A method for fault diagnosis during transient operation of an electric motor

Abstract

This invention relates to a fault diagnosis method for a motor during transient operation. It employs a current sensor to acquire the transient stator current signal of the permanent magnet synchronous motor (PMSM) under test in real time. Then, a variational mode decomposition algorithm optimized using the genetic-black tern algorithm is applied to the sampled transient stator current signal to obtain multiple modal component signals containing fault characteristics. Next, a Hilbert transform is performed on the obtained modal component signals to obtain the instantaneous frequency corresponding to each modal component signal. By analyzing whether the instantaneous frequencies corresponding to each modal component signal contain the characteristic frequency of inter-turn short-circuit faults, it is determined whether an inter-turn short-circuit fault has occurred in the PMSM. Compared with existing technologies, this invention can perform fault diagnosis not only under steady-state motor operation but also under transient motor operation, and significantly improves the accuracy and real-time performance of fault diagnosis.

Description

Technical Field

[0001] This invention relates to the field of electrical engineering technology, and specifically to a method for diagnosing faults in a motor during transient operation. Background Technology

[0002] In recent years, permanent magnet synchronous motors (PMSMs) have been widely used in high-tech fields such as aerospace, electric vehicles, CNC lathes, and intelligent robots due to their advantages such as high efficiency, high power factor, precise torque control, simple structure, and low noise, becoming an important component of advanced equipment. However, considering the harsh operating conditions, enclosed and confined working environment, and poor heat dissipation of PMSMs, they are highly susceptible to overload and overheating, resulting in a high failure rate. As a crucial core component of the system, the reliability of the motor directly affects the performance of the entire system; a failure can easily trigger a chain reaction, leading to the paralysis of the entire system. Therefore, improving methods for motor fault diagnosis is of great significance for enhancing the reliability and safety of motor operation.

[0003] In existing motor fault detection methods, current signal spectrum analysis is widely used due to its ease of acquisition and rich fault information. However, traditional current spectrum analysis is susceptible to fundamental frequency leakage and load fluctuations, making it difficult to extract fault characteristic components and easily leading to misjudgments. Furthermore, most traditional signal spectrum analysis methods directly employ Fourier transform, which can only process steady-state signals and is ineffective for transient signals.

[0004] Variational Mode Decomposition (VMD) is a completely non-recursive variational method that decomposes a signal into fast and slow oscillations. It is an adaptive, non-recursive, and quasi-orthogonal signal decomposition method, commonly used to analyze nonlinear and non-stationary transient signal sequences, and boasts a high signal-to-noise ratio. However, two crucial parameters of VMD—the penalty factor and the number of modal components—significantly influence the convergence speed and mode aliasing of each mode, thus requiring careful consideration and appropriate selection. This invention provides a fault diagnosis method for motors during transient operation, employing a genetic-black tern algorithm to optimize the penalty factor and the number of modal components. This accelerates the convergence speed of the VMD algorithm and resolves the mode aliasing phenomenon that may occur in the decomposition of transient stator current signals in motors. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a fault diagnosis method for motors during transient operation, which solves the technical problem that existing methods for diagnosing inter-turn short circuit faults in permanent magnet synchronous motors cannot analyze transient stator current signals, improves the real-time performance of fault diagnosis, ensures timely detection of motor faults, and enables the motor to operate safely and stably.

[0006] To address the aforementioned technical problems, this invention provides a method for diagnosing faults in a motor during transient operation, comprising the following steps: real-time acquisition of transient stator current signals of the motor to be identified and diagnosed within a preset time interval; modal decomposition of the acquired transient stator current signals using a variational mode decomposition algorithm optimized by the genetic-black tern algorithm to obtain multiple modal sub-signals containing fault characteristics; Hilbert transform of the obtained multiple modal sub-signals to obtain the instantaneous frequency corresponding to each modal sub-signal; and analysis of whether the instantaneous frequency corresponding to each modal sub-signal contains the fault characteristic frequency to determine whether a motor fault has occurred.

[0007] The specific process is as follows:

[0008] Step 1: Use a current sensor to collect the transient stator current signal of the permanent magnet synchronous motor to be identified and diagnosed within a preset time interval, wherein the sampling period is at least 10 fundamental cycles of the stator steady-state current signal, and at least 25 points are sampled per cycle.

[0009] Step 2: The range of values ​​for the two parameters of the variational mode decomposition algorithm, the penalty factor α and the number of mode components K, are set to [500 10000] and [1 20], respectively.

