristic of stator
assembly is studied.

 FIGURE 12   Test data vs correlated simulation data of

stator core
 Downloaded from SAE International by Harbin Engineering University, Saturday, July 22, 2023

6 SIMULATION AND VALIDATION OF STATOR MODES OF A HAIRPIN MOTOR

TABLE 2  Relative error between FEA model and test data of TABLE 3  Baseline material property parameters of winding
stator core lines and coils
FEA Model Test Data F AMP Winding line
No. F (Hz) AMP (dB) F (Hz) AMP (dB) ERR(%) ERR(dB) Material property parameter Cooper and coil
P1 822 18.4 809 18.6 1.6% 0.15 Mass density ρ (kg/mm3) 8.92E-06 5.28E-06
P2 2308 14.6 2300 14.6 0.35% 0.02 Elastic modulus E (mPa) EB = 114,000 EC/EB = 1
P3 4335 16.4 4345.5 19 0.24% 2.59 Poisson’s ratio ν νB = 0.3 νC /νB = 1
Shear modulus G (mPa) GB = 40,000 GC/GB = 1

Simulation and Validation  FIGURE 14   Test data vs simulation data based on
of Stator Assembly Modes cooper material

FRF Testing on Stator

Assembly
As shown in Figure 13, the coordinate system of stator
assembly is maintained the same as defined for stator core
only. Similar to the test on stator core, an accelerometer is
attached to the stator core at its 1/4 length and 12 o ‘clock
direction. Then modal hammer impact at this point is
conducted to generate the driving point FRF.
The mass density of winding lines and coils can
be obtained based on measured mass of physical part. As the
winding lines and coils are wound by long and thin copper
wires and fixed in the slots with insulating resin, it looks like  FIGURE 15   The predicted modes of stator assembly
a squirrel cage as shown in Figure 1, which determines that
its material property is not equivalent to the standard material
of copper, but “softer”. With the standard material of cooper
as baseline, the initial material property parameters of
winding lines and coils are defined in Table 3. The modes and
frequency response function are calculated by the FEA model,
and the result is shown in Figure 14.
Peaks P1-P5 in Figure 14 are controlled by the stator
assembly modes in Figure 15. The difference between the FEA
model and test data is significant, suggesting that although assembly model, the accuracy of material property parameters
accurate material representation of stator core has been of winding lines and coils should not be ignored.
achieved from reverse-engineering and applied to the stator
Correlation between FEA
Model and Test Data of Stator
 FIGURE 13   FRF testing on stator assembly Assembly
Using the FEA model and DOE method, the influence of
elastic modulus and shear modulus of winding lines and coils
on the vibration characteristic of stator assembly is studied,
and the results are shown in Figure 16 and 17.
The analysis results given in Figure 16 and 17 show that
with the decreasing or increasing of EC or G C , the modal
frequencies of the stator assembly decrease or increase accord-
ingly. The slopes of curves suggest that the change of EC or GC
has less impact on N=2 modes. According to above results, EC
and GC are taken as independent variables, and iterative corre-
lation study is carried out. Excellent correlation is achieved
as shown in Figure 18. Where, P1 is the mode of N=2, M=1,
P2 is the mode of N=2, M=2, P3 is the mode of N=3, M=1, P4
is the mode of N=3, M=2 and P5 is the mode of N=4, M=1.
There is a very good match of peak, valley and profile of
response curves between the FEA model and test data in
 Downloaded from SAE International by Harbin Engineering University, Saturday, July 22, 2023

SIMULATION AND VALIDATION OF STATOR MODES OF A HAIRPIN MOTOR 7

 FIGURE 16   EC vs modal frequency TABLE 4  Relative error between FEA model and test data of
stator assembly
FEA Model Test Data
AMP AMP F AMP
No. F (Hz) (dB) F (Hz) (dB) ERR(%) ERR(dB)
P1 690 -5.8 663.2 -8.2 4.04% 2.41
P2 752 10.4 760.4 14 1.11% 3.61
P3 2200 4.9 2126.1 -1.1 3.48% 5.99
P4 2300 4.2 2317.9 1.9 0.77% 2.28
P5 4267 8.9 4216.1 12.9 1.21% 3.97

At this point, an accurate FEA model is achieved with

reverse-engineered physical properties that can properly
represent this hairpin motor dynamics.

