Fault diagnosis method and system for servo drive system

Abstract

The invention relates to the technical field of servo driving systems, and discloses a fault diagnosis method and system for a servo driving system, and the method comprises the steps: carrying out the modeling of the servo driving system under a dq coordinate system, and obtaining a system parameter monitoring matrix; respectively injecting a positive virtual signal and a negative virtual signal into an observer of the servo driving system to obtain d-axis voltage and current response data and q-axis voltage and current response data; constructing a standardized feature vector and an enhanced fault feature matrix corresponding to the fault type according to the d-axis and q-axis voltage and current response data; calculating a system stability margin and a parameter health degree index of the servo driving system; and the standardized feature vector, the enhanced fault feature matrix and the parameter health degree index are input into a support vector machine classifier for fault identification, a fault severity level is obtained, a target fault diagnosis result is output in combination with a system stability margin, and the fault diagnosis stability and reliability of the servo drive system are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of servo drive systems, and in particular to a fault diagnosis method and system for a servo drive system. Background Art

[0002] The electrical parameters of the servo system may drift or fail during long-term operation, such as changes in motor resistance and inductance, abnormal flux parameters, voltage / current deviation, etc. If these faults are not detected and handled in time, they will lead to system performance degradation or even failure. Traditional servo system fault diagnosis methods are mainly based on signal processing or model comparison, but these methods usually require the injection of continuous disturbance signals or rely on high-precision system models, which will interfere with the normal operation of the system or make it difficult to obtain accurate diagnostic results in actual applications.

[0003] Fault diagnosis under low-speed conditions has always been a difficult problem in servo system maintenance. When the servo system is running at low speed, the back-EMF signal is weak, and traditional diagnostic methods based on current or speed fluctuations are difficult to obtain effective fault feature information. The diagnostic sensitivity is significantly reduced, resulting in a sharp drop in fault detection performance when the speed is below 2% of the rated speed. At the same time, traditional diagnostic methods are often affected by system load conditions and nonlinear characteristics of the driver (such as dead zone), which makes the diagnostic results lack consistency and reliability under different working conditions, and it is difficult to meet the fault detection needs within the full range of working conditions. Summary of the invention

[0004] The present invention provides a fault diagnosis method and system for a servo drive system, which greatly improves the fault diagnosis stability and reliability of the servo drive system.

[0005] In a first aspect, the present invention provides a method for diagnosing a fault of a servo drive system, the method comprising: Model the servo drive system in the dq coordinate system and obtain the system parameter monitoring matrix; Based on the system parameter monitoring matrix, positive virtual signals and negative virtual signals are respectively injected into the observer of the servo drive system to obtain d-axis and q-axis voltage and current response data; According to the d-axis and q-axis voltage and current response data, construct a standardized feature vector corresponding to the fault type and an enhanced fault feature matrix; Based on the standardized feature vector and the enhanced fault feature matrix, calculating the system stability margin and parameter health index of the servo drive system; The standardized feature vector, the enhanced fault feature matrix and the parameter health index are input into a support vector machine classifier for fault identification to obtain a fault severity level, and a target fault diagnosis result is output according to the system stability margin and the fault severity level.

[0006] In a second aspect, the present invention provides a fault diagnosis system for a servo drive system, the fault diagnosis system for the servo drive system comprising: A modeling module is used to model the servo drive system in a dq coordinate system and obtain a system parameter monitoring matrix; An injection module, used for injecting positive virtual signals and negative virtual signals into the observer of the servo drive system respectively based on the system parameter monitoring matrix to obtain d-axis and q-axis voltage and current response data; A construction module, used to construct a standardized feature vector corresponding to the fault type and an enhanced fault feature matrix according to the d-axis and q-axis voltage and current response data; A parameter estimation module, used for calculating the system stability margin and parameter health index of the servo drive system based on the standardized feature vector and the enhanced fault feature matrix; A fault identification module is used to input the standardized feature vector, the enhanced fault feature matrix and the parameter health index into a support vector machine classifier for fault identification, obtain a fault severity level, and output a target fault diagnosis result according to the system stability margin and the fault severity level.

