Abstract
The electromagnetism and rotor dynamics have been drawing great research interest due to their significance in the design of the high speed permanent magnet synchronous motor. However, it is challenging but essential to study the rotor supported by air foil bearings with cylindrical permanent magnet. Here the approach is reported to analyze a high speed PMSM rotor at the rating of 28.5 kW and 96,000 rpm. Through the analytic method and the finite element simulation, the expression for the magnetic flux density is obtained. Based on the rotor-bearing system discrete model, the critical speed is calculated by Campbell diagram. It comes to the conclusion that the radius and length of cylindrical permanent magnet is reciprocally related to the magnetic flux density and critical speed values. Finally, results of the PMSM experimental device operating at 96,000 rpm are represented by rotor frequency spectrum and orbit diagram.
Keywords
Introduction
The permanent magnet synchronous motor (PMSM) supported by air foil bearings are now gaining more and more popularity in the rotating machinery industry. In the high speed PMSM, the excitation magnetic field of the rotor permanent magnet interacts with the current density of the stator, and the electromagnetic torque is generated to drive the rotor rotation. The high speed PMSM has many advantages: firstly, the efficiency is high when the high speed PMSM works in the synchronous rotation [1]; secondly, the structure with a coaxial compressor or turbine can reduce the system complexity to be more compact [2]; thirdly, it has high specific energy and low power loss [3].
To analyze the rotor excitation magnetic field in theory, several researchers have worked on the analytic expression for different kinds of the permanent magnet [4–7]. The structure of the rotor and stator is different, so solution procedures are not suitable for the below-mentioned cases. Many other scholars have studied the rotor dynamics research of the high speed rotor supported by air foil bearings [2,8–10]. These references introduced the analysis of air foil bearing and rotor dynamic characteristics, including unbalance vibration response and critical speed. But effects of the permanent magnet on rotor dynamics have not been involved. References [1–3,11,12] present experimental realization in different levels of the power and rotation speed, which can be used as the reference for the PMSM performance experimental device. In this paper, it is necessary to obtain the analytic prediction solution of the excitation magnetic field in the bipolar cylindrical permanent magnet which is magnetized in parallel. Then the influence of permanent magnet dimensions on the magnetic flux density and critical speed can be taken into account. Therefore, this paper carried out further research on the electromagnetism, rotor-bearing system dynamics and performance experiment in the high speed PMSM.
This paper introduced the research organized as follows: first, through the analytic and simulation method, the magnetic flux density of the cylindrical permanent magnet is calculated. Then, the discrete model of the rotor-bearing system is built to analyze the critical speed. Finally, the experiment is carried out to validate the performance of the high speed PMSM. Through the reasonable and effective analysis of the electromagnetism and rotor dynamics, results obtained in this paper have significance to the theoretical study, design and manufacture of the high speed PMSM rotor.
Analytic and simulation analysis
The analytic and simulation analysis is based on a high speed PMSM driven by 28.5 kW at the rated speed of 96,000 rpm. The PMSM rotor has a nonferromagnetic sleeve and a bipolar cylindrical permanent magnet magnetized in parallel. The material of the sleeve and the permanent magnet is nickel-based superalloy and Sm2Co17 alloy, respectively.
Rotor electromagnetic analysis
Magnetic analytic expression of cylindrical permanent magnet
The magnetic field is directly related to the electromagnetic energy transfer between the rotor permanent magnet and stator. So the magnetic field analysis based on the permanent magnet is the important foundation for the high speed PMSM performance analysis and the characteristic calculation.
Figure 1 shows the excitation magnetic field model for the bipolar cylindrical permanent magnet which is magnetized in parallel. R 1, R 2 and R 3 represent the outer radius of the rotor sleeve, the radius of the permanent magnet and the inner radius of the stator, respectively. The regions of the permanent magnet and the gap are defined as P and G, respectively. The gap is regarded as nonmagnetic material.

Analytic model of excitation magnetic field.
In the excitation magnetic field model, B is permanent magnet magnetic flux density; M
R is rotor permanent magnet magnetization; H is magnetic field strength of the permanent magnet. They are in the same direction and the constitutive relation is written as
The analytic expression of the excitation magnetic field is defined in the polar coordinate system, which have polar coordinate parameters r and θ. For a cylindrical magnet that is magnetized in parallel, the magnetization vector
In the permanent magnet excitation magnetic field, the magnetic vector potential
In the static or steady-state magnetic field, the eddy current can be ignored inside the permanent magnet, i.e., 𝛻 ×
The separated variable method can be used when the periodic equation is assumed via A(r, θ) = A(r, θ + 2π). Then the general solution can be given by
In the radial and circumference magnetic flux density B
r
and B
θ, there are
For the excitation magnetic field in Fig. 1, there are the following boundary conditions:
Substituting Eqs (3) and (10) into Eq. (7), all of undetermined coefficients can be solved. Thus, based on Eqs (4) and (7), the analytic expression of the magnetic flux density can be written as
According to the finite element method (FEM), a finite element model of the stator and rotor is constructed in ANSYS Maxwell software and all parameters conform to the high speed PMSM driven by 28.5 kW. In order to improve the calculation accuracy, the element density in the gap region is largely raised as shown in Fig. 2. Totally, there are 24,882 elements in the finite element model. Through the electromagnetic simulation, Fig. 2 shows the distribution nephogram of the magnetic flux density in the stator and rotor. In the gap region, it can be seen that with a decrease in the distance away from the permanent magnet outer circle, the value of the magnetic flux density increases.

