Switched reluctance motor turn-to-turn short circuit fault diagnosis method

Abstract

The invention discloses a switched reluctance motor turn-to-turn short circuit fault diagnosis method. According to the method, a second harmonic characteristic value is extracted according to an acquired stator torque signal of the switched reluctance motor, and a fault diagnosis model based on OCSSA-VMD-CNN-BiLSTM is established. The method comprises the following implementation steps: 1) acquiring a signal; 2) feature value extraction; 3) preprocessing characteristic values; 4) extracting time-frequency domain features of the feature values, and labeling; (5) a VMD-CNN-BiLSTM model optimized by the OCSSA is constructed; (6) an OCSSA-VMD-CNN-BiLSTM model is trained and tested; and 7) inputting the time-frequency characteristics of a to-be-detected sample into the model to obtain a fault diagnosis result. The off-line and on-line identification methods are combined, the turn-to-turn short circuit fault of the switched reluctance motor and the degree of the turn-to-turn short circuit fault of the switched reluctance motor can be accurately diagnosed, and a feasible fault diagnosis method is provided for the turn-to-turn short circuit fault of the switched reluctance motor.

Description

Technical Field

[0001] The invention relates to a method for diagnosing short-circuit faults of switched reluctance motor windings. The method uses a variational mode decomposition (VMD) algorithm optimized by an Osprey-Cauchy-Sparrow Search Algorithm (OCSSA) that integrates Osprey and Cauchy variations to extract feature quantities, and combines a convolutional neural network (CNN) and a bidirectional gated recurrent unit (BiGRU) method to build a fault diagnosis prediction model. Background Art

[0002] The switched reluctance motor has the advantages of simple structure, high reliability and strong fault tolerance. At the same time, since the rotor has no permanent magnets and windings, there is no risk of demagnetization, so it can be used in harsh environments such as high temperature and has broad application prospects. The winding faults of the switched reluctance motor can be divided into turn-to-turn short circuit, phase-to-phase short circuit, rectifier coil short circuit, phase winding short circuit, ground short circuit and winding open circuit. As the most common fault of the switched reluctance motor, the turn-to-turn short circuit fault will not have a significant impact on the operation of the system in the early stage, but the large circulating current flowing through the short-circuited winding will cause the local temperature to rise, and even cause various other faults to occur, thereby expanding the scope of the fault. At present, due to the nonlinear characteristics of the switched reluctance motor itself and the non-sinusoidal and non-DC power supply mode, there is little research on the fault diagnosis technology for the stator winding turn-to-turn short circuit. Therefore, it is of great significance and engineering value to accurately detect and diagnose the turn-to-turn short circuit fault of the switched reluctance motor to improve the safety and stability of the system. Summary of the invention

[0003] The purpose of the present invention is to provide a method for diagnosing a turn-to-turn short-circuit fault of a switched reluctance motor by using a simple and effective characteristic signal through an intelligent algorithm, thereby improving the accuracy of fault diagnosis.

[0004] The present invention discloses a method for diagnosing a short-circuit fault of a switched reluctance motor, and the specific steps are as follows:

[0005] Step 1: Collect stator torque signals of the switched reluctance motor under normal conditions and with different degrees of inter-turn short-circuit faults;

[0006] Step 2: The second harmonic in the fault torque signal in the two-phase rotating coordinate system is converted into a DC signal by the following rotating coordinate transformation formula, and the square sum is calculated after filtering by the average value to obtain the torque signal fault characteristic value;

[0007]

[0008] Where T represents the coordinate transformation matrix, ω represents the motor angular velocity, and t represents time.

[0009] Step 3: Preprocess the fault characteristic value signal and extract the original stable characteristic value signal data by sliding window;

[0010] Step 4: Use the OCSSA algorithm to optimize the parameters of the processed eigenvalue signal data, and bring the optimal penalty factor and mode number of each fault eigenvalue into VMD;

[0011] Step 5: Use the VMD algorithm with optimal parameters to obtain the optimal IMF component, calculate its 14 time domain features and 5 frequency domain features, and set labels respectively;

[0012] Step 6: Divide the labeled normal and multiple fault data time-frequency feature data samples into training samples and test samples in proportion, and the number of training samples is greater than the number of test samples;

[0013] Step 7: Set parameters and build a CNN-BiLSTM model using training samples;

[0014] Step 8: Use the test samples to evaluate the quality of the CNN-BiLSTM model. According to the accuracy of the model in classifying and identifying the test samples, select the parameters with the highest accuracy during the test as the optimal model parameters and save the optimal model.

[0015] Step 9: Perform coordinate transformation, average value filtering, and data preprocessing on the torque signal of the switched reluctance motor to be tested, and then input it into the optimal model. According to the output of the OCSSA-VMD-CNN-BiLSTM model, determine whether the motor corresponding to the sample has a fault and the degree of the fault.

[0016] Beneficial Effects

[0017] The present invention proposes a method for diagnosing the short-circuit fault of a switched reluctance motor based on OCSSA-VMD-CNN-BiLSTM. The torque signal of the stator of the switched reluctance motor is used to extract its second harmonic characteristic value through online rotating coordinate transformation, and then input into the VMD-CNN-BiLSTM fault diagnosis model optimized by OCSSA after average value filtering, and the optimal diagnosis model is obtained by offline training and testing the OCSSA-VMD-BiLSTM model. The present invention combines offline and online identification methods, can accurately diagnose the short-circuit fault of the switched reluctance motor and its degree, and provides a feasible fault diagnosis method for the short-circuit fault of the switched reluctance motor, which improves the safety and stability of the system and has good engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1This is the troubleshooting process.

[0019] Figure 2 It is the fault diagnosis confusion matrix result diagram.