[0007] In the technical solution provided by the present invention, by designing a positive and negative virtual signal injection strategy, it is found that the fault characteristics are only related to the ratio of the q-axis and d-axis voltage fluctuations, so that the diagnosis result is completely independent of the position error and the nonlinear characteristics of the driver, which greatly improves the stability and reliability of the diagnosis. The high-frequency sampling mechanism combined with the resonance characteristic analysis can capture the characteristic information of the system in the high-frequency domain, and realize high-precision fault detection at low-speed operation (less than 2% of the rated speed), breaking through the limitation of poor diagnostic performance of traditional methods under low-speed conditions. The time-frequency analysis method based on the current phase angle realizes the comprehensive diagnosis of all parameters such as motor resistance, inductance, flux parameters and voltage and current deviations, and the diagnostic range is more comprehensive and more adaptable. By constructing a parameter health index and a system stability margin evaluation mechanism, not only can the fault type be identified, but also the fault severity can be quantified. The combination of standardized feature vectors and enhanced fault feature matrices significantly improves the accuracy of fault diagnosis under different load conditions. It has been verified that the diagnostic accuracy rate under various fault conditions is more than 99%. The algorithm combining recursive parameter estimation with support vector machine classification makes the fault diagnosis process adaptive and sensitive to parameter drift, so that potential problems can be discovered in time at the early stage of fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0009] Figure 1 A schematic diagram of the steps of a fault diagnosis method for a servo drive system according to an embodiment of the present invention; Figure 2 Schematic diagram of the structure of the fault diagnosis system of the servo drive system in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] The embodiment of the present invention provides a fault diagnosis method and system for a servo drive system. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0011] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the fault diagnosis method of the servo drive system in the embodiment of the present invention includes: Step S1, modeling the servo drive system in a dq coordinate system to obtain a system parameter monitoring matrix; It is understandable that the execution subject of the present invention may be a fault diagnosis system of a servo drive system, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject.

[0012] Specifically, the electrical part of the servo drive system is modeled, and the electrical characteristics of the system are described based on the voltage equation in the dq coordinate system. The voltage equation of the electrical part of the servo drive system in the dq coordinate system can reveal multiple key parameters involved in the system, such as stator resistance, d-axis inductance, q-axis inductance, and permanent magnet flux, reflecting the relationship between current, voltage, inductance, and flux. According to the characteristic equation, a set of virtual parameters including motor resistance, inductance, flux parameters, and voltage and current deviations are designed. These virtual parameters are used as the source of virtual interference signals to simulate the disturbances caused by different parameters to the dynamic characteristics of the system. The virtual disturbance signal can effectively simulate the response difference between the system in normal operation and when a fault occurs. By designing these virtual disturbance signals and injecting them into the characteristic equation of the system, an enhanced system model is obtained to reflect the dynamic behavior of the system under parameter disturbance. The enhanced system model can distinguish the system under normal operation from the system under fault conditions. The observer structure is constructed based on the enhanced system model, where the observer structure includes a state observer and a parameter identifier. The state observer is a component of the observer. The function of the state observer is to estimate the system state in real time, especially to estimate the dynamic response of the d-axis and q-axis currents. The state observer uses mathematical equations to describe its estimation process. These equations estimate the internal state variables of the system by using the system input and output information through feedback control. The dynamic response equations of the state observer can accurately estimate the d-axis current and q-axis current, providing accurate current data for subsequent fault feature extraction and fault diagnosis. In order to analyze the influence of system parameters on the dynamic characteristics of the system, the partial derivatives of each parameter in the enhanced system model are calculated to obtain the system parameter monitoring matrix, which describes the specific influence of each parameter change in the system on the dynamic characteristics of the system such as voltage and current. By calculating the system parameter monitoring matrix, the sensitivity of each parameter to the system behavior is quantified, and the response mode of the system under different parameter disturbances is revealed. The system parameter monitoring matrix is ​​rank analyzed to ensure the observability of the system. Observability refers to the ability to infer the internal state of the system under given system input and output data. In fault diagnosis, the observability of the system determines whether the internal parameters or fault status of the system can be accurately estimated based on externally available signals such as voltage and current. Through rank analysis, it is verified whether the system's parameter monitoring matrix meets the observability condition, that is, whether the rank of the matrix is ​​greater than or equal to the dimension of the parameter to be monitored. A virtual signal injection model that meets the system's observability condition is obtained.

