Electrical fault diagnosis method for permanent magnet synchronous motor

Abstract

The invention provides a permanent magnet synchronous motor electrical fault diagnosis method, and relates to the technical field of motor fault diagnosis. Comprising the following steps: S1, extracting historical electrical fault key characteristic quantity sets of a plurality of motors; s2, performing corresponding classification division on the plurality of historical electrical fault key characteristic quantity sets and different motor operation state categories to form a plurality of historical data sample sets; s3, calculating the prior probability of each motor operation state category; s4, extracting a newly collected electrical fault key characteristic quantity set to obtain a posterior probability of each motor operation state category; s5, supplementing and updating the characteristic quantity data to obtain the posterior probability of each motor operation state category; and S6, labeling the result, and storing the labeled information into a sample information base to update the historical data sample set. According to the method, the problem that most of current fault diagnosis methods cannot well meet the fault diagnosis requirement under the background of big data is solved.

Description

Technical Field

[0001] The invention relates to the technical field of motor fault diagnosis, and in particular to a method for diagnosing electrical faults of a permanent magnet synchronous motor. Background Art

[0002] With the continuous advancement of the technology of power electronic devices such as IGBT (insulated gate bipolar transistor) and SiC MOSFET (silicon carbide metal oxide semiconductor field effect transistor), the design and application of high-efficiency motors are rapidly developing in the direction of high power and high voltage. Permanent magnet synchronous motors (hereinafter referred to as motors) are mainly composed of three parts: stator (including winding), rotor and permanent magnet. The common faults that occur can be divided into mechanical faults, demagnetization faults and electrical faults according to their types. Among them: for mechanical faults, they can usually be diagnosed based on physical models using model-based methods such as Simulink simulation, and the diagnostic effect is relatively ideal; for demagnetization faults, intelligent algorithms can be used to identify them based on fault characteristics, and the application of this method is also relatively mature; however, for electrical faults, the diagnostic methods in the existing technology have limitations, specifically: Electrical faults include stator winding phase short circuit, stator winding open circuit and stator winding turn-to-turn short circuit faults. When an electrical fault occurs, fault-related information will appear in various physical signals represented by voltage, current, speed and torque. However, with the development of motors in terms of complexity and efficiency, the data reflecting the motor operating status presents "big data" characteristics such as massive, diversified, fast-flowing and low-value density. This makes most of the current fault diagnosis methods unable to meet the fault diagnosis needs in the context of big data. Specifically, it is reflected in the following points: First, when the motor works in harsh conditions, such as high temperature, vibration, magnetic field degradation, etc., the health of the motor winding will be affected through different mechanisms; due to the environmental noise and vibration interference of surrounding components during actual operation, the signal spectrum will be mixed with too many redundant harmonics, which will make it impossible to accurately extract fault features. In the prior art, although the motor status can be monitored by measuring motor signals such as vibration, noise, and temperature rise, this method requires the installation of additional signal sensors on the motor, which will impose many restrictions on the application scenarios of the motor.

[0003] Secondly, even if the motor works in a steady-state condition, various electrical fault features will change with different feature extraction windows due to the influence of various factors such as the environment, sampling equipment errors, and calculation errors, which will cause adverse effects such as decreased diagnostic accuracy and reduced diagnostic efficiency. In the prior art, although a method based on modeling and simulation is proposed, for example, the patent with publication number CN112926728A provides a fault diagnosis method, which utilizes the similarity between the simulation data and the real data of the permanent magnet synchronous motor, and uses the transfer learning method to transfer the deep learning model trained on the simulation data set to the real fault data set; however, compared with the simulation data of the motor, the fault characteristic value fluctuations will be greater during the actual operation, and some fault features with poor performance will lose the ability to characterize the motor state due to interference from external factors, and therefore cannot be used as the basis for motor fault diagnosis.

[0004] Third, there are differences in the probability of occurrence of different types of defects in equipment. This difference will cause the defect samples to have small sample characteristics, which makes it difficult to train an accurate prediction model. Based on this, if defects are judged only by relying on motor controller monitoring indicators and thresholds, it will be difficult to identify abnormal conditions when the monitoring data does not exceed the limit value and grows steadily.

[0005] In summary, the present invention provides a method for diagnosing electrical faults of a permanent magnet synchronous motor. Summary of the invention

[0006] The purpose of the present invention is to provide a permanent magnet synchronous motor electrical fault diagnosis method to solve the problem mentioned in the above background technology that most of the current fault diagnosis methods cannot well meet the fault diagnosis needs in the context of big data.

