Permanent magnet synchronous motor driver fault diagnosis method based on fault classifier

Abstract

The invention belongs to the technical field of motor testing, and particularly relates to a permanent magnet synchronous motor driver fault diagnosis method based on a fault classifier, which comprises the following steps: in a diagnosis period, acquiring operation environment data of an environment where a permanent magnet synchronous motor driver is located, and judging whether the operation environment of the permanent magnet synchronous motor driver is normal or not; if the operation environment is abnormal, acquiring real-time operation data of the permanent magnet synchronous motor driver, and inputting the real-time operation data into a fault classifier pre-stored in a database; if the operation environment is normal, evaluating the instruction completion condition of the permanent magnet synchronous motor driver, and judging whether the instruction completion condition of the permanent magnet synchronous motor driver is normal or not; and if the instruction completion condition is normal, acquiring real-time operation data of the permanent magnet synchronous motor driver, and judging whether the operation state of the permanent magnet synchronous motor driver is normal or not. According to the invention, the efficiency and accuracy of fault diagnosis are improved, and corresponding maintenance or repair measures can be taken in time.

Description

Technical Field

[0001] The invention belongs to the technical field of motor testing, and in particular relates to a permanent magnet synchronous motor driver fault diagnosis method based on a fault classifier. Background Art

[0002] Permanent magnet synchronous motors are widely used in AC speed control systems due to their small size, high power density, and high efficiency. At the same time, condition monitoring and fault diagnosis technologies that can achieve preventive maintenance of motor drives (inverters) have also received widespread attention. Motor drives are affected by many different types of faults, such as stator faults, rotor faults, mechanical faults, sensor faults, and power switching device faults. Fault diagnosis of permanent magnet synchronous motor drives involves detecting and identifying various faults that may occur in the motor, including motor winding faults, bearing damage, rotor bar breakage, etc. These faults affect the efficiency and performance of the motor drive and may even cause the motor to shut down.

[0003] Some existing permanent magnet synchronous motor drive fault diagnosis technologies have the problem of difficulty in combining operating environment data and real-time operating data to quickly and accurately diagnose permanent magnet synchronous motor drive faults in a complex and changing operating environment. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a permanent magnet synchronous motor drive fault diagnosis method based on a fault classifier, which can effectively distinguish between environmental anomalies, instruction execution anomalies and abnormal operating status of the motor itself, thereby achieving accurate judgment of the motor fault type.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: The permanent magnet synchronous motor drive fault diagnosis method based on the fault classifier comprises the following steps: In a diagnostic cycle, the operating environment data of the environment in which the permanent magnet synchronous motor drive is located is obtained, and based on the operating environment data, it is determined whether the operating environment of the permanent magnet synchronous motor drive is normal. The operating environment data includes ambient temperature, electromagnetic interference level, and power supply stability: If the operating environment is abnormal, the real-time operating data of the permanent magnet synchronous motor drive is obtained and input into the fault classifier pre-stored in the database. The real-time operating data includes motor winding impedance, circuit harmonic content, DC link voltage and rotor offset; If the operating environment is normal, evaluate the command completion status of the permanent magnet synchronous motor driver to determine whether the command completion status of the permanent magnet synchronous motor driver is normal: If the instruction completion is abnormal, real-time operation data of the permanent magnet synchronous motor driver is obtained, and the real-time operation data is input into a fault classifier pre-stored in the database; If the instruction is completed normally, the real-time operation data of the permanent magnet synchronous motor driver is obtained, and whether the operation status of the permanent magnet synchronous motor driver is normal is determined based on the real-time operation data: If the operating status is normal, enter the next diagnostic cycle; If the operating state is abnormal, real-time operating data of the permanent magnet synchronous motor driver is obtained, and the real-time operating data is input into a fault classifier pre-stored in a database; The fault classifier pre-stored in the database diagnoses the real-time operation data to determine the fault type of the permanent magnet synchronous motor drive.

[0006] Preferably, judging whether the operating environment of the permanent magnet synchronous motor driver is normal based on the operating environment data comprises the following steps: Obtain the operating environment parameter data stored in the database, including the ambient parameter temperature, the parameter electromagnetic interference level and the parameter power supply stability; The operating environment data is integrated and analyzed with the operating environment parameter data to obtain the operating environment evaluation index; Determine whether the operating environment assessment value is greater than the operating environment assessment threshold stored in the database: If the operating environment evaluation value is greater than the operating environment evaluation threshold stored in the database, the operating environment is abnormal; If the operating environment evaluation value is not greater than the operating environment evaluation threshold stored in the database, the operating environment is normal.

[0007] Preferably, the method for obtaining the operating environment evaluation value is as follows: ; Wherein, Yp is the operating environment assessment value, WsH is the ambient temperature, Ds is the electromagnetic interference level, Ws is the power supply stability, WcH is the ambient parameter temperature, Dc is the parameter electromagnetic interference level, and Wc is the parameter power supply stability.

[0008] Preferably, evaluating the instruction completion status of the permanent magnet synchronous motor driver to determine whether the instruction completion status of the permanent magnet synchronous motor driver is normal includes the following steps: Obtain the command completion status data of the permanent magnet synchronous motor drive, including the average command response time, dynamic response time and steady-state achievement time; Obtain instruction completion parameter data stored in the database, including instruction parameter average response time, parameter dynamic response time, and parameter steady-state achievement time; Obtain the allowable deviation data of the instruction completion status, including the allowable deviation value of the dynamic response time and the allowable deviation value of the steady-state achievement time; Obtaining an instruction completion evaluation value based on instruction completion data, instruction completion parameter data, and instruction completion allowable deviation data; Determine whether the instruction completion evaluation value is greater than the instruction completion evaluation threshold stored in the database: If the instruction completion evaluation value is greater than the instruction completion evaluation threshold stored in the database, the instruction completion is abnormal; If the instruction completion status evaluation value is not greater than the instruction completion status evaluation threshold stored in the database, the instruction completion status is normal.