[0013] Step S2: Based on the system parameter monitoring matrix, inject positive virtual signals and negative virtual signals into the observer of the servo drive system respectively to obtain d-axis and q-axis voltage and current response data; Specifically, according to the system parameter monitoring matrix, a disturbance vector containing parameters such as motor resistance, inductance, magnetic flux, and voltage and current deviation is designed. These disturbance vectors represent positive disturbance vectors and negative disturbance vectors, respectively. The design of positive disturbance vectors and negative disturbance vectors is to simulate different fault modes in the system and help detect and diagnose faults in the system. When designing these virtual signals, it is necessary to ensure that the disturbance signal can cover different system states and can effectively distinguish between faults and normal working states to obtain positive virtual signals δP+ and negative virtual signals δP-. Amplitude constraints are imposed on the positive virtual signals δP+ and the negative virtual signals δP- to ensure that the amplitude of the disturbance signal is not too large, so as not to affect the stability of the system. The setting of the amplitude constraint is generally determined according to the working range and tolerance range of the system to ensure that the disturbance signal is small enough and will not destroy the normal operation of the system. At the same time, it can effectively simulate the response behavior of the system when a fault occurs. By imposing appropriate amplitude constraints on the positive virtual signals and the negative virtual signals, the purpose of effective disturbance can be achieved without affecting the stability of the system. Based on the system parameter monitoring matrix and the positive virtual signal δP+, the response of the system is calculated to obtain the d-axis and q-axis positive voltage equations containing parameter disturbances. After the analog signal is injected, the voltage and current changes of the system when it is disturbed are observed, and the corresponding voltage equations are derived. Through these equations, the voltage and current changes caused by parameter disturbances are captured and used as one of the key features of subsequent fault diagnosis. Similarly, based on the system parameter monitoring matrix and the negative virtual signal δP-, the response of the system is calculated to obtain the d-axis and q-axis negative voltage equations containing parameter disturbances. Negative disturbances can also provide key information about the changes in system states, and together with positive disturbances, help accurately distinguish between normal and fault states. The positive virtual signal δP+ and the negative virtual signal δP- are alternately injected to ensure that the disturbance signal is injected according to a specific timing. The alternating injection strategy is designed to avoid excessive disturbances in the system while ensuring the continuity and effectiveness of the disturbance signal. The alternating injection timing is set to be carried out within a certain period of time, usually involving a certain time interval to ensure that the system can return to a stable state after each disturbance, so as not to interfere with the long-term stability of the system. On this basis, in order to ensure that the disturbance signal can transition smoothly and avoid sudden large fluctuations in the system, a smooth transition function is designed to control the transition mode of the disturbance signal during the injection process, so that the change of the disturbance signal is smoother and the oscillation phenomenon generated during the operation of the system is reduced. After completing the design of the alternating injection timing and smooth transition function, the injected signal will gradually change within the set time period. The voltage and current response data of the system will change under the action of these disturbance signals, and the observer of the servo drive system will record these data in real time.Through the observer, the voltage and current data of the system before and after the disturbance signal is injected are captured, and the d-axis and q-axis voltage and current response data are obtained, which reflect the actual response of the system under the injection of virtual signals.

[0014] Step S3, constructing a standardized feature vector and an enhanced fault feature matrix corresponding to the fault type according to the d-axis and q-axis voltage and current response data; Specifically, according to the injection of positive virtual signals and negative virtual signals, the deviations of the d-axis and q-axis voltages are calculated to obtain a voltage deviation sequence, which reflects the voltage changes of the system under different working conditions. The deviation sequence is subjected to sliding average filtering to effectively remove the influence of noise on the signal and obtain the filtered voltage deviation value. According to the filtered voltage deviation value, the ratio of the d-axis voltage deviation to the q-axis voltage deviation is calculated to obtain the voltage fluctuation ratio parameter under positive and negative injection conditions to help identify different types of faults. Threshold judgment processing is applied to the voltage fluctuation ratio parameter to eliminate errors caused by factors such as noise and ensure that the obtained ratio value represents the actual state of the system. If the voltage fluctuation ratio is too small or close to zero, it indicates that the response of the system under this working condition is not obvious, so a threshold is set to determine whether there is an obvious fault feature. Through threshold judgment processing, a more accurate voltage fluctuation ratio difference is obtained as a key fault feature parameter. Based on the key fault feature parameters, motor speed, d-axis current and q-axis current, characteristic components are calculated to represent the behavioral characteristics of the system under different working conditions, including the electrical characteristics and mechanical motion characteristics of the motor. Through comprehensive analysis of these characteristic components, a four-dimensional fault feature vector is obtained. Each component of the fault feature vector is normalized by the corresponding scaling factor to eliminate the influence caused by the dimension difference between different feature components, and obtain the standardized feature vector corresponding to the fault type. The time-frequency characteristics of the current phase angle of the servo drive system are analyzed by using the d-axis and q-axis voltage and current response data to obtain the spectrum characteristics of the system under different working conditions. Through methods such as short-time Fourier transform, the change of the current phase angle is converted into features in the frequency domain, revealing the frequency response of the system under different load conditions or fault conditions. By extracting these frequency features and combining them with the standardized feature vector, an enhanced fault feature matrix containing time and frequency information is constructed.