[0007] The present invention is achieved by adopting the following technical solutions: A method for diagnosing electrical faults of a permanent magnet synchronous motor comprises the following steps: Step S1: Establish a monitoring data set, and based on the correlation between the monitoring data set and the electrical fault state, extract a set of key characteristic quantities of historical electrical faults of several motors; Step S2: classifying and dividing a number of historical electrical fault key feature quantity sets into corresponding categories of different motor operation states to form a number of historical data sample sets; Step S3: Calculate the prior probability of each motor operation state category based on a plurality of historical data sample sets; Step S4: extracting a newly collected set of electrical fault key feature quantities, classifying them, and obtaining the posterior probability of each motor operating state category after classification through Bayesian reasoning; Step S5: Supplement the updated feature quantity data to the newly collected electrical fault key feature quantity set to form an updated electrical fault key feature quantity set, classify it, and obtain the posterior probability of each motor operation state category after classification; Step S6: Mark the classification results in step S5 according to the on-site confirmation results, and store the marked information in the sample information library to update the historical data sample set.

[0008] Furthermore, the step S1 includes the following sub-steps: Step S1-1: establishing a monitoring data set according to the real-time monitoring results and monitoring records, wherein the monitoring state quantity in the monitoring data set includes at least one of the rotational speed, torque, three-phase voltage, three-phase current, and motor temperature; Step S1-2: obtaining the dq axis current and dq axis voltage based on the three-phase current, and obtaining the fundamental wave and harmonics of the three-phase voltage, three-phase current, dq axis current and dq axis voltage by performing feature extraction; Step S1-3: respectively obtaining historical data information of a plurality of motors in normal operation and electrical fault state, and extracting relevant monitoring state quantities from the historical data information to form a data set in normal operation and a data set in electrical fault state; Step S1-4: Calculate and obtain the correlation between the relevant monitoring state quantity and the electrical fault state; Step S1-5: Based on the obtained correlation, extract a set of key characteristic quantities of historical electrical faults of several motors; the types of key characteristic quantities of electrical faults in the set of key characteristic quantities of historical electrical faults include at least one of the third harmonic of zero-sequence voltage, odd-order harmonics of zero-sequence voltage, third harmonic amplitude of phase current of the fault phase, second harmonic amplitude of d-axis voltage of the fault phase, second harmonic amplitude of d-axis current of the fault phase, torque, and speed.

[0009] Furthermore, in step S2, the permanent magnet synchronous motor finite element simulation and on-site operation electrical data are classified and divided to form a historical data sample set denoted as x= , ,…… ,in , … is the key characteristic quantity of electrical fault, Represents the operating status category.

[0010] Further, determine the historical data sample set H When calculating the conditional probability of the key characteristic quantities of each historical electrical fault under different motor operating status categories, , then based on the conditional probability Calculating conditional probabilities , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is .

[0011] Furthermore, the operating status categories include normal, stator winding phase-to-phase short circuit, stator winding open circuit and stator winding turn-to-turn short circuit fault.

[0012] Further, in step S4, the newly collected electrical fault key feature quantity set is recorded as n= , ,…… ; Classify and divide them to obtain When calculating the conditional probability of each newly collected key characteristic quantity of electrical fault under different motor operation status categories, , then based on the conditional probability Calculating conditional probabilities , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is , and the maximum conditional probability That is, the posterior probability of each motor operating state category after classification.

[0013] Further, in step S5, the updated electrical fault key feature quantity set is recorded as z= ,…… ; Classify and divide them to obtain When the conditional probability of each updated electrical fault key characteristic quantity under different motor operation status categories is first obtained by statistical analysis. , then As a priori knowledge, according to = Calculate to get , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is , and the maximum conditional probability That is, the posterior probability of each motor operating state category after classification.

[0014] Furthermore, in step S4, the conditional probability is calculated When the continuous newly collected electrical fault key feature quantity is , assuming that it obeys the normal distribution, the mean value of each newly collected electrical fault key feature quantity under different motor operation status categories is firstly calculated. and variance , and then calculate according to the following formula: .

