Permanent magnet synchronous motor fault detection method based on self-learning boundary reconstruction

Abstract

The invention discloses a permanent magnet synchronous motor fault detection method based on self-learning boundary reconstruction. According to the method, a fault estimator is constructed based on system parameters and a control law. Considering that an observed permanent magnet synchronous motor system is an actual system and is inevitably affected by noise, rotating speeds of the observed permanent magnet synchronous motor and a fault estimator are extracted through a sliding mode estimator and a phase-locked loop, and random noise is added into rotating speed information of the fault estimator to restore the actual situation. Self-learning boundary reconstruction is carried out on the rotating speed information of the fault estimator to form upper and lower bounds of the fault estimator; the fault detection of the permanent magnet synchronous motor is realized by checking whether the rotating speed curve of the observed permanent magnet synchronous motor is between the upper and lower bounds of the fault observer. Due to a self-learning mechanism, the detection method can flexibly adjust the accuracy degree of fault detection by adjusting the number of iterations according to actual requirements. Compared with a traditional detection method, the method has the advantages that the method participates in a simple application environment, and the method has good engineering value.

Description

Technical Field

[0001] The invention relates to a fault detection method for a permanent magnet synchronous motor, and in particular to a fault detection method for a permanent magnet synchronous motor based on self-learning boundary reconstruction. Background Art

[0002] Permanent magnet synchronous motors have significant advantages such as wide speed range, simple structure, high efficiency, high power density, high output torque, fast speed regulation response, and small size and light weight. They have been widely used in many fields such as industrial production, rail transportation, electric vehicles, aerospace, etc.

[0003] Due to the manufacturing defects of the permanent magnet synchronous motor itself and the effects of wear, deformation and corrosion during operation, its performance will gradually deteriorate as the performance of the components deteriorates, causing faults such as uniform and local demagnetization and bearing failure. In severe cases, it may even cause downtime accidents, causing serious economic losses. Therefore, it is very important to diagnose the faults of permanent magnet synchronous motors in a timely and accurate manner. At present, the diagnostic technology for large motors is relatively mature. The diagnostic system for large motors uses multiple sensors to monitor motor faults, including electrical signals, temperature signals, noise and vibration. In the field of small and medium power, due to cost factors, a single sensor diagnostic system is more often used. The limitations of a single signal affect the diagnostic accuracy to a certain extent, so it is necessary to study and adopt advanced technologies and methods to deal with it. Summary of the invention

[0004] The purpose of the present invention is to propose a permanent magnet synchronous motor fault detection method based on self-learning boundary reconstruction, which can effectively detect the faults of the permanent magnet synchronous motor.

[0005] The specific technical solution of the present invention is as follows: A permanent magnet synchronous motor fault detection method based on self-learning boundary reconstruction comprises the following steps:

[0006] The observed permanent magnet synchronous motor voltage equation is as follows:

[0007] Where L d , L q Represent the d-axis and q-axis inductances, ω e represents the angular velocity, p represents the differential operator, u a ,u β Represent the voltage of the stator α and β axes respectively, i α ,i β They represent the currents of the stator α and β axes, respectively, and E a , E β denote the back electromotive force of α and β axis respectively;

[0008] The back-EMF-based super-helical sliding mode estimator is further shown as follows:

[0009] In the formula Indicates i a The estimated value of Indicates i β Estimated value, Indicates i a The estimation error of Indicates i β The estimation error, γ α Indicates E α The estimated value of γ β Indicates E β The estimated value of , k1 and k2 represent the sliding mode estimator gains, express The absolute value of express The absolute value of express The symbol function of express The symbol function of Express To perform integration, Express Perform integration;

[0010] Further, based on the speed extraction of the phase-locked loop, the specific steps are as follows:

[0011] The observed back electromotive force equation of the permanent magnet synchronous motor is as follows:

[0012] Where E α , E β are the back electromotive force of the α and β axes, ψ f represents the flux linkage of the permanent magnet, θ e Indicates the rotor position;

[0013] make Then get By performing PI adjustment on ΔE, the position information of the observed permanent magnet synchronous motor can be obtained;

[0014] Add random noise to the speed information of the fault estimator. The specific steps are as follows: Probability distribution function of noise d:

[0015] Where d i represents the i-th element of vector d, f(d i ) means di The probability distribution function of

[0016] Further divide each element of d by the standard deviation σ of d. The specific steps are as follows:

[0017] In the formula represents the mean of d, iter represents the number of iterations, d i，1 It means that after the first step, the new d = d i，1 ;

[0018] Further subtract the mean of d from each element of d. The specific steps are as follows:

[0019] In the formula It represents the new mean value of d after the first step, d i，2 It means that after the second step, the new d = d i，2 ;