[0015] According to the d-axis and q-axis voltage and current response data, the amplitude and phase angle of the current vector are calculated, and the current phase characteristics of the system are analyzed to reflect the performance of the servo drive system under different operating conditions. The amplitude of the current vector is calculated by the square root of the sum of the squares of the d-axis and q-axis currents, while the phase angle is calculated by the inverse tangent function to calculate the relationship between the d-axis and q-axis currents. Based on the current phase characteristics, a short-time Fourier transform is performed to obtain the time-frequency distribution characteristics of the current phase angle. The short-time Fourier transform can convert the characteristics of the current phase angle changing with time into a representation in the frequency domain, revealing the change law of the current phase angle in different frequency ranges. This process helps to identify the frequency characteristics of the servo drive system under various working conditions, especially in terms of high-frequency resonance characteristics, which can help identify the frequency characteristic changes of the system when a fault occurs. After short-time Fourier transform processing, the obtained time-frequency distribution characteristics contain the frequency response information of the system at each time point. From the time-frequency distribution characteristics, the spectrum of the current phase angle is extracted to obtain the phase angle spectrum reflecting the high-frequency resonance characteristics of the system. These high-frequency resonance characteristics indicate that the system has nonlinear behavior or potential faults in certain frequency ranges, revealing the characteristics of the current phase angle change when a fault occurs. For the d-axis and q-axis currents, normalization mapping is performed on them, and their current vectors are mapped to the unit circle to eliminate the influence of the current amplitude difference on the current phase characteristic analysis, so that the current vector trajectory under different working conditions is more consistent and comparable. The current vector after normalization mapping can reflect the trajectory changes of the current under different working conditions. Based on the current vector trajectory mapping function, the partial derivatives of each system parameter with respect to the phase angle are calculated, and the influence of the change of system parameters (such as motor resistance, inductance, etc.) on the current phase angle is analyzed to obtain the sensitivity matrix of the phase angle, characterize the sensitivity of different system parameter changes to the current phase angle, and then reveal which parameters have the greatest influence on the current characteristics of the system. The standardized eigenvector and the high-frequency resonant phase angle characteristics are combined according to different scaling factors to obtain an enhanced fault feature matrix.

[0016] Step S4, calculating the system stability margin and parameter health index of the servo drive system based on the standardized feature vector and the enhanced fault feature matrix; Specifically, the deviations of motor resistance, inductance, flux parameters and voltage and current deviations relative to the rated values ​​are calculated to obtain the relative deviation vector of system parameters, which characterizes the degree of change of the current state of the system relative to the normal working condition and reflects the drift of each parameter during long-term operation or when a fault occurs. Based on the standardized eigenvector and the enhanced fault characteristic matrix, the observation vector and observation matrix are constructed to establish a mathematical model for recursive parameter estimation. The construction of the observation vector depends on the combination of characteristic variables such as current, voltage, and phase angle, while the observation matrix characterizes the influence of different parameters on the system response by combining the eigenvector with the system dynamics equation. The core goal of this mathematical model is to continuously update the estimated values ​​of system parameters using observation data so that it can adapt to the dynamic changes of the system and ensure that abnormal signals can be captured in time when a fault occurs. The construction of the observation vector and observation matrix not only needs to ensure observability, but also take into account the dynamic characteristics of the system under different working conditions to enhance the applicability and robustness of the model. The least squares method is applied to the mathematical model of recursive parameter estimation to obtain the real-time estimated values ​​of system parameters. In recursive estimation, the least squares method can continuously update the parameters when new observation data is continuously added, so as to ensure that the estimation results are continuously adjusted as the system state changes. This estimation method can reduce the impact of noise on parameter estimation and ensure the convergence of the estimated value, so that the system can obtain stable parameter update results in a dynamic environment. The real-time parameter estimation values ​​calculated by the least squares method are used to obtain the latest values ​​of parameters such as motor resistance, inductance and flux under the current working conditions, and compare them with the rated values ​​to analyze the health status of the system. In order to balance the influence of new and old data, the adaptive forgetting factor is calculated, and the recursive parameter estimation algorithm is optimized based on it. The calculation of the adaptive forgetting factor depends on the prediction error in the parameter estimation process. When the prediction error is large, a higher weight is given to the new data to ensure that the system can respond quickly to sudden faults. When the error is small, sufficient attention should be paid to historical data to ensure the stability of the estimation results. Through the adaptive mechanism, the estimation process is dynamically adjusted, so that the parameter estimation can respond to rapidly changing working conditions and maintain the accuracy of long-term trends. After obtaining the real-time estimated values ​​of the system parameters, the eigenvalues ​​of the system characteristic equation are calculated, and based on this, the stability margin of the system is calculated to obtain a quantitative indicator reflecting the stability of the system. The eigenvalues ​​of the system characteristic equation determine the dynamic stability of the system, and the stability margin is an important indicator to measure whether the system can maintain normal operation under external disturbances. If the stability margin is too low, it means that the system's resistance to external interference is weakened, resulting in instability or even failure. Calculating the stability margin of the system can help the system determine whether the current parameter state will affect the overall stability of the system, and provide early warning information accordingly. Apply an exponential function transformation to the relative deviation vector of the system parameters to obtain the parameter health index.The relative deviation of the parameters is mapped to a reasonable health range, so that the health index can more intuitively represent the health status of the system. The value of the health index is between 0 and 1. When it is close to 1, it means that the system is in good condition, while when it is close to 0, it means that the system has experienced serious parameter drift or failure. The health index can be used to quantitatively evaluate the overall health of the servo drive system, and maintenance plans or preventive measures can be formulated based on the trend of health changes to extend the service life of the system and improve operational reliability.