[0015] The beneficial effects achieved by the present invention are: A permanent magnet synchronous motor electrical fault diagnosis method is provided. By setting steps S1 to S6, a good diagnosis effect of the permanent magnet synchronous motor electrical fault is achieved, wherein: a large amount of sample data is used to study the correlation between the monitoring data set and the electrical fault state. Compared with the existing method based on modeling and simulation, this method is more comprehensive, and can make continuous corrections to the diagnosis model, and can expand people's understanding of the laws of motor electrical faults. In addition, based on the collected normal operation data and small sample data of electrical faults, the prior information of the small sample from multiple sources can be effectively combined through Bayesian reasoning to obtain the posterior probability, thereby providing a decision-making basis for the classification of motor electrical faults. In addition, the probability of the fault category can be updated according to the updated feature quantity data, and the prior probability index of the motor fault in the diagnosis model can be automatically adjusted based on the labeled conclusion, thereby effectively ensuring the accuracy of fault classification. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a schematic flow chart of the diagnostic method according to an embodiment of the present invention; Figure 2 It is a schematic diagram of the motor operation control process. DETAILED DESCRIPTION

[0016] To clearly illustrate the solution in the present invention, further description is given below in conjunction with the accompanying drawings: Example 1 This embodiment provides a method for diagnosing electrical faults of a permanent magnet synchronous motor. Figure 1 , the method comprises the following steps: Step S1: Establish a monitoring data set, and based on the correlation between the monitoring data set and the electrical fault state, extract a set of key characteristic quantities of historical electrical faults of several motors. Specifically: Step S1 includes the following sub-steps: Step S1-1: Establish a monitoring data set according to the real-time monitoring results and monitoring records, wherein the monitoring state quantity in the monitoring data set includes at least one of the rotational speed, torque, three-phase voltage, three-phase current, and motor temperature.

[0017] Step S1-2: Based on the three-phase current, the dq axis current and dq axis voltage are obtained, and by performing feature extraction, the fundamental wave and harmonics of the three-phase voltage, three-phase current, dq axis current and dq axis voltage are obtained. In this step: The three-phase current is converted by Clark transformation and Park transformation to obtain dq axis current. The specific transformation equation used in this embodiment is as follows:

[0018] ; Furthermore, the dq axis voltage can be obtained according to the dq axis voltage and current equation. The dq axis voltage and current equation is as follows:

[0019] ; In the formula, , are the d-axis and q-axis voltages of the motor respectively, is the motor speed, is the stator resistance, is the permanent magnet flux of the motor, , They are the d-axis and q-axis inductances of the motor respectively.

[0020] Please refer to Figure 2 , in the specific solution: according to the target torque table, the target current value can be determined , , which can be collected through the corresponding sensors , , , and then through Clark transformation to get , and then through Park transformation to get the actual , ; Get the target current value , and actual , Then, the target can be obtained by using PI proportional integral control. , In addition, , Performing the inverse Park transform yields , , through SVPWM (space vector pulse width modulation) , Convert to , , , , , It can be input into the three-phase inverter to control the actual operation of the motor.

[0021] Feature extraction is performed through a wavelet packet transform method based on an adaptive random window. Specifically, it can be divided into two steps. The first step is to decompose the signal to obtain the coefficients of the fundamental wave and each harmonic component in the corresponding scale space; the second step is to reconstruct the fundamental wave and each harmonic according to the coefficients obtained in the first step to complete the harmonic detection process.

[0022] Step S1-3: Obtain historical data information of several motors in normal operation and electrical fault state respectively, and extract relevant monitoring state quantities from the historical data information to form a data set in normal operation and a data set in electrical fault state. In this step: The relevant monitoring state quantities include at least one of the speed, torque, fundamental wave and harmonics of three-phase voltage, fundamental wave and harmonics of three-phase current, fundamental wave and harmonics of dq axis current, fundamental wave and harmonics of dq axis voltage, and motor temperature.

[0023] Step S1-4: Calculate the correlation between the relevant monitoring state quantity and the electrical fault state. In this step: Based on the historical data information under normal operating conditions and electrical fault conditions obtained in step S1-3, the monitoring values ​​of the relevant monitoring state quantities of the motor are divided into intervals, and the distance between the maximum monitoring value and the minimum monitoring value of the relevant monitoring state quantity is equally spaced into n interval segments; in each interval segment, the number of motors with the same type of faults and the number of motors in normal operation are counted, and the failure rate of motors of the same type in the current interval segment = the number of motors with the same type of faults / (the total number of faulty motors + the number of normal motors). Then, the correlation coefficient between the monitoring value of the relevant monitoring state quantity and the failure rate of motors of the same type is calculated using the distance correlation algorithm, and the correlation coefficient represents the correlation (correlation) between the relevant monitoring state quantity and the electrical fault state. Wherein: The mean of the two end values ​​in each interval segment can be used as the quantized value of the interval segment to form a new monitoring value and failure rate sequence; for example, the monitoring value of the relevant monitoring state quantity is 1~100, which is divided into 4 interval segments, namely 1~25, 26~50, 51~75 and 76~100. The quantized value of each interval segment is converted to 13, 38, 63, and 88, thereby obtaining a new monitoring value and failure rate sequence, which is expressed as [13, failure rate 1], [38, failure rate 2], [63, failure rate 2], [88, failure rate 2].

