Permanent magnet synchronous motor cogging torque suppression method based on fxlms
By generating a noise reference signal through an FxLMS hybrid filtering strategy and combining it with an FxLMS filter for cogging torque cancellation, the problem of complex hardware and system model dependence in existing technologies is solved, and a simple and efficient cogging torque suppression for permanent magnet synchronous motors is achieved, which is applicable to a variety of motor systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THE UNIV OF NOTTINGHAM NINGBO CHINA
- Filing Date
- 2026-02-04
- Publication Date
- 2026-06-02
AI Technical Summary
Existing methods for suppressing cogging torque in permanent magnet synchronous motors are complex, requiring expensive hardware or complex system models, and are not suitable for rapid deployment and field development, making it difficult to achieve effective suppression under sensorless conditions.
An FxLMS hybrid filtering strategy is adopted, which generates a noise reference signal by using a static low-pass and high-pass second-order Butterworth filter, and combines it with the FxLMS filter to cancel cogging torque and generate a q-axis current cancellation waveform to achieve closed-loop control.
It achieves efficient cogging torque suppression without the need for external sensors and system models, and is applicable to motor systems of different speeds, loads and sizes, simplifying the development and deployment process and reducing hardware and computational burden.
Smart Images
Figure CN122137292A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of permanent magnet synchronous motor, in particular to a permanent magnet synchronous motor cogging torque suppression method based on FxLMS. BACKGROUND
[0002] Permanent magnet synchronous motor (PMSM) based systems must cope with the inherent cogging torque problem, which is caused by the interaction between the magnetic fluxes inside the motor. This undesired characteristic introduces ripples in the output torque and speed, so various methods have been proposed to suppress this disturbance, including compensation through motor structure design and through motor control strategy. However, these motor control methods are complex in development and implementation, and may not be suitable for low-cost motor systems with limited control hardware performance.
[0003] For harmonic injection and cancellation, an accurate reference "noise" signal is ideally required to generate the accurate cancellation waveform. This is the working principle of active noise cancellation (ANC) systems. The concept was first proposed in 1936, and the basic idea is to use an error microphone and a loudspeaker to generate a cancellation effect. In this system, the cancellation waveform is actively adjusted to suppress the constantly changing error signal. For an ideal motor system, the error "microphone" can be realized by a torque sensor to directly obtain the torque ripple and distortion; however, most actual motor systems are not equipped with torque sensors. Therefore, the torque output can only be derived and predicted based on internal information of the control system, such as current, rotor position angle, and speed parameters. As can be seen, a practical cogging torque suppression strategy should only rely on signal processing and internal data to generate a reference "noise" signal, thereby avoiding the addition of extra sensors or hardware modifications, so that all improvements and implementations can be completed through software or code.
[0004] Existing suppression methods typically rely on prior information about the motor system, including transfer functions and system models. However, this is impractical in real-world applications, as end-users or system integrators usually lack the expertise to analyze motor control systems and should not be required to perform detailed flux analysis to derive cogging torque disturbances. For example, the Active Disturbance Rejection Controller (ADRC) utilizes multiple extended state observers to detect and respond to disturbances, including cogging torque, but its optimal performance depends on accurate and detailed system models, making it unsuitable for rapid deployment or field development. Furthermore, machine learning and artificial intelligence-based methods have been proposed, such as iterative learning methods or Q-learning algorithms, but these also require prior training and preparation, and typically necessitate high-performance AI hardware for the training process.
[0005] A "general-purpose" solution would help simplify the development and integration process, achieving cogging torque suppression without requiring expensive hardware, complex preparation, or knowledge of transfer functions and system models. This solution should not increase the hardware and computational burden of existing motor systems, nor introduce additional sensors for signal processing; all processing should be performed entirely within a software control environment. Furthermore, the solution should have a streamlined development and deployment process, and its control modules should be compatible with existing field-oriented control (FOC) systems and related programming environments. Finally, system parameters (such as cutoff frequency) should be minimized to further reduce deployment complexity. Summary of the Invention
[0006] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a FxLMS-based method for suppressing cogging torque of permanent magnet synchronous motors with a simple hardware structure and requiring only a few system tuning parameters to achieve maximum suppression performance.