[0017] Step S5: input the standardized feature vector, enhanced fault feature matrix and parameter health index into the support vector machine classifier for fault identification, obtain the fault severity level, and output the target fault diagnosis result according to the system stability margin and the fault severity level.

[0018] Specifically, the standardized feature vector, enhanced fault feature matrix and parameter health index are combined into a unified vector to obtain a complete representation of the fault feature space. The fault feature space representation is input into the support vector machine classifier and mapped to the high-dimensional feature space through the radial basis kernel function to construct an optimized classification hyperplane. The support vector machine finds an optimal hyperplane so that fault features of different categories can be effectively distinguished in the high-dimensional space. Since there is a nonlinear relationship between different fault types, the radial basis kernel function can effectively capture these nonlinear features and improve the generalization ability of the classifier. The calculation form of the radial basis kernel function can ensure that the distance metric between features remains stable in the high-dimensional space, thereby enhancing the classifier's adaptability to complex fault modes. In this process, the solution of the optimized classification hyperplane involves a constrained optimization problem, the goal of which is to maximize the interval between categories while minimizing the classification error to ensure that the classifier can provide high-confidence fault identification results when facing new samples. Based on the optimized classification hyperplane, the multi-category fault classification function value and fault confidence index are calculated for the input sample. The fault classification function value is used to measure the similarity between the input sample and different fault categories, while the fault confidence index is used to quantify the credibility of the classification result. In this way, each input sample corresponds to a candidate fault type and is accompanied by a confidence score. The confidence score is calculated based on the output distribution of the classification function. A probability value ranging from 0 to 1 is obtained through normalization transformation. The closer this value is to 1, the higher the confidence that the sample belongs to this fault type. In this way, the most likely fault type is effectively screened out and its confidence is marked in the classification result, making the diagnosis result more interpretable and reliable. The fault severity level is divided according to the candidate fault type and parameter health index. The fault severity is mainly divided based on the offset of the key parameters of the system and the system stability margin of the system, so that the fault is divided into different levels, such as minor, moderate, severe and dangerous faults. When the parameter health index is close to 1, it indicates that the system is in good condition. When the health index is close to 0, it means that the system parameters have drifted seriously, resulting in abnormal operation or system failure. The fault level is marked as severe or dangerous, and corresponding maintenance measures are recommended. Based on the calculation results of the support vector machine classifier, a complete target fault diagnosis result is obtained.

[0019] In the embodiment of the present invention, by designing the positive and negative virtual signal injection strategy, it is found that the fault characteristics are only related to the ratio of the q-axis and d-axis voltage fluctuations, so that the diagnosis result is completely independent of the position error and the nonlinear characteristics of the driver, which greatly improves the stability and reliability of the diagnosis. The high-frequency sampling mechanism combined with the resonance characteristic analysis can capture the characteristic information of the system in the high-frequency domain, realize high-precision fault detection at low-speed operation (less than 2% of the rated speed), and break through the limitation of poor diagnostic performance of traditional methods under low-speed conditions. The time-frequency analysis method based on the current phase angle realizes the comprehensive diagnosis of all parameters such as motor resistance, inductance, flux parameters and voltage and current deviations, and the diagnostic range is more comprehensive and more adaptable. By constructing the parameter health index and the system stability margin evaluation mechanism, not only can the fault type be identified, but also the fault severity can be quantified. The combination of standardized feature vectors and enhanced fault feature matrices significantly improves the accuracy of fault diagnosis under different load conditions. It has been verified that the diagnostic accuracy rate under various fault conditions is more than 99%. The algorithm combining recursive parameter estimation with support vector machine classification makes the fault diagnosis process adaptive and sensitive to parameter drift, so that potential problems can be discovered in time at the early stage of fault.

[0020] In a specific embodiment, the process of executing step S1 may specifically include the following steps: The voltage equation in the dq coordinate system is established for the electrical part of the servo drive system, and the characteristic equation including the stator resistance, d-axis inductance, q-axis inductance and permanent magnet flux is obtained; According to the characteristic equation, a virtual parameter set including motor resistance, inductance, flux parameters and voltage and current deviations is designed to obtain a virtual interference signal for injection; Injecting a virtual interference signal into the characteristic equation to obtain an enhanced system model including parameter disturbance quantities, and constructing an observer structure based on the enhanced system model to obtain a dynamic response equation for estimating d-axis current and q-axis current; The partial derivatives of each parameter of the enhanced system model with respect to the system dynamic characteristics are calculated to obtain the system parameter monitoring matrix. At the same time, the system parameter monitoring matrix is ​​rank analyzed to obtain a virtual signal injection model that meets the system observability conditions.