[0024] The distance correlation algorithm is a method for calculating the correlation coefficient. The basic formula is as follows: , In the formula, is the indicator value sequence consisting of the mean value in each interval segment, is a sequence of failure rates of motors of the same type, Represents the distance covariance between the monitoring value of the relevant monitoring state quantity and the failure rate of the same type of motor, ; Represents the covariance between the monitoring value of the relevant monitoring state quantity and the failure rate of the same type of motor, ,in , represents the sample mean of the kth row, represents the sample mean of the jth row, is the distance matrix of the sample means, is the Euclidean distance between the relevant monitoring state quantity index values. The distance correlation has The characteristics of When X and Y are independent of each other, When , X and Y are completely correlated.

[0025] Step S1-5: Based on the obtained correlation, extract a set of key characteristic quantities of historical electrical faults of several motors; the types of key characteristic quantities of electrical faults in the set of key characteristic quantities of historical electrical faults include at least one of the third harmonic of zero-sequence voltage, odd-order harmonics of zero-sequence voltage, third harmonic amplitude of phase current of the fault phase, second harmonic amplitude of d-axis voltage of the fault phase, second harmonic amplitude of d-axis current of the fault phase, torque, and speed. In this step: A correlation sorting algorithm is used to sort multiple correlation coefficients to determine the key characteristic quantities of historical electrical faults. Specifically: the correlation sorting algorithm will arrange the correlation coefficients between the relevant monitoring state quantity index values ​​and the failure rates of the same type of motors in order from large to small, thereby adjusting a set of unordered sequences into an ordered sequence, and then obtaining an arranged sequence of the strength of the correlation between the relevant monitoring state quantity and the electrical fault state; in this embodiment, the relevant monitoring state quantity with a correlation coefficient greater than 0.8 is determined as the key characteristic quantity of historical electrical faults.

[0026] Step S2: Classify several sets of historical electrical fault key feature quantities into corresponding categories of different motor operation states to form several sets of historical data samples. Specifically: The motor operation status category is divided into two categories: normal and abnormal. In order to clarify the cause of the equipment abnormality, the abnormal status is further subdivided into normal, stator winding phase short circuit, stator winding open circuit and stator winding turn-to-turn short circuit faults. The motor operation status category classification set way to express it.

[0027] In this step, the permanent magnet synchronous motor finite element simulation and on-site operation electrical data are classified and divided to form a historical data sample set denoted as x= , ,…… ,in , … is the key characteristic quantity of electrical fault, Represents the operating status category. Determine the historical data sample set x When calculating the conditional probability of the key characteristic quantities of each historical electrical fault under different motor operating status categories, , then based on the conditional probability Calculating conditional probabilities , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is The above process can be summarized as .

[0028] In addition, to help understand, here is a special addition to the principle of Bayesian reasoning: the conditional probability of each feature quantity under each category is obtained statistically, denoted as , then according to Bayes' theorem, we know ; Since each feature attribute condition is independent, ; For all probability calculations are the same value, so Bayesian inference can be simplified to , among which if ,but .

[0029] Step S3: Based on several historical data sample sets, calculate the proportion of each motor operation status category in several historical data sample sets, which is the prior probability .

[0030] Step S4: extract the newly collected electrical fault key feature quantity set, classify it, and obtain the posterior probability of each motor operation state category after classification through Bayesian reasoning. Specifically: In this step, the newly collected electrical fault key feature quantity set is recorded as n= , ,…… ; Classify and divide them to obtain When calculating the conditional probability of each newly collected key characteristic quantity of electrical fault under different motor operation status categories, , then based on the conditional probability Calculating conditional probabilities , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is , and the maximum conditional probability That is, the posterior probability of each motor operating state category after classification.