[0007] The technical solution of this invention is to provide a method for suppressing cogging torque in a permanent magnet synchronous motor based on FxLMS, comprising the following steps: S1, the unfiltered torque demand waveform output by the PI speed loop; S2. The torque demand is processed through a static low-pass second-order Butterworth filter to obtain the filtered torque demand waveform. S3 and torque demand are simultaneously processed by a static high-pass second-order Butterworth filter to obtain a waveform containing only the cogging torque component, thereby forming a noise reference signal; S4. Input the noise reference signal into the modified FxLMS filter to obtain the anti-noise signal; S5. The anti-noise signal is input to the summing module and superimposed with the filtered torque demand waveform obtained in step S2 to generate a combined torque demand for cogging torque cancellation. S6. Input the combined torque requirement to the gain module to generate the q-axis current reference value, and input the reference value to the PI current loop; S7. The combined torque demand is simultaneously fed back to the PI speed loop and combined with the actual speed error of the motor to generate a new torque demand value. The system then returns to step S1 to form a closed-loop operation.
[0008] Preferably, the unfiltered torque demand waveform described in step S1 includes a cogging torque component.
[0009] Preferably, the filtered torque demand waveform described in step S2 is smooth and basically free of cogging torque components.
[0010] Preferably, the average value of the noise reference signal mentioned in step S3 should be zero.
[0011] Preferably, the anti-noise signal described in step S4 is used to generate a compensation waveform for destructive interference.
[0012] Preferably, the reference value mentioned in step S6 is based on the offset injection performed in the aforementioned stages to modulate the torque output of the system.
[0013] With the above structure, the FxLMS-based permanent magnet synchronous motor cogging torque suppression method of the present invention has the following advantages compared with the prior art: The FxLMS hybrid filtering strategy employed in this invention can generate noise reference signals without the need for external sensors and without relying on the motor system model or specific parameter information. The system only needs to define the cutoff frequency of the cogging torque, which is used to construct a static low-pass second-order Butterworth filter and a static high-pass second-order Butterworth filter. By using the aforementioned static digital filters, a wide range of vibration frequencies can be suppressed without requiring a precise system model. The FxLMS algorithm is initialized with a virtual master path transfer function and adaptively corrects this transfer function during system operation, ultimately converging to the optimal system path to achieve maximum suppression performance.
[0014] Furthermore, this strategy proposes a novel cogging torque cancellation injection method. Specifically, the FxLMS hybrid filtering strategy transforms the cancellation waveform in traditional acoustic active noise cancellation into a q-axis current cancellation waveform that can be directly used for motor control, thereby introducing active noise cancellation technology into the field of motor control. In this analogy, the low-frequency desired torque component can be regarded as a low-frequency signal, while the cogging torque component can be regarded as a high-frequency signal. Classifying these components as acoustic signals helps in the design and implementation of the suppression system and makes the FxLMS hybrid filtering strategy more widely applicable to motor systems of different speeds, loads, and sizes.
[0015] Furthermore, the FxLMS-based cogging torque suppression method for permanent magnet synchronous motors of this invention is platform-independent. The FxLMS system is implemented in C language code and can be deployed in existing simulation or control environments, such as MATLAB, SIMULINK, PLECS, and TI control platforms. This method achieves active cancellation and current injection while requiring very few system tuning parameters; specifically, only the cutoff frequencies of the high-pass and low-pass static digital filters need to be set. Attached Figure Description
[0016] Figure 1 This is a comparison diagram of the reference torques of a static second-order filter and an FxLMS hybrid filter.
[0017] Figure 2 This is a comparison of the simulated actual torque of the static second-order filter and the FxLMS hybrid filter.
[0018] Figure 3 This is a comparison of the Fast Fourier Transform of a static second-order filter and an FxLMS hybrid filter.
[0019] Figure 4 The graph shows a comparison between the reference torque and the simulated actual torque of the original system, the static second-order filter, and the FxLMS hybrid filter.
[0020] Figure 5 The image shows a comparison of the Fast Fourier Transform of the original system, the static second-order filter, and the FxLMS hybrid filter.
[0021] Figure 6 The simulation speed ripple comparison diagram shows the original system, the static second-order filter, and the FxLMS hybrid filter.
[0022] Figure 7 The simulated rotational speed response comparison diagrams are shown for the original system, the static second-order filter, and the FxLMS hybrid filter.
[0023] Figure 8 This is a magnified view of the simulated rotational speed response.
[0024] Figure 9 This is a flowchart of the hybrid filtering process of the present invention. Detailed Implementation
[0025] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.