 FIGURE 17   GC vs modal frequency

Conclusions
This paper presents a simulation method based on the finite
element model and defined orthotropic material properties
of the stator. The material property parameters of the stator
core and hairpin windings are reverse-engineered through
iterative correlations to test data. High accuracy FEA model
is achieved. The following conclusions can be drawn:
1. The material property of stator core with laminated
multi-sheet structure can be represented by
orthotropic material properly. Through FEA study, it
is found that:
•• Ex (Ez) impacts stator modal frequencies;
•• Ey has no significant influence on
modal frequencies;
Figure 18. Meanwhile, the relative error is given in Table 4, •• Gxy (Gyz) has no obvious influence on the
showing that the frequency errors of P1 - P5 are less than 5% frequencies of symmetric modes (M=1) but is
and amplitude errors are less than 6dB, which indicates that important for the asymmetric modes (M=2);
the accuracy of FEA model is high and can represent the NVH •• Gxz has influence on stator modal frequencies.
performance of physical sample accurately. The modal
damping values used in the correlation are from 0.2% to 4.5%. 2. The contribution of stiffness and mass of hairpin
The amplitude error can be  further reduced by further windings composed of winding lines and coils to
adjusting the damping characteristics if needed. stator assembly should not be ignored, where:
•• The variation of EC or GC is proportional to the
modal frequency variation;
 FIGURE 18   Test data vs correlated simulation data of •• N=2 modes are insensitive to EC or GC change.
stator assembly
An excellent correlation is achieved between FEA model
prediction and test results. The achieved accurate model
provides a foundation for full EDU assembly modeling for
system level vibration and acoustic predictions.

References
1. Wei, Y., Han, X., Lu, L., Wang, H. et al., “Technology
Prospects of Carbon Neutrality-oriented New-energy
Vehicles and Vehicle-grid Interaction,” Journal of Automotive
Engineering 44, no. 4 (2022): 449-464.
 Downloaded from SAE International by Harbin Engineering University, Saturday, July 22, 2023

8 SIMULATION AND VALIDATION OF STATOR MODES OF A HAIRPIN MOTOR

2. Sun, Z., Steyer, G., and Ley, J., “Geartrain Noise

Optimization in an Electrical Drive Unit,” SAE Technical
Acknowledgments
Paper 2015-01-2365 (2015), https://doi.org/10.4271/2015-01- The authors wish to thank AAM for supporting this work and
2365. permission of this publication. The authors also would like to
3. Goetchius, G., “Leading the Charge - The Future of Electric acknowledge the support and contributions of their colleagues
Vehicle Noise Control,” Journal of Sound and Vibration in AAM.
(2011): 5-6.
4. Han, W., Jia, Q., and Qiu, J., “Vibration and Modal Analysis
for Stator of an Induction Motor,” Journal of Vibration and Abbreviations
Shock 31, no. 17 (2012): 91-94.
ICE - Internal combustion engine
5. Guo, X., Zhong, R., Ding, D. et al., “Fast Computation of
Radial Vibration in Switched Reluctance Motors,” IEEE FEA - Finite element analysis
Trans. Ind. Electron. 65, no. 6 (2018): 4588-4598. NVH - Noise vibration harshness
6. Arabi, S., Steyer, G., Sun, Z., and Nyquist, J., “Vibro OEM - Original equipment manufacture
-Acoustic Response Analysis of Electric Motor,” SAE rpm - Revolutions per minute
Technical Paper 2017-01-1850 (2017), https://doi.
org/10.4271/2017-01-1850. DOE - Design of experiment
FRF - Frequency response function
EDU - Electric drive unit
Contact Information ERR - Error
Fulong Zhang is currently Lead NVH Analytical Engineer AMP - Amplitude
at AAM Investment Management (Shanghai) Co., Ltd. on 22F,
Baohua Center, 355 West Guangzhong Road, Jing’an District, F - Frequency
Shanghai 200072, China. He  can be  reached via: Phone:
(86)185 1615 8096, e-mail: Elvis.zhang@aam.com

© 2023 SAE International. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE International.

Positions and opinions advanced in this work are those of the author(s) and not necessarily those of SAE International. Responsibility for the content of the work lies
solely with the author(s).

ISSN 0148-7191