[0021] Specifically, first. Model the electrical part of the servo drive system, and establish the voltage equation in the dq coordinate system to accurately describe the electrical dynamic characteristics of the system. Convert the voltage, current and flux variables in the three-phase stationary coordinate system to express them in the dq rotating coordinate system, thereby eliminating the time variation term, making the motor mathematical model more intuitive and easy to analyze. Based on this modeling process, the characteristic equation of the system is obtained, which contains key electrical parameters such as stator resistance, inductance and permanent magnet flux. In the servo motor, the d-axis and q-axis voltage equations are expressed as follows: ; ; in, and are the d-axis and q-axis voltages, and are the d-axis and q-axis currents, is the stator resistance, and are the d-axis and q-axis inductances, is the electrical angular velocity, is the permanent magnet flux, represents the derivative of the d-axis current with respect to time, represents the derivative of the q-axis current with respect to time. This equation describes the dynamic relationship between voltage, current, inductance and flux. In order to enhance the observability of the system, a virtual parameter set including motor resistance, inductance, flux parameters and voltage and current deviation is designed based on the above characteristic equation, and a virtual interference signal injection system is constructed. Assuming that the key parameters in the servo system will change during long-term operation, such as resistance increase, inductance decrease or flux weakening, a virtual parameter set is defined , contains all the parameters that need to be observed: ; in, and are the deviation compensation parameters of voltage and current, respectively, which are used to describe the influence of external noise, sensor error or nonlinear effect. Design virtual interference signal Inject into the system to enhance the observability of the system model. Set positive and negative virtual disturbances and , whose expression is: ; ; in, The parameters in are the positive perturbations of the parameters in P. The parameters in are the negative disturbances of the parameters in P. After introducing these disturbances into the characteristic equation, an enhanced system model containing parameter disturbances is obtained, where the d-axis and q-axis voltage equations after disturbance are as follows: ; ; This enhanced system model takes into account the changes in multiple key parameters, making the system more sensitive to parameter drift and thus more effectively identifying faults. In order to ensure that the injection of virtual signals does not affect the stability of the system, appropriate disturbance amplitude constraints are set so that the disturbance signal satisfies the following relationship: ; in, is the disturbance amplitude coefficient, and its typical value range is between 0.01 and 0.05 to ensure that the disturbance is small enough not to affect the normal operation of the system. On this basis, a state observer is constructed based on the enhanced system model to estimate the dynamic response of the d-axis and q-axis currents. The mathematical expression of the state observer is as follows: ; ; in, and are the d-axis and q-axis currents estimated by the observer, and is the feedback gain coefficient. The role of this observer is to use the input voltage and current feedback to continuously adjust the estimated value so that it can effectively track the actual state of the system. In order to analyze the impact of each parameter on the dynamic characteristics of the system, the partial derivatives of the enhanced system model are calculated to construct the system parameter monitoring matrix : ; To ensure system observability, the matrix Perform rank analysis, i.e. check Is it true. If the rank of the system parameter monitoring matrix is ​​greater than or equal to the parameter dimension, it means that all parameters can be accurately identified by the observer, otherwise the design of the virtual signal needs to be adjusted to enhance observability. For example, in some cases, it is necessary to add additional disturbance signals or adjust the sampling frequency to ensure the identification capability of all key parameters. Through the above process, the electrical model of the servo drive system is established, and by introducing virtual interference signals, enhancing system observability and constructing a state observer, the system can estimate key parameters in real time during operation, and detect and diagnose problems in time when a fault occurs.

[0022] In a specific embodiment, the process of executing step S2 may specifically include the following steps: According to the system parameter monitoring matrix, a positive disturbance vector and a negative disturbance vector including motor resistance, inductance, magnetic flux and voltage and current deviation parameters are set to obtain a positive virtual signal δP+ and a negative virtual signal δP-; Apply amplitude constraints to the positive virtual signal δP+ and the negative virtual signal δP- to ensure that the system stability is not affected, and obtain the positive virtual signal δP+ and the negative virtual signal δP- after amplitude constraints; The system response is calculated based on the positive virtual signal δP+ after amplitude constraint, and the d-axis and q-axis forward voltage equations including parameter disturbances are obtained; Based on the amplitude-constrained negative virtual signal δP-, the system response is calculated and the negative voltage equations of the d-axis and q-axis including parameter disturbances are obtained. For the positive virtual signal δP+ after amplitude constraint and the negative virtual signal δP- after amplitude constraint, alternating injection timing and smooth transition function are designed to obtain a time-continuous injection signal. The voltage and current data of the system in different time periods before and after the injection of the time-continuous injection signal are recorded through the observer of the servo drive system to obtain the d-axis and q-axis voltage and current response data.

