Comprehensive diagnosis method for open-circuit fault of permanent magnet synchronous motor driving system

Abstract

The invention discloses a comprehensive diagnosis method for an open-circuit fault of a permanent magnet synchronous motor driving system, and belongs to the technical field of reliability improvement of motor driving systems. According to the method, phase current is used as original fault data, normalization processing is firstly carried out, then a positive and negative half-cycle average value of a phase current fundamental wave cycle is calculated based on fault analysis and is used as a fault feature, then feature data is input into an optimized support vector machine for training, a mapping relation between the fault feature and a fault label is learned, and the fault is obtained. And finally, a fault diagnosis model is formed. The diagnosis model is 24 decision functions, and the model output result is a fault label. According to the invention, accurate diagnosis of the open-phase fault, the open-circuit fault of the switching tube and the disconnection fault of the current sensor of the permanent magnet synchronous motor driving system can be realized only by depending on the phase current data without depending on a motor model and artificial experience, so that wrong diagnosis caused by factors such as manual work can be effectively avoided; and the system can be integrated in an electric drive control program to realize on-line diagnosis.

Description

Technical Field

[0001] The present invention belongs to the technical field of reliability improvement of motor drive systems, and in particular relates to a comprehensive diagnosis method for open-circuit faults of permanent magnet synchronous motor drive systems. Background Art

[0002] Multi-phase motors have the characteristics of high power density and small torque pulsation. At the same time, dual three-phase motors are easy to process and can be realized by adjusting the windings of traditional three-phase motors. Therefore, dual three-phase motors are of great significance to the research of high-reliability and high-precision drive systems.

[0003] However, in practical applications, such as elevators, electric vehicles, and high-precision CNC machine tools, the operating conditions of motors are complex and changeable, so a series of faults may occur due to problems such as motor overheating and voltage spikes. Open circuit faults are one of the most common types of faults, especially phase loss faults, switch tube open circuit faults, and current sensor faults in electrical faults. When an open circuit fault occurs, it may cause torque fluctuations, reduced control accuracy, motor overheating, secondary faults, and even paralysis of the entire system. Therefore, it is of great significance to develop fault diagnosis methods for permanent magnet synchronous motor drive systems.

[0004] Existing fault diagnosis methods are divided into signal-based, model-based and data-driven. Signal-based and model-based methods generally require detailed analysis or precise modeling of specific signals and rely on manually preset thresholds. Both types of methods are difficult to diagnose multiple faults. Although the existing data-driven methods have a wide range of fault diagnosis, the feature extraction is generally complex, the computational burden of the diagnostic model is large, it is difficult to run online, and there is a lack of indication of unknown faults. Summary of the invention

[0005] In view of the problems mentioned in the background technology that the data-driven diagnosis method has high complexity, is difficult to operate online, and has unknown fault indications, the present invention provides a comprehensive diagnosis method for open-circuit faults in a permanent magnet synchronous motor drive system.

[0006] To solve the above problems, the technical solution proposed by the present invention is as follows:

[0007] A comprehensive diagnostic method for open circuit faults of a permanent magnet synchronous motor drive system, comprising:

[0008] Step 1: obtain the six-phase current of the permanent magnet synchronous motor through the current sensor, transform the six-phase current into the xy coordinate system, calculate the fault occurrence indicator variable through the x-axis current and y-axis current obtained after the coordinate transformation, and compare it with the fault indicator variable threshold to determine the occurrence of the fault;

[0009] Step 2, normalizing the six-phase currents and calculating the positive and negative half-cycle average values ​​of the normalized six-phase current values;

[0010] Step 3, support vector machine diagnostic model training: according to the one-to-rest training principle, 24 binary support vector machine classifiers are established, and they are trained using the support vector machine diagnostic model training set to obtain the support vector machine diagnostic model; the training set of the support vector machine diagnostic model comes from the positive and negative half-cycle average values ​​calculated in step 2;

[0011] Step 4: Fault diagnosis: After a fault occurs in the motor drive system, the six-phase current collected within a fundamental cycle is first extracted to obtain feature data. Feature extraction includes normalization processing and calculation of the average value of the positive and negative half cycles. The feature data is then input into the support vector machine diagnosis model for diagnosis to determine the specific fault.

[0012] Compared with the prior art, the present invention has the following beneficial effects:

[0013] (1) Compared with the existing fault diagnosis methods, the data-driven low-complexity online fault diagnosis method for permanent magnet synchronous motor drive system proposed in the present invention can identify more types of faults and improve the limitation of single fault diagnosis type.