[0020] Each element of d is further linearly transformed. The specific steps are as follows: d i，3 =0.5+0.1×d i,2

[0021] Where d i，3 It means that after the third step, the new d = d i，3 ;

[0022] The discrete dynamic model of the permanent magnet synchronous motor fault estimator is as follows:

[0023] Where x(k) represents the system state, u(k) represents the system input, y(k) represents the system output, and d(k) represents the noise;

[0024] A self-learning boundary reconstruction method is designed to reconstruct the system state. Multiple self-learning methods are used to continuously correct the estimated value to improve the estimation accuracy. The specific steps are as follows: Iteration 0: Iteration i: In the formula and They represent the estimated values ​​of the 0th and i-th system states respectively, and Respectively represent the estimated values ​​of the 0th and i-th disturbances, L, R, G are the observer gains to be designed;

[0025] The reconstruction algorithm of the present invention can realize the safety margin required for fault detection, and the process is as follows:

[0026] C001: Select a composite energy function of the following form:

[0027] C002: represents the state observation error of the system, represents the estimated error of the disturbance, P represents the Lyapunov variable matrix;

[0028] C003: Calculate the V between the k+1th sampling time and the kth sampling time [i] (k) and consider H ∞ Performance, including: In the formula γ represents H ∞ Performance level indicators, Respectively represent the estimated values ​​of gain matrices A, B, and D, Represent the transformed rank matrices of A, B, and D respectively;

[0029] C004: Further consideration According to C003, the following formula can be obtained:

[0030] C005: Therefore, from C004, we can get that when Θ<0,

[0031] C006: According to C005, the permanent magnet synchronous motor system is asymptotically stable under the event-triggered control scheme designed by the present invention; Obtain the normal operation data of the permanent magnet synchronous motor and analyze the reachable set of the observed error signal of each iterative batch: In the formula e [i] (k) and Respectively represent the upper and lower bounds of the observation error, Ω [i] (0) represents the reachable set of the initial value of the observation error, Box(·) represents the interval boundary, represents the Minkowski sum, and further calculates the average residual signal interval, thereby generating the interval of the permanent magnet synchronous motor system state during normal operation; In the formula x [i] (k) and They represent the upper and lower bounds of the permanent magnet synchronous motor system state respectively; therefore, the safety boundary required for fault detection can be reconstructed from the upper and lower bounds of the state reachability. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0031] Figure 2 The estimation interval width at different iteration times of the method proposed in the present invention is used as an example;

[0032] Figure 3 This is a diagram showing the effect of magnetic leakage fault detection using the method proposed in the present invention;

[0033] Figure 4 This is a diagram showing the effect of mechanical fault detection using the method proposed in the present invention; DETAILED DESCRIPTION

[0034] The present invention is further explained below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0035] like Figure 1 As shown, a permanent magnet synchronous motor fault detection method based on self-learning boundary reconstruction includes the following steps:

[0036] Step 1: extracting speed information from the fault estimator;

[0037] Step 2: adding noise to the speed information of the fault estimator;

[0038] Step 3: Reconstruct the speed information of the fault estimator through self-learning boundary;

[0039] Step 4: extracting the speed information of the observed permanent magnet synchronous motor;

[0040] Step 5: Utilize the fault estimator to perform fault detection on the observed permanent magnet synchronous machine;

[0041] Step 6: An embodiment of the present invention is described below: Considering the permanent magnet synchronous motor fault detection method based on self-learning boundary reconstruction, the observed permanent magnet synchronous motor is selected, and its parameters are: P r =4,L dr =0.0085H,L qr =0.0085H, R sr =2.875Ω,ψ fr =0.175Wb,J r=0.003Kg·m 2 ,ω T =1000rpm.

[0042] Where P r Indicates the number of motor pole pairs, L dr , L qr Respectively represent the stator d-axis and q-axis inductance, R sr is the stator resistance, ψ fr represents the permanent magnet flux, J r Represents the motor inertia, ω T Indicates the target speed;

[0043] Figure 1 is a method flow chart of an embodiment of the present invention; applying the proposed method, Figure 2 represents the estimated interval width of the method proposed in the present invention at different iteration times, Figure 3 It means that the observed permanent magnet synchronous motor has a sudden load fault at t=1s, which causes the speed to drop and exceeds the safety limit. Figure 3 It indicates that the observed permanent magnet synchronous motor has a leakage fault at t=1s, causing the speed to drop and exceed the safety limit. From these two figures, it can be seen that the proposed method has a good application effect in the permanent magnet synchronous motor fault detection system, and can obtain a satisfactory detection performance. Moreover, the more iterations, the faster the fault can be detected.