[0009] Preferably, obtaining instruction completion status allowable deviation data includes the following steps: Based on the device information of the permanent magnet synchronous motor drive, a matching data set is obtained from a database, where the matching data set includes a plurality of deviation value matching data, and the deviation value matching data includes an electromagnetic interference matching value, a power supply stability matching value, and a daily average temperature of the drive; Obtain the data to be matched of the permanent magnet synchronous motor drive, including the electromagnetic interference level, power supply stability and actual temperature of the drive; Compare the data to be matched with each deviation value matching data one by one, and determine the deviation value matching data closest to the data to be matched; The instruction completion status allowable deviation data corresponding to the deviation value matching data stored in the database is obtained.

[0010] Preferably, comparing the data to be matched with each deviation value matching data one by one to determine the deviation value matching data closest to the data to be matched comprises the following steps: Obtaining a matching value after comparing the data to be matched with each deviation value matching data; Determine the minimum matching value; The deviation value matching data corresponding to the minimum matching value is output as the deviation value matching data closest to the data to be matched.

[0011] Preferably, the method for obtaining the matching value is as follows: ; Where PP is the matching value, Ds is the electromagnetic interference level, Ws is the power supply stability, Qsw is the actual temperature of the driver, Dp is the electromagnetic interference matching value, Wp is the power supply stability matching value, Qcw is the daily average temperature of the driver, e is a natural constant, and ln is a logarithmic function, which means the logarithm with e as the base.

[0012] Preferably, judging whether the operating state of the permanent magnet synchronous motor driver is normal based on the real-time operating data comprises the following steps: Obtain the operating status parameter data of the permanent magnet synchronous motor drive stored in the database, including the motor winding parameter impedance, the parameter circuit harmonic content, the DC link parameter voltage and the rotor parameter offset; Obtaining the allowable deviation data of the operating status, including the allowable deviation value of the motor winding parameter impedance and the allowable deviation value of the DC link parameter voltage; The operating status of the permanent magnet synchronous motor drive is evaluated based on real-time operating data, operating status parameter data and operating status allowable deviation data to determine whether the operating status is normal.

[0013] Preferably, the operating state of the permanent magnet synchronous motor driver is evaluated based on the real-time operating data, the operating state parameter data and the operating state allowable deviation data to determine whether the operating state is normal, including the following steps: Obtaining an operation status evaluation index based on real-time operation data, operation status parameter data, and operation status allowable deviation data; Determine whether the running status evaluation index is greater than the running status evaluation threshold stored in the database: If the running status evaluation index is greater than the running status evaluation threshold stored in the database, the running status is abnormal; If the running status evaluation index is not greater than the running status evaluation threshold stored in the database, the running status is normal.

[0014] Preferably, obtaining the operating state allowable deviation data comprises the following steps: Obtain the operating status matching data of the permanent magnet synchronous motor drive, including the average daily continuous use time of the drive, the electromagnetic interference level and the power supply stability; Compare the running status matching data with each state deviation matching data stored in the database one by one to obtain the comparison coefficient. The state deviation matching data includes the matching parameters of the average continuous use time of the drive per day, the electromagnetic interference level matching parameters and the power supply stability matching parameters; Determine the minimum comparison coefficient, and extract the operating status allowable deviation data corresponding to the minimum comparison coefficient from the database.

[0015] The present invention has the following beneficial effects: The present invention monitors operating environment data and real-time operating data and uses a pre-stored fault classifier for analysis. This method can effectively distinguish between environmental anomalies, instruction execution anomalies and abnormal operating status of the motor itself, thereby achieving accurate judgment of the motor fault type, solving the technical problem of quickly and accurately diagnosing permanent magnet synchronous motor drive faults in complex and changing operating environments, not only improving the efficiency and accuracy of fault diagnosis, but also helping to take corresponding maintenance or repair measures in a timely manner, reducing downtime and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a framework flow chart of the JITGP-ELM method in an embodiment of the present invention; Figure 3 Schematic diagram of the network structure of the extreme learning machine in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The embodiments of the present invention are further described below in conjunction with the accompanying drawings: This embodiment uses a permanent magnet synchronous motor driver fault diagnosis method based on a fault classifier to first determine whether the operating environment of the motor driver is normal, and the factors considered include ambient temperature, electromagnetic interference level and power supply stability. This step is basic because environmental factors often have a significant impact on the normal operation of the equipment. If the environment is abnormal, it directly enters the stage of obtaining the real-time operating data of the motor, such as motor winding impedance, circuit harmonic content, etc., to determine the type of fault. If the environment is normal, check whether the command completion of the motor driver meets expectations. According to the real-time operating data and command completion of the motor, it is determined whether the operating state of the motor is normal. If any abnormal state is found, the real-time data is input into a pre-set fault classifier for analysis to diagnose the specific fault type.

[0018] like Figure 1 As shown, a permanent magnet synchronous motor drive fault diagnosis method based on a fault classifier includes the following steps: within a diagnosis cycle, the operating environment data of the environment in which the permanent magnet synchronous motor drive is located is obtained, the operating environment data includes the ambient temperature (using a thermistor, a thermocouple or an infrared sensor to monitor the ambient temperature, these sensors can accurately measure the temperature of the drive and its surrounding environment, and transmit the data to the monitoring system in real time, wherein the thermistor can be a PT100 or NTC thermistor), the electromagnetic interference level (using a special electromagnetic interference detection device, such as a spectrum analyzer, to measure the electromagnetic noise level in the environment) and the power supply stability (using a voltage sensor and a current sensor to monitor the voltage or current stability of the power supply line, such as the voltage mean square error or the current mean square error), based on the operating environment data, judging whether the operating environment of the permanent magnet synchronous motor drive is normal, including the following steps: First, the real-time operating data of the environment in which the permanent magnet synchronous motor drive is located is collected, and the outputs of all sensors are connected to a data acquisition system. The system is responsible for collecting, storing and preprocessing data, and using appropriate communication protocols (such as Modbus, CAN bus or Ethernet) to send the collected data to the central control system or cloud platform for further analysis and processing. Then, the operating environment parameter data stored in the database is obtained, and the operating environment parameter data includes the environmental parameter temperature, the parameter electromagnetic interference level and the parameter power supply stability; the operating environment data is integrated and analyzed with the operating environment parameter data to obtain the operating environment evaluation index; it is determined whether the operating environment evaluation value is greater than the operating environment evaluation threshold stored in the database: if the operating environment evaluation value is greater than the operating environment evaluation threshold stored in the database, the operating environment is abnormal; if the operating environment evaluation value is not greater than the operating environment evaluation threshold stored in the database, the operating environment is normal.