[0014] (2) The present invention uses the average values ​​of positive and negative half cycles to reduce the dimension of fault features, which significantly improves the problem of computational complexity compared to spectrum analysis.

[0015] (3) The present invention further simplifies and optimizes the diagnostic model. Compared with the traditional offline method, the diagnostic model can be conveniently deployed in a conventional industrial microcontroller and is applicable to various working conditions.

[0016] (4) The present invention not only takes known faults into consideration, but also determines and indicates unknown faults through decision values, thereby optimizing the problem that traditional data-driven methods cannot identify unknown faults and lead to misdiagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A control and diagnosis block diagram of a dual three-phase permanent magnet synchronous motor drive system according to an embodiment of the present invention.

[0018] Figure 2 The present invention is a flow chart of a comprehensive diagnostic method for an open-circuit fault of a permanent magnet synchronous motor drive system.

[0019] Figure 3 1 is a current waveform and a diagnostic response when a current sensor fault occurs according to an embodiment of the present invention, wherein (a) shows the ABC phase current; (b) shows the DEF phase current; (c) shows the fault occurrence indication variable I xyChanges; (d) shows the fault location label and fault type label. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention, that is, the embodiments described only represent selected embodiments of the present invention, not all embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0021] like Figure 1 As shown, this embodiment relates to a dual three-phase motor drive system, including a reference signal generating unit, a pulse width modulation unit, a power driving unit, a position speed detection unit, a current conversion unit and a fault detection unit, wherein the reference signal generating unit generates a current reference value through And the feedback current Calculate the reference voltage ; The pulse width modulation unit passes the reference voltage The pulse width modulation signal is calculated; the power drive unit feeds power to the motor through the pulse width modulation signal, and the six-phase current feedback value It is obtained by the sampling module including the current sensor, i represents the current, Indicates phase sequence; Position and speed detection unit detects the rotor position and motor speed of the motor, and Output to the current conversion unit to convert the motor speed Output to the reference signal generation unit; the current conversion unit performs coordinate transformation on the six-phase current (also called sampling current) to obtain the xy plane current components in the rotating coordinate system , and outputs feedback current to the reference signal generation unit The fault detection unit receives the current from the current conversion unit. , receives phase current from the power drive unit, performs fault detection, and outputs specific faults.

[0022] This embodiment involves Figure 1 The comprehensive diagnostic method for open circuit fault of permanent magnet synchronous motor drive system is shown in FIG. Figure 2 As shown, the method specifically includes:

[0023] Step 1: obtain the six-phase current of the permanent magnet synchronous motor through the current sensor, transform the six-phase current into the xy coordinate system, calculate the fault occurrence indicator variable through the x-axis current and y-axis current obtained after the coordinate transformation, and compare it with the fault indication variable threshold to determine the occurrence of the fault.

[0024] The faults include phase failure, switch tube open circuit fault and current sensor fault, totaling 24 fault modes, each of which includes different operating conditions. The different operating conditions refer to different speeds and loads.

[0025] The six-phase current is transformed into the xy coordinate system, which includes: the coordinate transformation of the sampled current can obtain the xy plane current component in the rotating coordinate system. , specifically:

[0026] ,

[0027] Normal and fault conditions :Since the control strategy performs closed-loop control on the xy plane, under normal circumstances and are all approximately 0, and then the fault occurrence detection can be performed by establishing the fault indicator variable. The calculation of the fault occurrence indicator variable by the x-axis current and the y-axis current obtained after the coordinate transformation in step 1 includes:

[0028] ,

[0029] Indicates the fault occurrence indicator variable. Under normal circumstances The calculation result is approximately 0. When a fault occurs will increase immediately, such as Figure 3 Therefore, the present invention sets a fault indication variable threshold , by real-time calculation and observation of fault indicator variables , and then with the fault indicator variable threshold In contrast, when the fault occurs, the observed value of the indicator variable , it is judged that the dual three-phase motor drive system has a fault, otherwise there is no fault; once it is judged that a fault has occurred, the specific fault type and location are further identified, and the six-phase current is collected and input into the trained and optimized support vector machine diagnosis model for identification and output.

[0030] Step 2: normalize the six-phase currents and calculate the positive and negative half-cycle average values ​​of the normalized six-phase current values.

[0031] Normalizing the six-phase current includes: after detecting a fault, the real-time data of the six-phase current collected by the current sensor and the q-axis reference current of the motor in the rotating coordinate system , normalize the six-phase currents:

[0032] ,

[0033] in, It is the six-phase current feedback value, which is collectively referred to as , is the normalized six-phase current value.