[0010] Step 3: Set the population size to Q, where Q is generally greater than 10. Based on the range of values ​​for the penalty factor α and the number of modal components K from Step 2, randomly generate Q population particles [α, K] = [α1, K1], [α2, K2]...[α...]. i K i ] is used as the initialization parameter, where i = 1, 2, 3, ..., Q, and the maximum number of iterations is set to M, where M is generally greater than 100.

[0011] Step 4: The generated parameters [α] i K i Substituting this into the variational mode decomposition algorithm, the transient stator current signal obtained in step 1 is then subjected to variational mode decomposition to obtain K. i One modal component.

[0012] Step 5: Based on the sparsity characteristics of the transient stator current signal obtained in Step 1, the envelope entropy is selected as the fitness function. The fitness function formula is as follows: (where j = 1, 2, 3, ..., N, and N is the signal length of the modal component sub-signal), where a(j) is the envelope signal obtained after demodulation of the modal component obtained in step 4.

[0013] Step 6: Calculate K obtained in Step 4 respectively. i The fitness function values ​​of each modal component are calculated, and the minimum value is selected as the individual fitness value f. i .

[0014] Step 7: Repeat the fitness values ​​of the Q individuals obtained in steps 4-6, and calculate their average fitness value.

[0015] Step 8: Compare the individual fitness value obtained in Step 6 with the average fitness value obtained in Step 7.

[0016] Step 9: Based on the comparison results in Step 8, if the individual fitness value is less than the average fitness value, it indicates that the neighboring search region [α, K] is promising, so a genetic algorithm with strong local search is used for the search; if the individual fitness value is greater than the average fitness value, the local search strategy is not used, the Black-and-White algorithm is used for the search, and the process jumps to Step 15 to continue running.

[0017] Step 10: The genetic algorithm iteratively updates [α, K].

[0018] Step 11: Substitute [α, K] obtained in Step 10 into the variational mode decomposition algorithm to perform variational mode decomposition on the current signal and calculate the individual fitness value.

[0019] Step 12: Compare the individual fitness values ​​obtained in Step 11 and select the minimum value as the local minimum.

[0020] Step 13: Compare the local minima obtained in step 12 after each iteration, select the minimum value as the new local minima, and use the [α, K] corresponding to the new local minima as the input parameter for the next iteration update.

[0021] Step 14: Determine if the maximum number of iterations M has been reached. If it has, proceed to step 20 to continue running; otherwise, proceed to step 10 to continue running.

[0022] Step 15: The Black-and-White Algorithm iteratively updates [α, K].

[0023] Step 16: Substitute [α, K] obtained in Step 15 into the variational mode decomposition algorithm to perform variational mode decomposition on the current signal and calculate the individual fitness value.

[0024] Step 17: Compare the individual fitness values ​​obtained in Step 16 and select the minimum value as the local minimum.

[0025] Step 18: Compare the local minima obtained in step 17 after each iteration, select the minimum value as the new local minima, and use the [α, K] corresponding to the new local minima as the input parameter for the next iteration update.

[0026] Step 19: Determine if the maximum number of iterations M has been reached. If it has, proceed to step 20 to continue running; otherwise, proceed to step 15 to continue running.

[0027] Step 20: Compare the local minima obtained in Step 13 and Step 18, and select the smaller one as the global minima.

[0028] Step 21: Output the global minimum of the fitness function and its corresponding [α0, K0]. At this point, α0 and K0 are the optimal penalty factor and the number of modal components.

[0029] Step 22: Set the parameters for variational mode decomposition based on the optimal parameter penalty factor and the number of signal components obtained in Step 21.

[0030] Step 23: Use the variational mode decomposition algorithm optimized by the parameters in step 22 to perform mode decomposition on the transient stator current signal obtained in step 1, and obtain K0 mode component quantum signals.

[0031] Step 24: Calculate the instantaneous frequencies of the K0 modal component quantum signals obtained in Step 23 using the Hilbert transform.

[0032] Step 25: By analyzing whether the instantaneous frequency corresponding to each modal sub-signal contains the characteristic frequency f0 of the inter-turn short circuit fault of the permanent magnet synchronous motor, when a frequency of (f0±2)Hz is detected, it is the fault characteristic frequency, and thus the motor fault is diagnosed.

[0033] The beneficial effects of this invention are:

[0034] This invention employs a fault diagnosis method for motors during transient operation. Verification has shown that a variational mode decomposition algorithm optimized using the genetic-black tern algorithm is applied to diagnose inter-turn short-circuit faults in permanent magnet synchronous motors. This method can perform inter-turn short-circuit fault diagnosis under transient motor operation conditions, significantly improving the real-time performance of fault diagnosis. Attached Figure Description

[0035] Figure 1 This is a flowchart of a fault diagnosis method for a motor during transient operation according to the present invention.