[0019] Timely identification of adverse operating environmental conditions, such as excessive temperature, excessive electromagnetic interference or unstable power supply, can have a negative impact on motor performance. Through early identification and response, possible failures or damage can be prevented, thus avoiding more serious consequences.

[0020] The method for obtaining the operating environment evaluation value is as follows: ; In the formula, Yp is the operating environment assessment value, WsH is the ambient temperature, Ds is the electromagnetic interference level, Ws is the power supply stability, WcH is the ambient temperature, Dc is the electromagnetic interference level, and Wc is the power supply stability. This formula is designed to evaluate the operating environment quality of permanent magnet synchronous motor drives, taking into account three key factors: ambient temperature, electromagnetic interference level, and power supply stability. Each parameter directly affects the performance and life of the motor, so a comprehensive evaluation of these parameters is essential to ensure reliable operation of the motor.

[0021] Ambient temperature has a direct impact on the thermal state of the motor. High temperatures may cause the motor to overheat, affect insulation performance, and increase the risk of failure. Therefore, it is critical to monitor the temperature difference from the reference standard. Electromagnetic interference can affect the normal operation of the electronic equipment of the motor drive system, such as the microprocessor and other sensitive electronic components of the control system. Therefore, evaluating the ratio of actual interference to Dc is key to determining whether the environment is suitable for normal operation. Power supply stability directly affects the performance and safety of the motor. Power supply fluctuations may cause abnormal operation or damage to the motor control system, so it is necessary to monitor the ratio of actual power supply to reference power supply stability. Using a combination of absolute value differences and ratios, this design aims to quantify the specific contribution of each parameter to environmental adaptability and ensure that significant deviations from any one parameter can be captured and reflected in the total evaluation value.

[0022] By comparing real-time data to preset standards, this approach provides a quantitative and precise way to assess the environment the motor is operating in. This precise matching ensures operation under ideal conditions, thereby increasing the efficiency and life of the equipment.

[0023] If the operating environment is abnormal, the real-time operating data of the permanent magnet synchronous motor driver is obtained and input into the fault classifier pre-stored in the database. The real-time operating data includes motor winding impedance, circuit harmonic content, DC link voltage and rotor offset. If the operating environment is normal, the instruction completion status of the permanent magnet synchronous motor driver is evaluated to determine whether the instruction completion status of the permanent magnet synchronous motor driver is normal. By comparing the actual instruction response time with the set standard, it can be confirmed whether the motor driver is working normally within the designed performance parameter range.

[0024] Obtain the instruction completion status data of the permanent magnet synchronous motor driver, which includes the average instruction response time (the average time required for the motor driver to start executing after receiving the instruction), the dynamic response time (the time for the motor to accelerate to the target state, the target state is to reach the set speed or position, etc.) and the steady-state achievement time (the time required for the motor to stabilize in the state after reaching the target state); obtain the instruction completion status parameter data stored in the database, which includes the instruction parameter average response time, parameter dynamic response time and parameter steady-state achievement time. These data are determined based on performance tests, manufacturer specifications or empirical standards, and provide a benchmark for evaluating whether the current system performance is within an acceptable range.

[0025] Get the instruction completion tolerance data, which defines the acceptable deviation range of dynamic response time and steady-state time. This is important because in actual operation, slight performance fluctuations may occur due to various factors (such as load changes, power supply fluctuations, etc.). The instruction completion tolerance data includes the dynamic response time tolerance value and the steady-state time tolerance value.

[0026] Based on the device information of the permanent magnet synchronous motor drive, a matching data set is obtained from the database, and the matching data set includes multiple deviation value matching data, and the deviation value matching data includes electromagnetic interference matching value, power supply stability matching value and daily average temperature of the drive; the data to be matched of the permanent magnet synchronous motor drive is obtained, and the data to be matched includes electromagnetic interference level (electromagnetic interference can interfere with the control system of the drive, especially affecting sensor readings and communication buses. This interference may cause misjudgment or delay in processing control signals, thereby affecting the dynamic response time. High electromagnetic interference levels may cause transient errors in the control system and affect the ability of the device to reach a steady state), power supply stability (unstable power supply will cause fluctuations in the motor power supply, which not only affects the starting and acceleration process of the motor, but also may cause fluctuations when trying to reach a stable operating state, thereby extending the dynamic response time and the steady state reaching time) and the actual temperature of the drive (temperature increase may affect the performance of the insulation material inside the motor, reduce motor efficiency, and increase resistance, thereby affecting the response time and stability of the motor. At high temperatures, the drive may take longer to reach a stable operating state because factors such as thermal expansion may affect the operation of mechanical components). By monitoring these parameters, the actual performance of the permanent magnet synchronous motor drive under specific environmental and power supply conditions can be better understood. This helps set more realistic and feasible performance benchmarks, such as the allowable deviation values ​​for dynamic response time and steady-state achievement time.

[0027] The data to be matched are compared with each deviation value matching data one by one to determine the deviation value matching data closest to the data to be matched; and the instruction completion status allowable deviation data corresponding to the deviation value matching data stored in the database is obtained.

[0028] Determining the deviation value matching data closest to the data to be matched includes the following steps: obtaining the matching value after comparing the data to be matched with each deviation value matching data; determining the minimum matching value, and selecting the historical data point that is most similar to the current condition (i.e., the minimum matching value). This indicates that under similar conditions, the performance and reaction of the motor are known and can be predicted and evaluated; outputting the deviation value matching data corresponding to the minimum matching value as the deviation value matching data closest to the data to be matched, and using this as a basis to predict the current performance condition or perform fault diagnosis.