[0034] The normalized six-phase current values ​​refer to: within one fundamental wave cycle of the phase current, the average values ​​are accumulated and calculated according to the current polarity.

[0035] Furthermore, the average value of the positive and negative half cycles of a fundamental wave cycle is calculated as the input of the support vector machine diagnosis model:

[0036] ,

[0037] in, It represents the average value of the six-phase current in the positive half cycle after normalization. It represents the average value of the negative half cycle after the six-phase current is normalized, and T represents the time of one fundamental cycle of the phase current.

[0038] Step 3: Support vector machine diagnostic model training: According to the one-to-rest training principle, 24 binary support vector machine classifiers are established and trained using the support vector machine diagnostic model training set to obtain the support vector machine diagnostic model. The training set of the support vector machine diagnostic model comes from the average values ​​of the positive and negative half cycles calculated in step 2.

[0039] The support vector machine diagnosis model includes a one-to-many support vector machine, which uses a one-to-many support vector machine (SVM) algorithm, uses known fault data and labels to perform offline training on the one-to-many support vector machine to establish a fault classification model, optimizes the output rules of the fault classification model according to the uniqueness and difference of the fault, and establishes a fault diagnosis model. By inputting the normalized positive and negative half-cycle average values ​​of the phase current to be diagnosed online into the trained and optimized fault diagnosis model, accurate identification of 24 faults (6 phase loss faults, 12 switch tube open circuit faults and 6 current sensor faults) in the motor drive system is achieved.

[0040] The one-to-many support vector machine uses a linear kernel function. The one-to-many principle means that the 24 types of data are classified into 1 type and the remaining 23 types to obtain a binary classification training set, totaling 24. The training method of the support vector machine is to input each binary classification training set into the support vector machine diagnosis model according to the +1 and -1 labels. The 24 faults are divided into 24 groups according to 1:23 for training, and finally a support vector machine diagnosis model containing 24 decision functions can be trained. The 24 decision functions represent 24 sub-classifiers, and each sub-classifier is a binary support vector machine classifier. The binary support vector machine classifier can be represented by a decision function:

[0041] ,

[0042] in, is the current characteristic value, and Obtained through training, is the decision function value.

[0043] Step 4: Fault diagnosis: After a fault occurs in the motor drive system, the six-phase current collected within a fundamental cycle is first extracted to obtain feature data. Feature extraction includes normalization processing and calculation of the average value of the positive and negative half cycles. The feature data is then input into the support vector machine diagnosis model for diagnosis to determine the specific fault.

[0044] Step 4 may include: inputting the characteristic data value into the support vector machine diagnosis model to obtain 24 decision values, and comprehensively counting and judging to obtain a diagnosis label or an unknown type of fault, including:

[0045] By inputting the characteristic data into 24 decision functions for calculation, 24 fault decision values ​​can be obtained.

[0046] The fault decision value is a real number, which can be positive or negative. The fault decision value represents the classification result of 24 sub-classifiers and can be used Indicates, 1 means it belongs to one of the categories, and 0 means it belongs to the remaining 23 categories:

[0047] ,

[0048] Support vector machine diagnostic model judgment output: Based on the 24 classification results obtained by calculation, analysis is performed according to the following judgment conditions:

[0049] ,

[0050] in, Represents the sub-classification result, i represents the classifier label 1-24, Indicates that it is not the fault corresponding to i. Indicates the fault corresponding to i, Indicates the final diagnosis result.

[0051] When the output result of one and only one classifier is 1, the diagnostic label is the fault type corresponding to the classifier. In other cases, the output is -1, indicating that the fault is a fault with an unknown label, thereby avoiding misdiagnosis.

[0052] In this embodiment, fault location and fault type judgment are separately represented, and the fault type and occurrence location are calibrated. In the location label, 1-6 represent the six phases of ABCDEF respectively, and 0 represents an unknown position fault; in the type label, 1-4 represent a phase loss fault, an upper tube open circuit fault of the switch tube, a lower tube open circuit fault of the switch tube, and a current sensor fault respectively, and 0 represents an unknown type fault.