[0036] Figure 2 This is a flowchart of the variational mode decomposition algorithm based on the genetic-black tern algorithm of the present invention.

[0037] Figure 3 This is a modal component diagram of the stator current transient signal obtained by the genetic-terrier optimized variational mode decomposition algorithm of this invention.

[0038] Figure 4 This is the instantaneous frequency diagram corresponding to the modal component IMF1.

[0039] Figure 5 This is the instantaneous frequency diagram corresponding to the modal component IMF2.

[0040] Figure 6 This is the instantaneous frequency diagram corresponding to the modal component IMF3. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0042] like Figure 1-3 As shown, the genetic-terrier algorithm is first used to optimize the penalty factor and the number of modal components in the variational mode decomposition algorithm. Then, the initial parameters of the variational mode decomposition algorithm are set according to the optimized penalty factor and the number of modal components. The transient stator current signal of the motor to be identified and diagnosed is acquired in real time within a preset time interval. The acquired transient stator current signal is then subjected to mode decomposition using the optimized variational mode decomposition algorithm to obtain multiple modal sub-signals containing fault characteristics. Then, a Hilbert transform is performed on the obtained multiple modal sub-signals to obtain the instantaneous frequency corresponding to each modal sub-signal. By analyzing whether the instantaneous frequency corresponding to each modal sub-signal contains the fault characteristic frequency, the occurrence of a motor fault can be diagnosed.

[0043] A method for diagnosing faults in a motor during transient operation includes the following steps:

[0044] Step 1: Use a current sensor to collect the transient stator current signal of the permanent magnet synchronous motor to be identified and diagnosed within a preset time interval. The sampling period is 12 fundamental cycles, the sampling frequency is 2000Hz, the sampling time is 0.2s, and the number of sampling points is 400.

[0045] Step 2: The range of values ​​for the two parameters of the variational mode decomposition algorithm, the penalty factor α and the number of mode components K, are set to [500 10000] and [1 20], respectively.

[0046] Step 3: Set the population size to Q = 30. Based on the range of values ​​for the penalty factor α and the number of modal components K in Step 2, randomly generate 30 population particles [α, K] = [α1, K1], [α2, K2]...[α...]. i K i ] is used as the initialization parameter, where i = 1, 2, 3, ..., 30, and the maximum number of iterations is set to 120.

[0047] Step 4: The generated parameters [α] i K i Substituting this into the variational mode decomposition algorithm, the transient stator current signal obtained in step 1 is then subjected to variational mode decomposition to obtain K. i One modal component.

[0048] Step 5: Based on the sparsity characteristics of the transient stator current signal obtained in Step 1, the envelope entropy is selected as the fitness function. The fitness function formula is as follows: (where j = 1, 2, 3, ..., 400), where a(j) is the envelope signal obtained after demodulation of the modal components obtained in step 4.

[0049] Step 6: Calculate K obtained in Step 4 respectively. i The fitness function values ​​of each modal component are calculated, and the minimum value is selected as the individual fitness value f. i .

[0050] Step 7: Repeat steps 4-6 to obtain the fitness values ​​of the 30 individuals and calculate their average fitness value.

[0051] Step 8: Compare the individual fitness value obtained in Step 6 with the average fitness value obtained in Step 7.

[0052] Step 9: Based on the comparison results in Step 8, if the individual fitness value is less than the average fitness value, it indicates that the neighboring search region [α, K] is promising, so a genetic algorithm with strong local search is used for the search; if the individual fitness value is greater than the average fitness value, the local search strategy is not used, the Black-and-White algorithm is used for the search, and the process jumps to Step 15 to continue running.

[0053] Step 10: The genetic algorithm iteratively updates [α, K].

[0054] Step 11: Substitute [α, K] obtained in Step 10 into the variational mode decomposition algorithm to perform variational mode decomposition on the current signal and calculate the individual fitness value.

[0055] Step 12: Compare the individual fitness values ​​obtained in Step 11 and select the minimum value as the local minimum.

[0056] Step 13: Compare the local minima obtained in step 12 after each iteration, select the minimum value as the new local minima, and use the [α, K] corresponding to the new local minima as the input parameter for the next iteration update.

[0057] Step 14: Determine if the maximum number of iterations (120) has been reached. If it has, proceed to step 20 to continue running; otherwise, proceed to step 10 to continue running.