Magnetic flux density distribution nephogram (28.5 kW PMSM).
The simulation result can be compared with the analytic solution to verify Eq. (11). In the middle of gap region, the radius is 15.55 mm as shown in Fig. 2. The magnetic flux density results (in radius r = 15.55 mm) of the analytic solution and simulation is illustrated in Fig. 3. It can be seen that curves comprise analytic and simulation results of the radial component B r and the circumference component B θ. The distribution of the magnetic flux density is close to the sine function. The comparative analysis of analytic values and FEM simulation results reveals their close fit for both radial and circumference component. This implies that the analytic expression is suitable for the magnetic flux density of the cylindrical permanent magnet parallel magnetized. So a feasibility calculation method can be applied for the magnetic field of the PMSM rotor.

Results of finite element simulation and analytic solution.
According to Eq. (11), it can be suggested that the magnetic flux density of the permanent magnet can be determined by the permanent magnet radius R
2 when the stator inner radius R
3 is a given value. So it is necessary to analyze the relationship between the magnetic flux density and the permanent magnet radius. The radial and circumferential maximum amplitude of the magnetic flux density can be written as
From the calculation results, the relationship between R 2 and B rmax, B θmax is shown as Fig. 4. It indicates that with an increase in R 2, the radial maximum amplitude B rmax increases and the circumferential maximum amplitude B θmax decreases. Thus, permanent magnet radius affects the maximum amplitude of magnetic flux density, and the numerical relationship is expressed by Eq. (12).

Relationship between radius and maximum amplitude of magnetic flux density.
Above results covered only the magnetic flux density generated by permanent magnets. While, the torque comes from the interaction between the magnetic field of rotor permanent magnets and the current surface density of stator currents in the air gap. Mechanical vibrations in a PMSM can be caused by torque ripples due to many factors. So the proposed approach with the magnetic field calculation may be prospectively extended and used for mechanical vibration in the follow-up studies.
Rotor discrete model and critical speed
The structure of the rotor supported by radial air foil bearings (I and II) is shown in Fig. 5. The rotor is double-span combined and connected by a shaft coupling. The permanent magnet diameter D is 31.8 mm and the length L is 70 mm.

Discrete model of the rotor-bearing system (mm).
The dynamics equation of the rotor-bearing system is given by:
FEM is a commonly used method in rotor dynamics analysis. A finite element model of the rotor-bearing system is constructed via ANSYS as shown in Fig. 6. Shafts can be modeled using Timoshenko beam element, and the thrust disc is treated as a rigid body. Air foil Bearings can be simplified as COMBIN14 element and equivalent to the form of stiffness and damping in the linear range. In the finite element model, the bearing supported stiffness is 1 × 106 N/m and Coriolis force is added to allow for the effect of spin softening.

Finite element model.
By solving the eigenvalue of Eq. (12), damp algorithm is used to determine eigenvalues under the different speed. The Campbell diagram of the rotor whirling frequency is obtained through the imaginary part of eigenvalues. As shown in Fig. 7, the horizontal axis represents the rotor speed range which is from 0 to 250,000 rpm, and the vertical axis stands for the rotor whirling frequency calculated through FEM. The rotor forward whirl and backward whirl are shown by different symbol lines. The crossing points of whirl lines and a spin line (slope = 1) represent the critical speed.

Campbell diagram of the rotor.
The Campbell diagram shows that due to the gyroscopic effect, with an increase in the rotor speed, the backward whirling frequency reduces and the forward whirling frequency increases. The forward whirling frequency need to be considered in the rotor critical speed. Therefore, the critical speed of forward whirling from 1st to 5th are 26,182 rpm, 28,693 rpm, 35,307 rpm, 84,225 rpm and 203,250 rpm, respectively. Whirling orbit plots of the vibration mode are shown in Fig. 8. Operating of the rotor supported by air foil bearings is not affected by the critical speed of parallel and conical whirling, but the critical speed of the bending whirling should be avoided [2]. As shown in Fig. 8, the whirling orbits from 1st to 4th are parallel and conical whirling within the rated speed range (0–96,000 rpm). The rated speed is far away from the bending whirling critical speed, so the performance of rotor dynamics has sufficient safety margin. According to the bending whirling orbits, the bending area mainly appears in the part of the rotor sleeve and the permanent magnet.