[0023] Specifically, based on the system parameter monitoring matrix, the positive disturbance vector and negative disturbance vector of key parameters including motor resistance, inductance, magnetic flux, voltage and current deviation are set to obtain the positive virtual signal and negative virtual signal During system operation, parameters may shift due to temperature changes, electrical aging or other environmental factors. Therefore, the purpose of introducing virtual signal disturbance is to observe the response characteristics of the system under different disturbance conditions. For servo drive systems, the key parameter set is Including stator resistance , d-axis inductance , q-axis inductance , permanent magnet flux , voltage deviation and current deviation , so the positive and negative perturbation vectors are expressed as: ; ; in, and Respectively represent the positive and negative disturbances of the stator resistance, and the other parameters are similar. In order to ensure that these disturbance signals do not affect the stability of the system, amplitude constraints are imposed on them, that is, the amplitude of the disturbance is limited so that it does not exceed the range that the system can withstand. The maximum allowable value of the disturbance is usually a fixed proportion of the parameter, such as between 1% and 5%, so the constraint is expressed as: ; in, is the disturbance coefficient, and its typical value range is between 0.01 and 0.05. By imposing these amplitude constraints, the amplitude of the disturbance signal is guaranteed to be small enough not to cause system instability, but large enough to ensure that the fault characteristics of the system can be clearly reflected. , calculate the response of the system under the forward disturbance condition, and obtain the d-axis and q-axis forward voltage equations containing parameter disturbances. Since the disturbance signal directly affects the electrical parameters of the system, the d-axis and q-axis voltages after disturbance are expressed as: ; ; in, and are the d-axis and q-axis voltages under positive disturbance respectively. Similarly, based on the system parameter monitoring matrix and the negative virtual signal , calculate the response of the system under negative disturbance conditions, and obtain the negative voltage equations of the d-axis and q-axis including parameter disturbances, which are in the following form: ; ; In order to ensure that the positive and negative virtual signal injections have the least impact on the system and ensure the continuity of data acquisition, a suitable alternating injection sequence is designed and a smooth transition function is used so that the switching of signal injection does not cause drastic dynamic changes in the system. The injection sequence is set in a periodic alternating manner, that is, at time Internal injection , then at time Internal injection ,in and is the disturbance duration. In order to make the injected signal transition smoothly, the cosine transition function is used. The change of the control disturbance signal is defined as follows: ; in, is the transition time, which is set to 1ms to make the disturbance signal change more smoothly and avoid impact on system operation. The final injection signal is expressed as: ; During this process, the observer of the servo drive system records the voltage and current data of the system before and after the disturbance is injected for subsequent analysis. The observation data contains the following information: ; These data can reflect the dynamic response of the system and are used to further extract fault characteristics and identify the changing trends of key parameters.

[0024] In a specific embodiment, the process of executing step S3 may specifically include the following steps: Based on the d-axis and q-axis voltage and current response data, the d-axis and q-axis voltage deviations caused by the injection of positive virtual signals and negative virtual signals are calculated to obtain a voltage deviation sequence; Perform sliding average filtering on the voltage deviation sequence to obtain a filtered voltage deviation value, and calculate the ratio of the d-axis voltage deviation to the q-axis voltage deviation based on the filtered voltage deviation value to obtain a voltage fluctuation ratio parameter under positive and negative injection conditions; Apply threshold judgment processing to the voltage fluctuation ratio parameter, and calculate the voltage fluctuation ratio difference under positive and negative injection conditions to obtain key fault characteristic parameters; The characteristic components are calculated based on the key fault characteristic parameters, motor speed, d-axis current and q-axis current to obtain a four-dimensional fault characteristic vector; Each component of the fault feature vector is normalized by a corresponding scaling factor to obtain a standardized feature vector corresponding to the fault type; The time-frequency characteristics of the current phase angle of the servo drive system are analyzed using the d-axis and q-axis voltage and current response data, and an enhanced fault feature matrix is ​​obtained.