[0053] Figure 3 1 is a current waveform and a diagnostic response when a current sensor fault occurs according to an embodiment of the present invention, wherein (a) shows the ABC phase current; (b) shows the DEF phase current; (c) shows the fault occurrence indication variable I xy (d) shows the fault location label and fault type label. Figure 3 As shown, when the motor drive system operates normally, I xy Approximately 0, when the fault occurs (time = 20.5ms), I xy Increase immediately, when it is greater than the preset threshold I xy_th After that, data collection and processing are started. After data collection and processing for one fundamental wave cycle, the obtained feature data is input into the support vector machine diagnosis model for identification. As shown in the figure, the position label indicates 2, indicating that the fault occurs in phase B; the type label indicates 4, indicating that the fault is a current sensor fault. Therefore, the fault in this embodiment is a phase B current sensor fault. Excluding the data collection process, the present invention can complete fault identification in an instant.

[0054] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed in the specific embodiment, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be within the scope of the concept described in the present invention, through the above teachings to change the technology or knowledge in the relevant field. The changes and modifications made by those skilled in the art do not deviate from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A comprehensive diagnostic method for open circuit faults in a permanent magnet synchronous motor drive system, characterized in that: include: Step 1: obtain the six-phase current of the permanent magnet synchronous motor through the current sensor, transform the six-phase current into the xy coordinate system, calculate the fault occurrence indicator variable through the x-axis current and y-axis current obtained after the coordinate transformation, and compare it with the fault indicator variable threshold to determine the occurrence of the fault; Step 2, normalizing the six-phase currents and calculating the positive and negative half-cycle average values ​​of the normalized six-phase current values; Step 3, support vector machine diagnostic model training: according to the one-to-rest training principle, 24 binary support vector machine classifiers are established, and they are trained using the support vector machine diagnostic model training set to obtain the support vector machine diagnostic model; the training set of the support vector machine diagnostic model comes from the positive and negative half-cycle average values ​​calculated in step 2; Step 4: Fault diagnosis: After a fault occurs in the motor drive system, the six-phase current collected within a fundamental cycle is first extracted to obtain feature data. Feature extraction includes normalization processing and calculation of the average values ​​of the positive and negative half cycles. The feature data is then input into the support vector machine diagnosis model for diagnosis to determine the specific fault.

2. The comprehensive diagnosis method for open circuit fault of permanent magnet synchronous motor drive system according to claim 1 is characterized in that: Transforming the six-phase current into the xy coordinate system includes: performing coordinate transformation on the six-phase current to obtain the xy plane current components in the rotating coordinate system , : ， in, is the six-phase current feedback value; The calculation of the fault occurrence indicator variable by using the x-axis current and the y-axis current obtained after the coordinate transformation includes: ， Indicates the fault occurrence indicator variable; Setting the fault indicator variable threshold , when the observed value of the indicator variable is , it is judged that the dual three-phase motor drive system has a fault, otherwise there is no fault.

3. The comprehensive diagnosis method for open circuit fault of permanent magnet synchronous motor drive system according to claim 2 is characterized in that: Normalizing the six-phase current includes: after detecting a fault, the real-time data of the six-phase current collected by the current sensor and the q-axis reference current of the motor in the rotating coordinate system , normalize the six-phase currents: ， in, is the six-phase current value, is the normalized six-phase current value; Calculate the average value of the positive and negative half cycles of a fundamental cycle of the normalized six-phase current value as the input of the support vector machine diagnosis model: ， in, It represents the average value of the six-phase current in the positive half cycle after normalization. It represents the average value of the negative half cycle after the six-phase current is normalized, and T represents the time of one fundamental cycle of the phase current.

4. The comprehensive diagnosis method for open circuit fault of permanent magnet synchronous motor drive system according to claim 3 is characterized in that: The one-to-rest principle means that the 24 types of data are classified into one type and the other 23 types to obtain a binary classification training set. Each binary classification training set is input into the support vector machine diagnosis model according to the +1 and -1 labels to obtain a classification decision model containing 24 decision functions. The 24 decision functions represent 24 sub-classifiers, each of which is a binary support vector machine classifier. The binary support vector machine classifier is represented by the decision function: ， Among them, x j is the current characteristic value, w ij and b j Obtained through training, is the decision function value.

5. The comprehensive diagnosis method for open circuit fault of permanent magnet synchronous motor drive system according to claim 4 is characterized in that: Step 4 includes: inputting the characteristic data into 24 decision functions for calculation to obtain 24 fault decision values, using r i Indicates, 1 means it belongs to one of the categories, and 0 means it belongs to the remaining 23 categories: ， Support vector machine diagnosis model judgment output: According to the 24 fault decision values ​​calculated, analysis is performed based on the following judgment conditions: ， Among them, r i Indicates the sub-classification result, i indicates the classifier label 1-24, 0 indicates that the fault does not correspond to i, and 1 indicates that the fault corresponds to i. Indicates the final diagnosis result.

Images