Whirling orbit plots from mode 1 to mode 5.
From the above analysis results, it can be seen that the bending whirling primarily appears in the rotor part including the permanent magnet. High speed rotor dynamics can be influenced by the complex interaction between the inertial properties and the critical speed [16]. So different dimension parameters of the permanent magnet affect the dynamics performance of the whole rotor-bearing system.
The applied FEM made it possible to construct rotor-bearing models with various diameter and length of the permanent magnet. For each of the following cases, diameter D is chosen from 21.8 mm to 41.8 mm, and length L is chosen from 70 mm to 140 mm. Influence of the permanent magnet on the critical speed is calculated through the finite element simulation. As shown in Fig. 9, the respective curves of the forward whirling critical speed corresponding to different modes are calculated in the spin velocity ranging from 0 to 250,000 rpm.
Figure 9 properly reflects the change trend of the forward whirling critical speed along with different diameters and lengths. With the same length (L = 70 mm) and an increase in the diameter, values of the critical speed is reduced gradually; meanwhile, with the same diameter (D = 31.8 mm) and an increase in the length, values of the critical speed increases gradually. In conclusion, the case results obtained show that the variation of the forward whirling critical speed can be found through different diameters and lengths of the permanent magnet.

Influence of diameter and length on the forward whirling critical speed.
Based on two sections of the case study, from the comprehensive analysis of calculation results it can be seen that different radius of the cylindrical permanent magnet affects the magnetic flux density; meanwhile, different diameter and length affect the rotor critical speed. Given this, dimension parameters of the permanent magnet is reciprocally related to the magnetic flux density and critical speed.
In order to identify the performance in electromagnetism and rotor dynamics, a PMSM experimental device has already been made and illustrated in Fig. 10. In the experimental setup, there are a stator and a rotor-bearing system, including two sets of radial air foil bearings a shaft coupling, a thrust disc and shaft sleeves as show in Fig. 10(b–d). The bipolar cylindrical permanent magnet is press-fit in the rotor sleeve. The stator with the Wye distributed winding is assembled in the PMSM as shown in Fig. 10(a). Two high accuracy displacement sensors are arranged in the horizontal and vertical direction to measure the rotor orbit. The experiment process and data acquisition system are shown in Fig. 10(e,f).
When the high speed PMSM experiment was carried out at the speed of 96,000 rpm, all performances of the rotor-bearing system were verified successfully. As shown in Fig. 11(a), the rotor frequency spectrum indicates the horizontal vibration values at 96,000 rpm, and the maximum amplitude is 2.337 μm corresponding to the frequency of 1,600 Hz. Figure 11(b) shows the rotor orbit diagram obtained from two displacement sensors in the horizontal and vertical position at 96,000 rpm. The vertical displacement Y and the horizontal displacement X are both within the limits of 11 μm. The experimental results obtained at the rated speed suggest that the rotor supported by air foil bearings has the high stability and little vibration. It is anticipated that above results are useful for the design and analysis of the rotor electromagnetism and dynamics.

PMSM experimental device.

Rotor experimental results obtained at 96,000 rpm.
In this paper, several analysis methods have been applied to deal with the high speed PMSM rotor at the rating of 28.5 kW and 96,000 rpm. Firstly, the analytic expression for the magnetic flux density generated by parallely magnetized cylindrical permanent magnet is given. The results obtained via analytic solution and FEM simulation are found to be approximate. Secondly, the rotor-bearing system discrete model is built and rotor dynamics analysis is carried out. The bending whirling critical speed is higher than the rated speed as proven in the Campbell diagram. Thirdly, based on results of the case study, dimension parameters of the permanent magnet are reciprocally related to the magnetic flux density and rotor critical speed. This result is instrumental in the integration design of the magnetic field and mechanical structure. Lastly, a PMSM experimental device has been constructed to identify the performance in electromagnetism and rotor dynamics. Through the realization of PMSM operation experiment at 96,000 rpm, rotor frequency spectrum and orbit diagram are both obtained. As a result, the performances of the rotor-bearing system are verified well.
Footnotes
Acknowledgements
The work is supported by Chinese National Natural Science Foundation (51805407), Natural Science Foundation in Shaanxi Province of China (Grant No. 2018JQ5148) and Key Research and Development Project in Shaanxi Province of China (Grant No. 2017ZDXM-GY-054).