[0026] Example: like Figures 1-9 As shown, this invention discloses a method for suppressing cogging torque in a permanent magnet synchronous motor based on FxLMS, comprising the following steps: S1, the PI speed loop outputs an unfiltered torque demand waveform, which contains a cogging torque component. S2. The torque demand is processed by a static low-pass second-order (biquad) Butterworth filter to obtain the filtered torque demand waveform, which is smooth and basically free of cogging torque components. S3 and torque demand are simultaneously processed by a static high-pass second-order Butterworth filter to obtain a waveform containing only the cogging torque component, thereby forming a noise reference signal. The average value of the noise reference signal should be zero. S4. The noise reference signal is input into the modified FxLMS filter to obtain an anti-noise signal, which is used to generate a compensation waveform for destructive interference. S5. The anti-noise signal is input to the summing module and superimposed with the filtered torque demand waveform obtained in step S2 to generate a combined torque demand for cogging torque cancellation. S6. Input the combined torque demand to the gain module to generate a q-axis current reference value, and input the reference value to the PI current loop. The reference value modulates the torque output of the system based on the cancellation injection performed in the previous stages. S7. The combined torque demand is simultaneously fed back to the PI speed loop and combined with the actual speed error of the motor to generate a new torque demand value. The system then returns to step S1 to form a closed-loop operation.
[0027] To compare the effectiveness of the FxLMS hybrid filtering strategy, it is compared with a static filtering strategy; the static filtering strategy uses only a low-pass second-order Butterworth digital filter to achieve cogging torque suppression.
[0028] 1. Steady-state – Simulation Torque Analysis like Figure 1 and Figure 2As shown, under steady-state conditions, compared to the case using only a static second-order Butterworth filter, although the simulated actual torque suppression effect is not as significant as that in the reference torque, the FxLMS hybrid filtering strategy still exhibits higher suppression capability in reducing torque waveform ripple caused by cogging torque. The resulting improvements are listed in Table 1 below:
[0029] Table 1 Table 1 compares the torque distortion improvement of the simulated system between the static second-order filter and the FxLMS hybrid filter (static low-pass second-order Butterworth filter and static high-pass second-order Butterworth filter).
[0030] As shown in Table 1, under steady-state conditions, the peak values of the torque demand and the actual torque obtained from the simulation were improved by 62.3% and 39.4%, respectively, thus verifying the superior performance of the FxLMS hybrid filter in suppressing cogging torque.
[0031] like Figure 3 As shown, the Fast Fourier Transform (FFT) analysis results indicate that the FxLMS hybrid filter is more effective than the static filtering strategy in attenuating the cogging torque components (approximately 470 Hz and 580 Hz) and low-frequency components. However, two additional frequency components remain at approximately 430 Hz and 520 Hz, which are residual components of the anti-noise injection waveform. This suggests that the generated canceled waveform is not entirely accurate, although these residual components have a significantly smaller impact compared to the two main cogging torque components.
[0032] like Figure 4 As shown, it presents the overall comparison results between the simulated output torque and the unfiltered original system; the data is organized as shown in Table 2 below.
[0033] Table 2 Table 2 is a comparison table of torque distortion improvement in the simulation system of unfiltered, static second-order filter and FxLMS hybrid filter.
[0034] Compared to the unfiltered state, the FxLMS hybrid filter improves the reduction of reference torque ripple by 19.1% and the reduction of simulated actual torque output ripple by 15%.
[0035] like Figure 5As shown, FFT analysis results for the three operating modes indicate that the FxLMS hybrid filtering strategy still performs better in terms of suppression. Compared to the static second-order Butterworth filter, this method introduces two additional residual frequency components at approximately 430 Hz and 520 Hz, but their amplitudes are still smaller than the unsuppressed cogging torque components at the same frequencies, and they are still attenuated overall. Slight distortion still exists in the low-frequency components, but this effect is significantly less than the distortion introduced by the static second-order Butterworth filter.
[0036] Therefore, it can be concluded that, under steady-state operating conditions, the FxLMS hybrid filter has superior performance in suppressing cogging torque compared to the static second-order Butterworth filter.
[0037] 2. Simulated rotational speed ripple analysis like Figure 6 As shown, the target speed was again set to 50 rad / s, and the simulated speed ripple was collected when the motor reached steady-state operation. Table 3 below shows a comparison of the system speed ripple improvements with unfiltered, static second-order filter, and FxLMS hybrid filter.
[0038] Table 3 Simulation results show that the rotational speed ripple is reduced by 40.5%, which is less than the 62.0% improvement achieved by the static filter. Although the FxLMS hybrid filter has better performance in cogging torque suppression, its performance in rotational speed ripple suppression is 20.5% lower than that of the static filter.