[0031] In addition, the conditional probability is calculated When the continuous newly collected electrical fault key feature quantity is , assuming that it obeys the normal distribution, the mean value of each newly collected electrical fault key feature quantity under different motor operation status categories is firstly calculated. and variance , then follow

[0032] This calculation formula is used for calculation; for discrete feature quantities, it is calculated by counting the frequency of the feature quantity in the feature quantity set, that is, .

[0033] Step S5: Supplement the updated feature quantity data to the newly collected electrical fault key feature quantity set to form an updated electrical fault key feature quantity set, classify it, and obtain the posterior probability of each motor operation state category after classification. Specifically: In this step, the updated electrical fault key feature set is recorded as z= ,…… ; Classify and divide them to obtain When the conditional probability of each updated electrical fault key characteristic quantity under different motor operation status categories is first obtained by statistical analysis. , then As a priori knowledge, according to = Calculate to get , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is , and the maximum conditional probability That is, the posterior probability of each motor operating state category after classification.

[0034] Step S6: Annotate the classification results in step S5 according to the on-site confirmation results, and store the annotated information in the sample information library to update the historical data sample set and the probability knowledge in the above steps; where the on-site confirmation results refer to the classification of the motor status confirmed manually on-site. In addition, the operation sample information of motors in different regions can be summarized, the prior probability and conditional probability knowledge under different operating status categories can be calculated, and the knowledge can be automatically updated to each regional level diagnosis model.

[0035] For the above steps, some example data are provided in this embodiment for reference, specifically: The monitoring data set established includes a speed of 3000 r / min, a torque of 120 N•m, and three-phase voltage and three-phase current; the dq axis current and dq axis voltage are obtained based on the three-phase current, and the fault features are extracted by the wavelet packet transform method based on the adaptive random window to obtain the fundamental wave and harmonics of the three-phase voltage, three-phase current, dq axis current and dq axis voltage. During the extraction, 100 sets of data are extracted for each fault feature, so a total of 400 sets of data. The historical data information of the motor under the electrical fault state is obtained, and based on this, a set of key characteristic quantities of historical electrical faults is constructed. After the key characteristic quantity set of historical electrical faults and the motor operating state categories are classified and divided, a set of historical data samples is formed and corresponding calculations are performed based on it. The prior probability of different motor operating state categories can be obtained as shown in the following table: Table 1 Prior probability of different motor operating status categories

[0036] Then, the newly collected electrical fault key feature quantity set is extracted through the motor online monitoring system and classified. For the newly collected electrical fault key feature quantity that obeys the normal distribution, the mean value of each newly collected electrical fault key feature quantity under different motor operating status categories is calculated. and variance For example, the newly collected electrical fault key feature quantity set specifically includes the third harmonic amplitude of the phase current of the fault phase, the second harmonic amplitude of the d-axis voltage of the fault phase, and the second harmonic amplitude of the d-axis current of the fault phase. The specific data is as follows: Table 2 Mean and variance calculation results of some newly collected key characteristics of electrical faults

[0037] The third harmonic amplitude of the phase current of the fault phase, the second harmonic amplitude of the d-axis voltage of the fault phase, and the second harmonic amplitude of the d-axis current of the fault phase obtained at a certain moment are 7A, 0.95V, and 3.64A respectively. The posterior probability of each motor operation state category after classification and division according to Bayesian reasoning is shown in Table 3. By sorting the probability values, it can be seen that the current motor has a stator winding inter-turn short circuit fault.

[0038] In summary, by executing the diagnostic method provided in this embodiment, a good diagnostic effect for electrical faults of permanent magnet synchronous motors can be achieved. In this diagnostic method, Bayesian reasoning is used to calculate the probability of operating state categories under different feature quantities, and then identify the fault type; based on new operating evidence and annotation information, the prior probability index in the diagnostic model can be automatically adjusted, which helps to ensure the accuracy of fault classification.

[0039] It should be noted that the parts not described in detail or expanded in the above scheme are all existing technologies, which do not belong to the improvements made by the present invention on the existing technologies, nor do they belong to the protection scope of the technical scheme of the present invention. Therefore, they will not be described in detail in this article.

[0040] Of course, the above contents are only preferred embodiments of the present invention and cannot be considered to limit the scope of the embodiments of the present invention. The present invention is not limited to the above examples, and equal changes and improvements made by ordinary technicians in the technical field within the essential scope of the present invention should all fall within the scope of the patent coverage of the present invention.