[0020] Figure 3 This is the fault diagnosis result diagram. DETAILED DESCRIPTION

[0021] In conjunction with the accompanying drawings and specific implementation methods, the technical solution of the present invention is described in detail below by taking the turn-to-turn short circuit fault diagnosis of a three-phase 12 / 8-pole switched reluctance motor as an example.

[0022] Specific steps:

[0023] Step 1: Collect stator torque signals of the switched reluctance motor under normal conditions and with different degrees of inter-turn short-circuit faults;

[0024] Step 2: The second harmonic in the fault torque signal in the two-phase rotating coordinate system is converted into a DC signal by the following rotating coordinate transformation formula, and the square sum is calculated after filtering by the average value to obtain the torque signal fault characteristic value;

[0025]

[0026] Where T represents the coordinate transformation matrix, ω represents the motor angular velocity, and t represents time.

[0027] Step 3: Preprocess the fault characteristic value signal and extract the original stable characteristic value signal data by sliding window;

[0028] Step 4: Use the OCSSA algorithm to optimize the parameters of the processed eigenvalue signal data, and bring the optimal penalty factor and mode number of each fault eigenvalue into VMD;

[0029] Step 5: Use the VMD algorithm with optimal parameters to obtain the optimal intrinsic mode function (IMF) component, calculate its 14 time domain features and 5 frequency domain features, and set labels respectively;

[0030] Step 6: Divide the labeled normal and multiple fault data time-frequency feature data samples into training samples and test samples in proportion, and the number of training samples is greater than the number of test samples;

[0031] Step 7: Set parameters and build a CNN-BiLSTM model using training samples;

[0032] Step 8: Use the test samples to evaluate the quality of the CNN-BiLSTM model. According to the accuracy of the model in classifying and identifying the test samples, select the parameters with the highest accuracy during the test as the optimal model parameters and save the optimal model.

[0033] Step 9: Perform coordinate transformation, average value filtering, and data preprocessing on the torque signal of the switched reluctance motor to be tested, and then input it into the optimal model. According to the output of the OCSSA-VMD-CNN-BiLSTM model, determine whether the motor corresponding to the sample has a fault and the degree of the fault.

[0034] By designing 9 types of switched reluctance motors with short-circuit turns and obtaining their torque signals, the torque signals of the fault-free motors are used as signal samples together. The collected three-phase torque signals are converted into signals in a two-phase rotating coordinate system through Clark transformation and Park transformation online, and then the DC signals are obtained through the coordinate transformation matrix. After filtering by the average value, the square sum is taken to obtain the fault characteristic value. The above 10 fault characteristic values ​​are used to optimize the optimal IMF components through the OCSSA algorithm, and their time-frequency domain features are extracted and labeled respectively, and all samples are divided into training set and test set at a ratio of 3:1.

[0035] The accuracy of the switched reluctance motor turn-to-turn short-circuit fault diagnosis method based on OCSSA-VMD-CNN-BiLSTM can reach 99.6667%.

Claims

1. A method for diagnosing a short-circuit fault in a switched reluctance motor, characterized in that: The following steps are involved: Step 1: Collect stator torque signals of the switched reluctance motor under normal conditions and with different degrees of inter-turn short-circuit faults; Step 2: The second harmonic in the fault torque signal in the two-phase rotating coordinate system is converted into a DC signal by the following rotating coordinate transformation formula, and the square sum is calculated after filtering by the average value to obtain the torque signal fault characteristic value; Where T represents the coordinate transformation matrix, ω represents the motor angular velocity, and t represents time. Step 3: Preprocess the fault characteristic value signal and extract the original stable characteristic value signal data by sliding window; Step 4: Use the OCSSA algorithm to optimize the parameters of the processed eigenvalue signal data, and bring the optimal penalty factor and mode number of each fault eigenvalue into VMD; Step 5: Use the VMD algorithm with optimal parameters to obtain the optimal IMF component, calculate its 14 time domain features and 5 frequency domain features, and set labels respectively; Step 6: Divide the labeled normal and multiple fault data time-frequency feature data samples into training samples and test samples in proportion, and the number of training samples is greater than the number of test samples; Step 7: Set parameters and build a CNN-BiLSTM model using training samples; Step 8: Use the test samples to evaluate the quality of the CNN-BiLSTM model. According to the accuracy of the model in classifying and identifying the test samples, select the parameters with the highest accuracy during the test as the optimal model parameters and save the optimal model. Step 9: Perform coordinate transformation, average value filtering, and data preprocessing on the torque signal of the switched reluctance motor to be tested, and then input it into the optimal model. According to the output of the OCSSA-VMD-CNN-BiLSTM model, determine whether the motor corresponding to the sample has a fault and the degree of the fault.

2. The method for diagnosing turn-to-turn short-circuit fault of a switched reluctance motor based on OCSSA-VMD-CNN-BiLSTM according to claim 1, characterized in that: When OCSSA is used to optimize the parameters of the VMD algorithm in step 4, the minimum envelope entropy is used as the fitness function, and the optimal IMF component is obtained when the fitness is minimum.

3. The method for diagnosing turn-to-turn short-circuit fault of a switched reluctance motor based on OCSSA-VMD-CNN-BiLSTM according to claim 1, characterized in that: In step 5, the time-frequency domain features of the optimal IMF are extracted, and the maximum value, minimum value, average value, peak-to-peak value, rectified average value, variance, standard deviation, kurtosis, skewness, effective value, waveform factor, peak factor, pulse factor, and margin factor are used as time domain features; the center of gravity frequency, mean square frequency, root mean square frequency, frequency variance, and frequency standard deviation are used as frequency domain features.

Images