[0029] The matching value is obtained as follows: ; Where PP is the matching value, Ds is the electromagnetic interference level, Ws is the power supply stability, Qsw is the actual temperature of the driver, Dp is the electromagnetic interference matching value, Wp is the power supply stability matching value, Qcw is the daily average temperature of the driver, e is a natural constant, and ln is a logarithmic function, which means the logarithm with e as the base.

[0030] For example, Ds=0.5dB, Ws is expressed as voltage standard deviation, which is 2.5V, and Qsw=25℃. The following table shows the matching values ​​of three sets of deviation matching data and the data to be matched: Table 1 Matching values ​​of three sets of deviation value matching data and data to be matched

[0031] It can be seen from the data table that the third group of deviation matching data has the smallest matching value.

[0032] Using absolute difference and , respectively, calculate the difference between electromagnetic interference and power supply stability. The absolute value difference emphasizes the direct deviation between the two quantities, whether positive or negative. Temperature difference The natural logarithm function is used for processing, in which the addition operation ensures that the logarithm function is always positive, avoiding calculation errors or meaningless results. The exponential function of the natural constant is used to process the combined deviation of electromagnetic interference and power supply stability, and the impact of the deviation value change on the final matching value is improved, so that even small changes can significantly affect the results. The temperature difference is processed through logarithmic operation, which smooths the impact of temperature difference on the matching value and avoids excessive impact of temperature abnormality on the overall matching result.

[0033] Obtain an instruction completion status evaluation value based on instruction completion status data, instruction completion status parameter data and instruction completion status allowable deviation data; determine whether the instruction completion status evaluation value is greater than the instruction completion status evaluation threshold stored in the database: if the instruction completion status evaluation value is greater than the instruction completion status evaluation threshold stored in the database, the instruction completion status is abnormal; if the instruction completion status evaluation value is not greater than the instruction completion status evaluation threshold stored in the database, the instruction completion status is normal.

[0034] The method for obtaining the instruction completion evaluation value is as follows: ; Where Zp is the instruction completion evaluation value, Tsp is the average instruction response time, Tsd is the dynamic response time, Tsw is the steady-state reaching time, Tcp is the instruction parameter average response time, Tcd is the parameter dynamic response time, and Tcw is the parameter steady-state reaching time. is the allowable deviation value of dynamic response time, It is the allowable deviation value of the time to reach steady state.

[0035] The calculation formula of the command completion evaluation value combines key performance indicators to evaluate the response efficiency and stability of the permanent magnet synchronous motor drive. This formula covers three main parameters: average command response time, dynamic response time and steady-state achievement time, compared with their respective parameter values, and taking into account the corresponding allowable deviation. Compare the actual average response time with the set average response time. The smaller Tsp is, the faster the response speed is.

[0036] Considering the combination of differences and deviations, if the dynamic response time is too fast, it may mean that the system changes state quickly after receiving the command, which may cause greater pressure and load on mechanical parts, electronic components and software processing. This may reduce the service life and reliability of the equipment in the long run. Long response time will affect production efficiency, especially in high-speed production lines or operating environments that require quick response. Response delays may cause slow or stagnant production processes. Reaching steady state too quickly may mean that the system does not have enough time to make appropriate adjustments, which may lead to insufficient system stability, such as large fluctuations or adjustment requirements after reaching steady state. If the steady state takes too long to reach, the system may be in a non-optimal working state for a long time, reducing operating efficiency and output.

[0037] If the instruction completion is abnormal, the real-time operation data of the permanent magnet synchronous motor driver is obtained, and the real-time operation data is input into the fault classifier pre-stored in the database; if the instruction completion is normal, the real-time operation data of the permanent magnet synchronous motor driver is obtained, and whether the operation state of the permanent magnet synchronous motor driver is normal is judged based on the real-time operation data, including the following steps: The operating state parameter data of the permanent magnet synchronous motor drive stored in the database are obtained. The operating state parameter data include motor winding parameter impedance (motor winding impedance is usually measured by a resistance tester or a motor test system. These devices can calculate the impedance by injecting a known current into the motor winding when the motor is running or stopped, and then measuring the voltage response of the motor winding), parameter circuit harmonic content (circuit harmonic content refers to the presence of non-fundamental frequency components in the power supply circuit, which is measured by a harmonic analyzer or a power quality analyzer. These devices can continuously monitor the voltage and current of the power supply system, thereby detecting and recording the harmonic level of the circuit in real time), DC link parameter voltage (DC link voltage usually refers to the voltage level on the DC side of the motor drive. This parameter can be directly measured by a voltage sensor installed in the motor control system) and rotor parameter offset (rotor offset refers to the position offset of the rotor relative to its theoretical center axis. The rotor position is directly monitored using a photoelectric sensor or a magnetic sensor).

[0038] The allowable deviation data of the operating status are obtained. The allowable deviation data of the operating status include the allowable deviation value of the motor winding parameter impedance and the allowable deviation value of the DC link parameter voltage. These deviation values ​​take into account normal fluctuations or slight changes that may occur in actual operation.

[0039] Obtain the operating status matching data of the permanent magnet synchronous motor drive. The operating status matching data includes the average daily continuous use time of the drive (continuous long-term use will cause the motor temperature to rise, affecting the resistance of the motor winding. The resistance increases with the increase in temperature, which requires adjustment of the allowable impedance deviation value to adapt to higher operating temperatures), electromagnetic interference level (high electromagnetic interference level may affect the signal processing of the motor control system, resulting in measurement errors and unstable operation. In a high-interference environment, the impedance measurement of the motor winding may be affected, and the allowable deviation needs to be adjusted to adapt to these conditions) and power supply stability (power supply instability will cause input voltage fluctuations, which will directly affect the stability of the DC link voltage. Voltage fluctuations may cause instability in the internal current of the motor, which in turn affects the stability of the winding impedance. Therefore, when designing the allowable deviation of the motor winding impedance and the DC link voltage, the power supply stability needs to be considered).