Claims

1. A method for diagnosing electrical faults of a permanent magnet synchronous motor, characterized in that: The steps include: Step S1: Establish a monitoring data set, and based on the correlation between the monitoring data set and the electrical fault state, extract a set of key characteristic quantities of historical electrical faults of several motors; Step S2: classifying and dividing a number of historical electrical fault key feature quantity sets into corresponding categories of different motor operation states to form a number of historical data sample sets; Step S3: Calculate the prior probability of each motor operation state category based on a plurality of historical data sample sets; Step S4: extracting a newly collected set of key characteristic quantities of electrical faults, classifying them, and obtaining the posterior probability of each motor operating state category after classification through Bayesian reasoning; Step S5: Supplement the updated feature quantity data to the newly collected electrical fault key feature quantity set to form an updated electrical fault key feature quantity set, classify it, and obtain the posterior probability of each motor operation state category after classification; Step S6: Mark the classification results in step S5 according to the on-site confirmation results, and store the marked information in the sample information library to update the historical data sample set.

2. The method for diagnosing electrical faults of a permanent magnet synchronous motor according to claim 1, characterized in that: The step S1 includes the following sub-steps: Step S1-1: establishing a monitoring data set according to the real-time monitoring results and monitoring records, wherein the monitoring state quantity in the monitoring data set includes at least one of the rotational speed, torque, three-phase voltage, three-phase current, and motor temperature; Step S1-2: obtaining dq axis current and dq axis voltage based on the three-phase current, and obtaining fundamental waves and harmonics of the three-phase voltage, three-phase current, dq axis current and dq axis voltage by performing feature extraction; Step S1-3: respectively obtaining historical data information of a plurality of motors in normal operation and electrical fault state, and extracting relevant monitoring state quantities from the historical data information to form a data set in normal operation and a data set in electrical fault state; Step S1-4: Calculate and obtain the correlation between the relevant monitoring state quantity and the electrical fault state; Step S1-5: Based on the obtained correlation, extract a set of key characteristic quantities of historical electrical faults of several motors; the types of key characteristic quantities of electrical faults in the set of key characteristic quantities of historical electrical faults include at least one of the third harmonic of zero-sequence voltage, odd-order harmonics of zero-sequence voltage, third harmonic amplitude of phase current of the fault phase, second harmonic amplitude of d-axis voltage of the fault phase, second harmonic amplitude of d-axis current of the fault phase, torque, and speed.

3. The method for diagnosing electrical faults of a permanent magnet synchronous motor according to claim 2, characterized in that: In step S2, classification is performed based on the permanent magnet synchronous motor finite element simulation and on-site operation electrical data to form a historical data sample set denoted as x= , ,…… ,in , … is the key characteristic quantity of electrical fault, Represents the operating status category.

4. The method for diagnosing electrical faults of a permanent magnet synchronous motor according to claim 3, characterized in that: Determine the historical data sample set x When calculating the conditional probability of the key characteristic quantities of each historical electrical fault under different motor operating status categories, , then based on the conditional probability Calculating conditional probabilities , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is .

5. The method for diagnosing electrical faults of a permanent magnet synchronous motor according to claim 3, characterized in that: The operating status categories include normal, stator winding phase-to-phase short circuit, stator winding open circuit and stator winding turn-to-turn short circuit fault.

6. The method for diagnosing electrical faults of a permanent magnet synchronous motor according to claim 4, characterized in that: In step S4, the newly collected electrical fault key feature quantity set is recorded as n= , ,…… ; Classify it to get When calculating the conditional probability of each newly collected key characteristic quantity of electrical fault under different motor operation status categories, , then based on the conditional probability Calculating conditional probabilities , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is , and the maximum conditional probability That is, the posterior probability of each motor operating state category after classification.

7. The method for diagnosing electrical faults of a permanent magnet synchronous motor according to claim 6, characterized in that: In step S5, the updated electrical fault key feature quantity set is recorded as z= ,…… ; Classify it to get When the conditional probability of each updated electrical fault key characteristic quantity under different motor operation status categories is first obtained by statistical analysis. , then As a priori knowledge, according to = Calculate to get , and finally we get multiple conditional probabilities Sort by the largest conditional probability In That is , and the maximum conditional probability That is, the posterior probability of each motor operating state category after classification.

8. The method for diagnosing electrical faults of a permanent magnet synchronous motor according to claim 6, characterized in that: In step S4, the conditional probability is calculated When the continuous newly collected electrical fault key feature quantity is , assuming that it obeys the normal distribution, the mean value of each newly collected electrical fault key feature quantity under different motor operation status categories is firstly calculated. and variance , and then calculate according to the following formula: .

Images