[0058] Step 15: The Black-and-White Algorithm iteratively updates [α, K].

[0059] Step 16: Substitute [α, K] obtained in Step 15 into the variational mode decomposition algorithm to perform variational mode decomposition on the current signal and calculate the individual fitness value.

[0060] Step 17: Compare the individual fitness values ​​obtained in Step 16 and select the minimum value as the local minimum.

[0061] Step 18: Compare the local minima obtained in step 17 after each iteration, select the minimum value as the new local minima, and use the [α, K] corresponding to the new local minima as the input parameter for the next iteration update.

[0062] Step 19: Determine if the maximum number of iterations (120) has been reached. If it has, proceed to step 20 to continue running; otherwise, proceed to step 15 to continue running.

[0063] Step 20: Compare the local minima obtained in Step 13 and Step 18, and select the smaller one as the global minima.

[0064] Step 21: Output the global minimum of the fitness function and its corresponding [α0, K0]. At this point, α0 and K0 are the optimal penalty factor and the number of modal components.

[0065] Step 22: Set the parameters for variational mode decomposition based on the optimal parameter penalty factor and the number of signal components obtained in Step 21.

[0066] Step 23: Use the variational mode decomposition algorithm optimized by the parameters in step 22 to perform mode decomposition on the transient stator current signal obtained in step 1, and obtain K0 mode component quantum signals.

[0067] Step 24: Calculate the instantaneous frequencies of the K0 modal component quantum signals obtained in Step 23 using the Hilbert transform.

[0068] Step 25: By analyzing whether the instantaneous frequency corresponding to each modal sub-signal contains the characteristic frequency f0 of the inter-turn short circuit fault of the permanent magnet synchronous motor, when a frequency of (f0±2)Hz is detected, it is the fault characteristic frequency, and thus the motor fault is diagnosed.

[0069] Example:

[0070] This test uses a permanent magnet synchronous motor with 4 pole pairs (p=4) and a stator winding short-circuit turns ratio of 0.1. The motor is operated at a speed of 900 r / min. At 3.1 s, the load torque is increased, and a current sensor is used to acquire the transient stator current signal of the permanent magnet synchronous motor during the time interval from 3 s to 3.2 s. The signal sampling frequency is 2000 Hz.

[0071] Based on steps 2 to 21 above, the optimal parameters for variational mode decomposition are determined: K = 4, α = 1850. These parameters are then used as the parameters for variational mode decomposition. The acquired transient stator current signal is then subjected to mode decomposition using the pre-defined variational mode decomposition parameters, and the results are as follows: Figure 3 As shown, the signal is decomposed into four modal component sub-signals: IMF1, IMF2, IMF3, and IMF. Then, the instantaneous frequencies of these four modal component sub-signals are calculated using Hilbert transform, as shown below. Figure 4 , Figure 5 and Figure 6 As shown.

[0072] When a permanent magnet synchronous motor experiences an inter-turn short-circuit fault, due to the asymmetry of the three-phase windings, the corresponding fault characteristic frequency is the third harmonic frequency of the fundamental frequency. The fault characteristic frequency is determined when a transient stator current signal with a frequency of (3f1±2) Hz is detected, where f1 is the fundamental frequency. The fundamental frequency of the tested motor is f1 = n*p / 60 = 60 Hz, and its corresponding third harmonic frequency is 3f1 = 180 Hz. The modal component quantum signal containing the fault characteristic frequency is IMF3, and its corresponding instantaneous frequency according to the Hilbert transform is as follows: Figure 5 As shown in the figure, the instantaneous frequency corresponding to IMF2 is approximately 180Hz, which is the third harmonic characteristic frequency containing inter-turn short circuit faults. Therefore, the method of the present invention can effectively detect inter-turn short circuit faults under the transient operation state of the motor.

[0073] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Any modifications or substitutions made by those skilled in the art within the scope of the technology disclosed in this invention should be included within the scope of this invention.