[0040] The operating status matching data is compared one by one with each state deviation matching data stored in the database to obtain the comparison coefficient. The state deviation matching data includes the matching parameters of the average daily continuous use time of the drive, the electromagnetic interference level matching parameters and the power supply stability matching parameters. The minimum comparison coefficient is determined, and the operating status allowable deviation data corresponding to the minimum comparison coefficient is extracted from the database.

[0041] The method for obtaining the comparison coefficient is: ; In the formula, i is the number of the state deviation matching data, is the ith comparison coefficient, is the average continuous use time of the drive per day, Matching parameters for the average daily continuous use time of the i-th drive, is the i-th electromagnetic interference level matching parameter, is the i-th power supply stability matching parameter.

[0042] The operating status of the permanent magnet synchronous motor drive is evaluated based on real-time operating data, operating status parameter data and operating status allowable deviation data to determine whether the operating status is normal.

[0043] Obtain an operation status evaluation index based on real-time operation data, operation status parameter data and operation status allowable deviation data; determine whether the operation status evaluation index is greater than the operation status evaluation threshold stored in the database: if the operation status evaluation index is greater than the operation status evaluation threshold stored in the database, the operation status is abnormal; if the operation status evaluation index is not greater than the operation status evaluation threshold stored in the database, the operation status is normal.

[0044] The method for obtaining the operating status evaluation index is as follows: ; Where, YzP is the operating status evaluation index, Rsc is the motor winding impedance, Dsx is the circuit harmonic content, Rsd is the DC link voltage, Dsp is the rotor offset, Rcc is the motor winding parameter impedance, Dcx is the parameter circuit harmonic content, Rcd is the DC link parameter voltage, and Dcp is the rotor parameter offset. The allowable deviation value of the motor winding impedance is determined. It is the allowable deviation value of DC link parameter voltage.

[0045] Changes in impedance can indicate electrical problems in the motor windings, such as insulation damage, poor contact, or overheating. Impedance measurements reflect the integrity and efficiency of the motor's internal electrical loop. The level of harmonic content reflects the quality of the power supply and the tolerance of the motor's electrical system. High harmonic levels can lead to reduced motor efficiency, increased heat, and shortened life, so they are an important indicator of motor operation quality.

[0046] The stability of the DC link voltage directly affects the performance of the motor drive. Voltage fluctuations may cause instability in the motor control system, affecting its response speed and accuracy. The position accuracy of the rotor has a direct impact on the operating efficiency and mechanical wear of the motor. Rotor deviation may cause mechanical vibration and increase bearing load, thereby reducing the operating efficiency and life of the motor.

[0047] Changes in motor winding impedance may affect the power consumption and heat generation of the motor, which in turn affects the stability of the DC link voltage; at the same time, high harmonic content in the power supply may further increase the thermal stress of the windings and increase the impedance changes. These parameters work together to provide a comprehensive reflection of the overall health of the motor in terms of electrical and mechanical performance. For example, an increase in rotor misalignment may cause changes in winding impedance, because changes in mechanical position may affect the electrical characteristics.

[0048] and It reflects the degree of deviation between the actual operating status and the set standard. The use of logarithmic functions can smooth large deviations, avoid excessive influence of extreme values ​​on the evaluation results, and make the evaluation index more robust and reliable. By comprehensively analyzing these parameters, a comprehensive portrait of the motor's operating status can be obtained, and potential electrical and mechanical problems can be discovered in a timely manner. This method not only improves the accuracy of fault diagnosis, but also helps to formulate more effective maintenance strategies to ensure the reliability and long-term stable operation of the motor system. This systematic analysis method makes the maintenance and monitoring of motors more scientific and efficient, helping to reduce unexpected downtime and improve production efficiency.

[0049] If the operating status is normal, the next diagnostic cycle is entered; if the operating status is abnormal, the real-time operating data of the permanent magnet synchronous motor driver is obtained, and the real-time operating data is input into the fault classifier pre-stored in the database; the fault classifier pre-stored in the database diagnoses the real-time operating data to determine the fault type of the permanent magnet synchronous motor driver.

[0050] The fault classifiers stored in the database are built based on just-in-time learning (JITL), such as Figure 2 As shown in the figure, in the proposed fault diagnosis framework, the instant Gaussian process is used to build a model of the healthy system to predict the healthy system behavior, and then form residuals with the behavior of the faulty system. These residuals are used as fault features in the subsequent steps. This process is based on the fault feature extraction of the instant Gaussian process. In the instant Gaussian process, the Gaussian process (GP) is used to build a local model in the instant learning. Since there is a sample selection step in the instant learning, the computational cost of Gaussian process modeling can be significantly reduced in the framework of the instant learning.

[0051] As a popular local learning method, just-in-time learning has the advantage of low computational cost. However, when there are few samples, traditional just-in-time learning based on linear local models may be sensitive to noise, resulting in a contradiction between speed and accuracy in practical applications. An improved algorithm is the just-in-time Gaussian process, whose local model is established by the Gaussian process model, replacing the traditional linear model. Compared with the traditional just-in-time learning algorithm, JITGP has the advantages of smoothness and noise resistance, while inheriting the characteristics of low computational complexity and online adaptability of traditional just-in-time learning. In addition, the GP model can also provide prediction variance to reflect the probability information of the prediction.

[0052] 1) Real-time Gaussian process modeling In the instant Gaussian process, the instant local model is modeled by Gaussian process regression. The Gaussian process model is a non-parametric model in Bayesian statistics. The system modeling of the Gaussian process model includes the following three stages: 11) Gaussian process prior model As a common type of function prior, the Gaussian process model assumes At any point x is a random variable ,in and are independent constants. Therefore, the joint multivariate Gaussian distribution over a set of variables can be expressed as , ..., ,in is a matrix, N is the number of samples, It can be expressed as represent and The variance between and represent any two variables, is the covariance function, and its common form is: (1); Where D is the dimension of variable x, parameter Describe each dimension of x for The importance of is the scale parameter. For an unknown function , ,in is a variance of White noise, that is ,So , ..., , , which follows the Gaussian process prior framework, where if but ,otherwise .