[0039] Nevertheless, it can be concluded that the FxLMS hybrid filter has a significant suppression effect on speed ripple under steady-state conditions, although its suppression level is slightly lower than that of the static filtering strategy.
[0040] 3. Variable speed response simulation analysis like Figure 7 and Figure 8 As shown, compared with the static filtering strategy, the FxLMS hybrid filter exhibits smaller overshoot and better acceleration ramp response during speed change, but its convergence time to reach the target speed and enter steady state is relatively longer. Overall, the simulated speed response can be considered similar to that of the static second-order Butterworth filter, although both filtering strategies exhibit larger overshoot compared to the unfiltered state.
[0041] The FxLMS-based cogging torque suppression method for permanent magnet synchronous motors of this invention employs a hybrid FxLMS filtering strategy that generates a noise reference signal without the need for external sensors and without relying on the motor system model or specific parameter information. The system only needs to define the cutoff frequency of the cogging torque, which is used to construct a static low-pass second-order Butterworth filter and a static high-pass second-order Butterworth digital filter. By using the aforementioned static digital filters, a wide range of vibration frequencies can be suppressed without requiring a precise system model. The FxLMS algorithm is initialized with a virtual master path transfer function and adaptively corrects this transfer function during system operation, ultimately converging to the optimal system path to achieve maximum suppression performance.
[0042] Furthermore, this strategy proposes a novel cogging torque cancellation injection method. Specifically, the FxLMS hybrid filtering strategy transforms the cancellation waveform in traditional acoustic active noise cancellation into a q-axis current cancellation waveform that can be directly used for motor control, thereby introducing active noise cancellation technology into the field of motor control. In this analogy, the low-frequency desired torque component can be regarded as a low-frequency signal, while the cogging torque component can be regarded as a high-frequency signal. Classifying these components as acoustic signals helps in the design and implementation of the suppression system and makes the FxLMS hybrid filtering strategy more widely applicable to motor systems of different speeds, loads, and sizes.
[0043] Furthermore, the FxLMS-based cogging torque suppression method for permanent magnet synchronous motors of this invention is platform-independent. The FxLMS system is implemented in C language code and can be deployed in existing simulation or control environments, such as MATLAB, SIMULINK, PLECS, and TI control platforms. This method achieves active cancellation and current injection while requiring very few system tuning parameters; specifically, only the cutoff frequencies of the high-pass and low-pass static digital filters need to be set.
[0044] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for suppressing cogging torque in a permanent magnet synchronous motor based on FxLMS, characterized in that: Includes the following steps, S1, the unfiltered torque demand waveform output by the PI speed loop; S2. The torque demand is processed through a static low-pass second-order Butterworth filter to obtain the filtered torque demand waveform. S3 and torque demand are simultaneously processed by a static high-pass second-order Butterworth filter to obtain a waveform containing only the cogging torque component, thereby forming a noise reference signal; S4. Input the noise reference signal into the modified FxLMS filter to obtain the anti-noise signal; S5. Input the anti-noise signal to the summing module and superimpose it with the filtered torque demand waveform obtained in step S2 to generate a combined torque demand for cogging torque cancellation. S6. Input the combined torque requirement to the gain module to generate the q-axis current reference value, and input the reference value to the PI current loop; S7. The combined torque demand is simultaneously fed back to the PI speed loop and combined with the actual speed error of the motor to generate a new torque demand value. The system then returns to step S1 to form a closed-loop operation.
2. The method for suppressing cogging torque of a permanent magnet synchronous motor based on FxLMS according to claim 1, characterized in that: The unfiltered torque demand waveform described in step S1 contains a cogging torque component.
3. The method for suppressing cogging torque of a permanent magnet synchronous motor based on FxLMS according to claim 2, characterized in that: The filtered torque demand waveform described in step S2 is smooth and basically free of cogging torque components.
4. The method for suppressing cogging torque of a permanent magnet synchronous motor based on FxLMS according to claim 3, characterized in that: The average value of the noise reference signal mentioned in step S3 should be zero.
5. The method for suppressing cogging torque of a permanent magnet synchronous motor based on FxLMS according to claim 1, characterized in that: The anti-noise signal mentioned in step S4 is used to generate a compensation waveform for destructive interference.
6. The method for suppressing cogging torque of a permanent magnet synchronous motor based on FxLMS according to claim 1, characterized in that: The reference value mentioned in step S6 modulates the torque output of the system based on the offset injections performed in the preceding stages.