Claims

1. A method for diagnosing a fault of an electric machine during transient operation, characterized in that The method comprises the following steps: collecting transient stator current signals of a motor to be identified and diagnosed within a preset time interval in real time, performing modal decomposition on the collected transient stator current signals by using a variational mode decomposition algorithm optimized by a genetic-tern algorithm, obtaining a plurality of modal sub-signals containing fault characteristics, and then performing Hilbert transform on the obtained plurality of modal component sub-signals to obtain instantaneous frequencies corresponding to the modal sub-signals; and determining whether a motor fault occurs by analyzing whether the instantaneous frequencies corresponding to the modal sub-signals contain a fault characteristic frequency. The method comprises the following implementation processes: Step 1: collecting transient stator current signals of a permanent magnet synchronous motor to be identified and diagnosed within a preset time interval by using a current sensor; Step 2: setting a value range of two parameters, a penalty factor α and a modal component number K, of the variational mode decomposition algorithm; Step 3: Set the population number as Q, Q is generally greater than 10, according to the value range of the penalty factor a and the number of modal components K in step 2, randomly generate Q population particles [a, K] = [a1, K1], [a2, K2]…[a i , K i ] as the initialization parameters, where i = 1, 2, 3…, Q, and set the maximum iteration number as M, M is generally greater than 100; Step 4: The generated parameters [a i , K i ] are brought into the variational modal decomposition algorithm, and then the transient stator current signal obtained in Step 1 is subjected to variational modal decomposition, and K i modal components are obtained; Step 5: According to the sparse characteristics of the transient stator current signal obtained in step 1, the envelope entropy is selected as the fitness function, and the fitness function formula is (wherein j = 1, 2, 3…, N, N is the signal length of the modal component sub-signal), wherein a(j) is the envelope signal obtained after the demodulation operation of the modal component obtained in step 4; Step 6: Calculate K obtained in Step 4 respectively. i The fitness function values ​​of each modal component are calculated, and the minimum value is selected as the individual fitness value f. i ; Step 7: repeating the Q individual fitness values obtained in steps 4-6, and calculating an average fitness value thereof; Step 8: comparing the individual fitness value obtained in step 6 with the average fitness value obtained in step 7; Step 9: according to the comparison result in step 8, if the individual fitness value is less than the average fitness value, it is indicated that the adjacent search region of [α, K] is promising, and therefore a local search strong genetic algorithm is used for searching; if the individual fitness value is greater than the average fitness value, a local search strategy is not used, and the tern algorithm is used for searching, and the process is continued in step 15; Step 10: iteratively updating [α, K] by using the genetic algorithm; Step 11: substituting [α, K] obtained in step 10 into the variational mode decomposition algorithm, performing variational mode decomposition on the current signal, and calculating an individual fitness value; Step 12: comparing the individual fitness values obtained in step 11, and selecting a minimum value as a local minimum value; Step 13: comparing the local minimum values obtained in step 12 after each loop iteration, selecting a minimum value as a new local minimum value, and taking [α, K] corresponding to the new local minimum value as an input parameter for the next iteration update; Step 14: determining whether a maximum iteration number M is reached, if yes, the process is continued in step 20, otherwise, the process is continued in step 10; Step 15: iteratively updating [α, K] by using the tern algorithm; Step 16: substituting [α, K] obtained in step 15 into the variational mode decomposition algorithm, performing variational mode decomposition on the current signal, and calculating an individual fitness value; Step 17: comparing the individual fitness values obtained in step 16, and selecting a minimum value as a local minimum value; Step 18: comparing the local minimum values obtained in step 17 after each loop iteration, selecting a minimum value as a new local minimum value, and taking [α, K] corresponding to the new local minimum value as an input parameter for the next iteration update; Step 19: determining whether a maximum iteration number M is reached, if yes, the process is continued in step 20, otherwise, the process is continued in step 15; Step 20: comparing the local minimum values obtained in steps 13 and 18, and selecting a smaller one as a global minimum value; Step 21: output the global minimum value of the fitness function and its corresponding [a0, K0], at this time a0 and K0 are the optimal penalty factor and the number of modal components; Step 22: according to the optimal parameter penalty factor and the number of signal components obtained in step 21, the parameter setting of variational modal decomposition is carried out; Step 23: using the variational modal decomposition algorithm optimized in step 22, the transient stator current signal obtained in step 1 is modal decomposed to obtain K0 modal component sub-signals; Step 24: the instantaneous frequency of the K0 modal component sub-signals obtained in step 23 is calculated by Hilbert transform; Step 25: by analyzing whether the instantaneous frequency corresponding to each modal sub-signal contains the fault characteristic frequency f0 of the inter-turn short circuit of the permanent magnet synchronous motor, when the frequency (f0±2) Hz is detected, it is the fault characteristic frequency, so as to diagnose whether the motor fault occurs.

2. The method of claim 1, wherein, The sampling period in step 1 is at least 10 periods of the stator steady-state current signal fundamental wave, and at least 25 points are sampled per period.

3. The method of claim 1, wherein, The average fitness value is 4. The method of claim 1, wherein, The value range of the two parameters, the penalty factor a and the number of modal components K, is set to [500 10000] and [1 20] respectively.

Images