[0053] 12) Hyperparameter Optimization Given N state observations , ..., And the corresponding target output , ..., , then use the probability distribution Represents the likelihood function based on the training data: (2); in Output training data matrix for the target, is the state data training matrix, is the hyperparameter vector, is the covariance matrix of the training data. Therefore, the optimal hyperparameters can be optimized by maximizing the likelihood function: (3); It can be seen that the optimization process requires calculation Partial derivatives with respect to each hyperparameter: 13) Prediction Finding the optimal hyperparameters After that, we can use the relevant training data , i=1,...,N for any state to be tested Make a prediction, and its output is expressed as Therefore, we can get An unbiased estimate of : (4); (5); in, is the covariance vector between the test sample and the training sample, is the covariance between the sample to be tested and itself.

[0054] 2) Fault feature extraction Details of the fault feature extraction method based on instantaneous Gaussian processes, including the selection of relevant samples using similarity estimation, local model estimation using Gaussian processes, and the generation of residual vectors as fault features based on predicted outputs and measurement results.

[0055] 21) Similar sample selection In the instant Gaussian process method, it is necessary to select a part of the data that is most relevant to the sample to be tested to establish a local model, so a similarity evaluation criterion is needed as a standard for selecting data. and The commonly used distance metrics and similarity evaluations are as follows: Distance Metrics: Euclidean distance is a commonly used distance metric that defines the spatial distance between two samples; Manhattan distance, also known as neighborhood distance; Minkowski distance is a generalization of Euclidean distance and Manhattan distance; Similarity evaluation: Cosine similarity considers the vector angle between two samples; it is used to measure the angle between two vectors. The smaller the angle, the greater the cosine similarity.

[0056] Pearson similarity reflects linear similarity.

[0057] The combined similarity criterion consisting of the exponential of the Euclidean distance between two vectors and the cosine similarity is used to evaluate the similarity between the current sample to be tested and the dataset to improve the estimation ability. With historical data samples The similarity between them can be written as: (6); in, is the weight factor, express and The angle between here , , Represents the current sample and The similarity of The bigger the and The higher the similarity between them. In addition, Must be greater than 0, that is, .if , it is considered that the two vectors that make up this angle are dissimilar, and the corresponding database samples are not considered when establishing the local model. .

[0058] After similarity evaluation, a data subset can be constructed using related data , Corresponding to greater similarity ,in The number of samples used to build the instant local model.

[0059] 22) Local model prediction based on Gaussian process Consider a sample state to be predicted, x, whose output is unknown and represented by y. To simplify the calculation, assume that the elements of each dimension in y are are independent of each other. The prediction problem of can be decomposed into the prediction problem of each element: According to the Gaussian process prediction method, after Gaussian process modeling and hyperparameter optimization, we can get Prediction: According to formula (4), we can get It is an unbiased estimate, and the mean square error of this estimate can be calculated by formula (5) . Then we can get the predicted value of y. The implementation of the above algorithm requires the inversion of the covariance matrix K, and its computational complexity is , the storage complexity is , when the amount of training data is large, the computational cost may be very high. This problem can be solved by the natural advantage of just-in-time learning: , i.e. training data The number of is much smaller than the amount of data in the data set N, the dimension of the K matrix will be greatly reduced, so the amount of calculation for its inversion will be greatly reduced.

[0060] 23) Residual calculation In order to generate residuals as fault features for fault detection and diagnosis, the instant Gaussian process simulates the behavior of actual nonlinear and dynamic systems in healthy conditions. The residual is the difference between the output estimate and the actual observed value, which eliminates the dynamics and nonlinearity of the process and extracts the essential information of the fault. It is calculated by the following formula: (7); Among them, y and They are the measured output and predicted output under the current state respectively.

[0061] 3) Fault classification based on extreme learning machine like Figure 3 , when the input of a single hidden layer feedforward neural network is an m-dimensional vector r, its output is: (8); Where L is the number of hidden neurons; is the input layer weight vector connecting the input vector and the i-th hidden layer neuron, m is the dimension of the input vector, is the weight vector connecting the output vector and the i-th hidden layer neuron, and o is the number of output layer nodes; is the hidden layer activation function, represents the bias of the i-th hidden layer neuron, is the output of the i-th hidden layer neuron.

[0062] Then, given N samples , ,in , , we can construct N output results similar to formula (8): (9); in, , , , , , .

[0063] H is the hidden layer output matrix, where The first The output of the jth hidden layer neuron can be used to distribute the random hidden layer neuron parameters. Then the minimum norm and least squares solution of formula (9) are: (10); in is the generalized inverse matrix of H.

[0064] For classification problems, the extreme learning machine can be set to a multi-output structure with o output nodes, where O is the number of categories in the classification problem. When the category label is k, the expected output vector is , the kth element is 1, and the rest are 0. When dealing with multi-classification problems, the index number of the maximum value of the output node is used as the predicted category of the test input. Therefore, when the input is When , the estimated fault state is: (11); in , is the output of the jth node.

[0065] In order to train the extreme learning machine network, a dataset failure dataset is required ,in The state-output data under normal and various fault conditions needs to be included. is the label data corresponding to the fault state. Then, combined with the estimated residual calculated by the instant Gaussian process and the corresponding fault status labels , construct the training data set for the extreme learning machine After determining the network weight parameters and activation function, the training process is completed and the ELM fault classifier can be obtained.

[0066] To more clearly illustrate how to use the ELM fault classifier to perform permanent magnet synchronous motor drive fault diagnosis, refer to the following case: 1. Building an Instant Gaussian Process Model 1. Data preparation Collect a large amount of historical data of permanent magnet synchronous motor drives in normal operation, including operating environment data (ambient temperature, electromagnetic interference level, power supply stability), command completion data (average command response time, dynamic response time, steady-state achievement time) and operating status data (motor winding impedance, circuit harmonic content, DC link voltage, rotor offset). Assume that we collected data every 10 minutes over the past month, and obtained a total of 4320 sets of data samples.