[0044] References

[0045] [1] W. Tang, Z. Wang, Y. Wang, T. and Y.Shen, “Interval estimation methods for discrete-time linear time-invariant systems,” IEEETrans.Autom.Control, vol.64, no.11, pp.4717-4724, Nov.2019., doi: 10.1109 / TAC.2019.2902673.

[0046] [2] J. Li, Z. Wang, T. and Y.Shen,“Unknown input observer design forlinear parameter-varying systems in a bounded error context,”IEEETrans.Autom.Control,vol.66,no.9,pp.4246-4251,Sep.2020.doi:10.1109 / TAC.2020.3027646.

Claims

1. A permanent magnet synchronous motor fault detection method based on self-learning boundary reconstruction, characterized in that: The following steps are involved: Self-learning is used to reconstruct the fault with added noise, and the speed of the observed permanent magnet synchronous motor is extracted based on the sliding mode estimator and the phase-locked loop. The voltage equation of the observed permanent magnet synchronous motor is as follows: Where L d , L q Respectively represent the d-axis and q-axis inductance, ω e represents angular velocity, p represents the differential operator, u α ,u β Represent the voltage of the stator α and β axes respectively, i α ,i β They represent the currents of the stator α and β axes, respectively, and E α , E β denote the back electromotive force of α and β axis respectively; The back-EMF based super-helical sliding mode estimator is shown below: In the formula Indicates i α The estimated value of Indicates i β Estimated value, Indicates i α The estimation error of Indicates i β The estimation error, γ ɑ Indicates E α The estimated value of γ β Indicates E β The estimated value of , k1 and k2 represent the sliding mode estimator gains, express The absolute value of express The absolute value of express The symbol function of express The symbol function of Express To perform integration, Express Perform integration; The specific steps of speed extraction based on phase-locked loop are as follows: The observed back electromotive force equation of the permanent magnet synchronous motor is as follows: Where E α , E β are the back electromotive force of the α and β axes, ψ f represents the flux linkage of the permanent magnet, θ e Indicates the rotor position; make Then get By performing PI adjustment on ΔE, the position information of the observed permanent magnet synchronous motor can be obtained; Add random noise to the speed information of the fault estimator. The specific steps are as follows: Probability distribution function of noise d: Where d i represents the i-th element of vector d, f(d i ) means d i The probability distribution function of Divide each element of d by the standard deviation σ of d: In the formula represents the mean of d, iter represents the number of iterations, d i,1 It means that after the first step, the new d = d i，1 ; Subtract the mean of d from each element of d: In the formula It represents the new mean value of d after the first step, d i，2 It means that after the second step, the new d = d i，2 ; Perform a linear transformation on each element of d: d i，3 =0.5+0.1×d i，2 Where d i，3 It means that after the third step, the new d = d i，3 ; The discrete dynamic model of the permanent magnet synchronous motor fault estimator is as follows: Where x(k) represents the system state, u(k) represents the system input, y(k) represents the system output, and d(k) represents the noise; Based on the above interference design process, a self-learning boundary reconstruction method is designed to reconstruct the system state. The estimated value is continuously corrected in multiple iterations to improve the estimation accuracy. The specific steps are as follows: Iteration 0: Iteration i: In the formula and They represent the estimated values ​​of the 0th and i-th system states respectively, and Respectively represent the estimated values ​​of the 0th and i-th disturbances, L, R, G are the observer gains to be designed; The reconstruction algorithm of the present invention can realize the safety margin required for fault detection, and the process is as follows: B001: Select a composite energy function of the following form: In the formula represents the state observation error of the system, represents the estimated error of the disturbance, P represents the Lyapunov variable matrix; B002: Calculate the k+1th sampling time V [i] (k+1) and the kth sampling time V [i] (k) and consider H ∞ Performance, including: In the formula γ represents H ∞ Performance level indicators, Respectively represent the estimated values ​​of gain matrices A, B, and D, Represent the transformed rank matrices of A, B, and D respectively; B003: Further consideration According to B002, the following formula can be obtained: B004: Therefore, from B003, we can get that when Θ<0, B005: According to B004, the permanent magnet synchronous motor system is asymptotically stable under the event-triggered control scheme designed by the present invention; Obtain the normal operation data of the permanent magnet synchronous motor and analyze the reachable set of the observed error signal of each iterative batch: In the formula e [i] (k) and Respectively represent the upper and lower bounds of the observation error, Ω [i] (0) represents the reachable set of the initial value of the observation error, Box(·) represents the interval boundary, represents the Minkowski sum, and further calculates the average residual signal interval, thereby generating the interval of the permanent magnet synchronous motor system state during normal operation; In the formula x [i] (k) and They represent the upper and lower bounds of the permanent magnet synchronous motor system state respectively; therefore, the safety boundary required for fault detection can be reconstructed from the upper and lower bounds of the state reachability.

Images