[0067] The data is preprocessed to ensure the accuracy and completeness of the data. For example, the ambient temperature data is detected and corrected for outliers to remove obviously erroneous measurements; the instruction completion data is time-series aligned to ensure that the execution time data of different instructions are on the same time scale.

[0068] 2. Gaussian process prior model setting Assume that the operating status of the motor driver can be described by a function Indicates that x is the input data vector (including the various operating data parameters mentioned above). In the Gaussian process prior model, for any point x, is assumed to be a random variable.

[0069] Determine the covariance function, for example, select the above formula (1). For this embodiment, is 1.5.

[0070] 3. Hyperparameter Optimization Given the above 4320 state observations and the corresponding target outputs (such as a mark indicating whether the motor is operating normally in this state, 0 for normal operation and 1 for failure), a target output training data matrix and a state data training matrix are constructed.

[0071] The probability distribution is used to represent the likelihood function based on the training data, and the hyperparameters are optimized by maximizing the likelihood function. The gradient descent algorithm or other optimization algorithms are used to solve the problem. During the optimization process, the partial derivative of the likelihood function for each hyperparameter is calculated, and the value of the hyperparameter is adjusted according to the direction and size of the partial derivative. After multiple iterations, the optimal hyperparameter value is finally obtained.

[0072] 2. Fault Feature Extraction 1. Similar sample selection When the current running state of the motor driver needs to be diagnosed, the running data at the current moment is obtained as the sample to be tested. The combined similarity between the current sample to be tested and the historical data sample is calculated. For example, a combined similarity criterion composed of the exponential of the Euclidean distance and the cosine similarity is used, formula (6).

[0073] 2. Local model prediction based on Gaussian process Consider the current state of the sample to be tested, whose output is unknown. Decompose the prediction problem into sub-problems of predicting its individual elements. According to the Gaussian process prediction method, the previously established Gaussian process model (with optimized hyperparameters) and the selected related samples are used to calculate the covariance vector between the sample to be tested and the training sample, as well as the covariance between the sample to be tested and itself.

[0074] According to formula (4), we can get It is an unbiased estimate, and the mean square error of this estimate can be calculated by formula (5) , and then we can get the predicted value of y.

[0075] 3. Residual calculation The actual output value of the current motor driver is measured, such as the actual measured motor winding impedance, the actual response time after the command is executed, etc. The residual is calculated, formula (7). This residual contains the difference information between the current system operation state and the predicted state based on the health model, which may indicate a potential fault. For example, if the residual of the calculated motor winding impedance exceeds the normal fluctuation range, it may indicate that there is a problem with the motor winding, such as insulation aging or local short circuit.

[0076] 3. Fault classification Construct an extreme learning machine (ELM) training data set, use the residual vector calculated above as the input feature, and combine it with the corresponding fault state label (determined by actual fault detection or expert judgment). For example, if a certain residual pattern is always accompanied by a motor winding fault in past fault cases, then the fault label corresponding to the residual pattern is a motor winding fault.

[0077] Determine the network structure of ELM, including selecting the appropriate number of hidden layer neurons, activation function, etc. For example, select the number of hidden layer neurons as 50 and the activation function as sigmoid function.

[0078] The input layer weight vector and hidden layer neuron bias are randomly initialized, the hidden layer output matrix is ​​calculated, and then the output layer weight vector is calculated according to formula (10) to complete the ELM training process.

[0079] 2. Fault type determination When new real-time operation data is acquired and the residual is calculated, the residual is input into the trained ELM to obtain the output result. For example, the output of the ELM may be a vector, where each element represents the probability or possibility of a certain fault in the motor drive.

[0080] The fault type is determined based on the output result, and the index number of the maximum value of the output node is usually selected as the predicted category of the test input. If the value of the third element in the output vector is the largest, and the fault type corresponding to the third element is a rotor offset fault, then it is determined that the current motor drive may have a rotor offset fault, thereby achieving accurate classification of the permanent magnet synchronous motor drive fault type.

[0081] In addition, a database including multiple fault types can also be established, in which fault feature data corresponding to various fault types and data on normal operation of the drive are stored. The real-time operation data is input into the database, and similarity analysis is performed between the real-time operation data and each fault feature data (including but not limited to Euclidean distance and Manhattan distance). The fault feature data closest to the real-time operation data (or the data on normal operation of the drive) can be found, that is, the minimum similarity value is determined, so that the fault type can be determined and fault classification can be achieved.

[0082] This embodiment can be implemented based on the following hardware: The electronic device comprises: a processor and a memory, wherein computer program instructions are stored in the memory, and when the computer program instructions are executed by the processor, the processor executes the above permanent magnet synchronous motor drive fault diagnosis method based on the fault classifier.

[0083] A computer-readable storage medium is used to store a program, and when the program is executed by a processor, the permanent magnet synchronous motor drive fault diagnosis method based on a fault classifier as described above is implemented.

Claims

1. A permanent magnet synchronous motor drive fault diagnosis method based on a fault classifier, characterized in that: The following steps are involved: In a diagnostic cycle, the operating environment data of the environment in which the permanent magnet synchronous motor drive is located is obtained, and based on the operating environment data, it is determined whether the operating environment of the permanent magnet synchronous motor drive is normal. The operating environment data includes ambient temperature, electromagnetic interference level, and power supply stability: If the operating environment is abnormal, the real-time operating data of the permanent magnet synchronous motor drive is obtained and input into the fault classifier pre-stored in the database. The real-time operating data includes motor winding impedance, circuit harmonic content, DC link voltage and rotor offset; If the operating environment is normal, evaluate the command completion status of the permanent magnet synchronous motor driver to determine whether the command completion status of the permanent magnet synchronous motor driver is normal: If the instruction completion is abnormal, real-time operation data of the permanent magnet synchronous motor driver is obtained, and the real-time operation data is input into a fault classifier pre-stored in the database; If the instruction is completed normally, the real-time operation data of the permanent magnet synchronous motor driver is obtained, and whether the operation status of the permanent magnet synchronous motor driver is normal is determined based on the real-time operation data: If the operating status is normal, enter the next diagnostic cycle; If the operating state is abnormal, real-time operating data of the permanent magnet synchronous motor driver is obtained, and the real-time operating data is input into a fault classifier pre-stored in a database; The fault classifier pre-stored in the database diagnoses the real-time operation data to determine the fault type of the permanent magnet synchronous motor drive.

2. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 1 is characterized in that: Judging whether the operating environment of the permanent magnet synchronous motor driver is normal based on the operating environment data includes the following steps: Obtain the operating environment parameter data stored in the database, including the ambient parameter temperature, the parameter electromagnetic interference level and the parameter power supply stability; The operating environment data is integrated and analyzed with the operating environment parameter data to obtain the operating environment evaluation index; Determine whether the operating environment assessment value is greater than the operating environment assessment threshold stored in the database: If the operating environment evaluation value is greater than the operating environment evaluation threshold stored in the database, the operating environment is abnormal; If the operating environment evaluation value is not greater than the operating environment evaluation threshold stored in the database, the operating environment is normal.

3. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 2 is characterized in that: The method for obtaining the operating environment evaluation value is as follows: ； Wherein, Yp is the operating environment assessment value, WsH is the ambient temperature, Ds is the electromagnetic interference level, Ws is the power supply stability, WcH is the ambient parameter temperature, Dc is the parameter electromagnetic interference level, and Wc is the parameter power supply stability.

4. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 1 is characterized in that: Evaluating the instruction completion status of the permanent magnet synchronous motor driver to determine whether the instruction completion status of the permanent magnet synchronous motor driver is normal includes the following steps: Obtain the command completion status data of the permanent magnet synchronous motor drive, including the average command response time, dynamic response time and steady-state achievement time; Obtain instruction completion parameter data stored in the database, including instruction parameter average response time, parameter dynamic response time, and parameter steady-state achievement time; Obtain the allowable deviation data of the instruction completion status, including the allowable deviation value of the dynamic response time and the allowable deviation value of the steady-state achievement time; Obtaining an instruction completion evaluation value based on instruction completion data, instruction completion parameter data, and instruction completion allowable deviation data; Determine whether the instruction completion evaluation value is greater than the instruction completion evaluation threshold stored in the database: If the instruction completion evaluation value is greater than the instruction completion evaluation threshold stored in the database, the instruction completion is abnormal; If the instruction completion status evaluation value is not greater than the instruction completion status evaluation threshold stored in the database, the instruction completion status is normal.

5. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 4 is characterized in that: Obtaining instruction completion tolerance data includes the following steps: Based on the device information of the permanent magnet synchronous motor drive, a matching data set is obtained from a database, where the matching data set includes a plurality of deviation value matching data, and the deviation value matching data includes an electromagnetic interference matching value, a power supply stability matching value, and a daily average temperature of the drive; Obtain the data to be matched of the permanent magnet synchronous motor drive, including the electromagnetic interference level, power supply stability and actual temperature of the drive; Compare the data to be matched with each deviation value matching data one by one, and determine the deviation value matching data closest to the data to be matched; The instruction completion status allowable deviation data corresponding to the deviation value matching data stored in the database is obtained.

6. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 5 is characterized in that: The data to be matched is compared with each deviation value matching data one by one, and the deviation value matching data closest to the data to be matched is determined, including the following steps: Obtaining a matching value after comparing the data to be matched with each deviation value matching data; Determine the minimum matching value; The deviation value matching data corresponding to the minimum matching value is output as the deviation value matching data closest to the data to be matched.

7. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 6 is characterized in that: The method for obtaining the matching value is as follows: ； Where PP is the matching value, Ds is the electromagnetic interference level, Ws is the power supply stability, Qsw is the actual temperature of the driver, Dp is the electromagnetic interference matching value, Wp is the power supply stability matching value, Qcw is the daily average temperature of the driver, e is a natural constant, and ln is a logarithmic function, which means the logarithm with e as the base.

8. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 1 is characterized in that: Judging whether the operating state of the permanent magnet synchronous motor driver is normal based on the real-time operating data includes the following steps: Obtain the operating status parameter data of the permanent magnet synchronous motor drive stored in the database, including the motor winding parameter impedance, the parameter circuit harmonic content, the DC link parameter voltage and the rotor parameter offset; Obtaining the allowable deviation data of the operating status, including the allowable deviation value of the motor winding parameter impedance and the allowable deviation value of the DC link parameter voltage; The operating status of the permanent magnet synchronous motor drive is evaluated based on real-time operating data, operating status parameter data and operating status allowable deviation data to determine whether the operating status is normal.

9. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 8 is characterized in that: The operating state of the permanent magnet synchronous motor drive is evaluated based on the real-time operating data, the operating state parameter data and the operating state allowable deviation data to determine whether the operating state is normal, including the following steps: Obtaining an operation status evaluation index based on real-time operation data, operation status parameter data, and operation status allowable deviation data; Determine whether the running status evaluation index is greater than the running status evaluation threshold stored in the database: If the running status evaluation index is greater than the running status evaluation threshold stored in the database, the running status is abnormal; If the running status evaluation index is not greater than the running status evaluation threshold stored in the database, the running status is normal.

10. The permanent magnet synchronous motor drive fault diagnosis method based on fault classifier according to claim 8, characterized in that: Obtaining the operating status allowable deviation data includes the following steps: Obtain the operating status matching data of the permanent magnet synchronous motor drive, including the average daily continuous use time of the drive, the electromagnetic interference level and the power supply stability; Compare the running status matching data with each state deviation matching data stored in the database one by one to obtain the comparison coefficient. The state deviation matching data includes the matching parameters of the average continuous use time of the drive per day, the electromagnetic interference level matching parameters and the power supply stability matching parameters; Determine the minimum comparison coefficient, and extract the operating status allowable deviation data corresponding to the minimum comparison coefficient from the database.

Images