[0025] Specifically, based on the d-axis and q-axis voltage and current response data, the d-axis and q-axis voltage deviations caused by the injection of positive virtual signals and negative virtual signals are calculated to obtain a voltage deviation sequence. During the virtual signal injection process, the voltage and current of the system will change, and these changes reflect the response of the system to the disturbance. By calculating the d-axis and q-axis voltage differences before and after the disturbance, the voltage deviations under positive and negative disturbances are obtained respectively. Assume and represent the d-axis and q-axis voltages after forward disturbance, respectively, and and They represent the d-axis and q-axis voltages after negative disturbance, respectively, and and is the steady-state voltage before the disturbance, then the voltage deviation sequence is expressed as: in, and represents the d-axis and q-axis voltage deviation of the system after the positive virtual signal is injected, and and Represents the voltage deviation of the system after the negative virtual signal is injected. These voltage deviation data contain the system's response information to the disturbance and are used to analyze the changes in the key parameters of the system. The voltage deviation data is subjected to a sliding average filter to eliminate random fluctuations in a short period of time and improve the stability of the feature. The calculation method of the sliding average filter is as follows: in, is the size of the sliding window, is the sampling period, and are the filtered values ​​of the d-axis and q-axis voltage deviations after the positive virtual signal is injected, and and is the filtered value after the negative virtual signal is injected. The random noise is eliminated through the sliding average filter, and the main change trend of the voltage deviation is retained. Based on the filtered voltage deviation value, the ratio of the d-axis voltage deviation to the q-axis voltage deviation is calculated, that is, the voltage fluctuation ratio parameter. This ratio is used to measure the relative changes of the d-axis and q-axis voltages under the disturbance, and is defined as follows: in, and They represent the voltage fluctuation ratio parameters of the system under positive disturbance and negative disturbance conditions respectively. In order to avoid numerical instability caused by the denominator approaching zero in the ratio calculation, a threshold judgment process is introduced, such as: like like sign in, It is the threshold setting, generally 0.5% of the rated voltage. The voltage fluctuation ratio difference under positive and negative injection conditions is calculated to obtain the key fault characteristic parameters: This parameter is used to characterize the changing trend of voltage fluctuation ratio under different fault conditions. For example, when the motor resistance fails, Mainly with current When the d-axis or q-axis inductance fails, With motor speed Directly proportional. And other key parameters, such as motor speed , d-axis current and q-axis current , calculate the characteristic components, and thus construct a four-dimensional fault feature vector: in, is the magnitude of the current vector. In order to make the different components of the eigenvector numerically consistent, they are normalized by the scaling factor: in, is the corresponding scaling factor, which is used to eliminate the scale differences between different physical quantities and make the eigenvector more stable under different working conditions. Using the d-axis and q-axis voltage and current response data, the current phase angle time-frequency characteristics of the servo drive system are analyzed to obtain the enhanced fault feature matrix. Calculate the current phase angle: right Perform short-time Fourier transform to obtain its time-frequency characteristics, extract high-frequency resonance phase angle characteristics, and construct a fault feature matrix : in, is the selected sensitive frequency point, is the corresponding scaling factor. Through this enhanced fault feature matrix, the accuracy of fault diagnosis is improved and the robustness of the fault identification algorithm is optimized.

[0026] In a specific embodiment, the process of performing the step of analyzing the time-frequency characteristics of the current phase angle of the servo drive system using the d-axis and q-axis voltage and current response data to obtain the enhanced fault feature matrix may specifically include the following steps: According to the d-axis and q-axis voltage and current response data, the amplitude and phase angle of the current vector are calculated to obtain the current phase characteristics reflecting the system operation status; Perform short-time Fourier transform on the current phase characteristics to obtain the time-frequency distribution characteristics of the current phase angle; The phase angle spectrum is extracted from the time-frequency distribution characteristics to obtain the high-frequency resonant phase angle characteristics, and the d-axis and q-axis currents are normalized and mapped according to the amplitude of the current vector to obtain the current vector trajectory mapping function under different working conditions; Based on the current vector trajectory mapping function, the partial derivatives of each system parameter with respect to the phase angle are calculated to obtain the phase angle sensitivity matrix that characterizes the parameter sensitivity. The phase angle sensitivity matrix is ​​used to determine the sensitive frequency points of the high-frequency resonance phase angle characteristics and their corresponding scaling factors. The high-frequency resonance phase angle characteristics are scaled at the determined sensitive frequency points to obtain a scaled phase angle feature vector. The scaled phase angle feature vector is combined with the standardized feature vector to obtain an enhanced fault feature matrix.

[0027] Specifically, the amplitude and phase angle of the current vector are calculated based on the d-axis and q-axis voltage and current response data, the current phase characteristics reflecting the system operation state are obtained, and the amplitude and phase angle of the current vector are defined. The amplitude of the current vector is expressed as: in, and They are the current components of the d-axis and q-axis, respectively, and their amplitudes are It reflects the magnitude of the composite current of the motor in the dq coordinate system. The phase angle of the current vector is used to characterize the relative relationship between the d-axis and q-axis currents. The calculation formula is: This phase angle It reflects the directionality of the current on the dq plane. It changes under different loads, different motor parameters or different fault conditions. Therefore, it is used as a characteristic parameter reflecting the operating status of the system. After obtaining the current phase characteristics, in order to analyze the evolution characteristics of the phase angle with time and frequency, the current phase characteristics are subjected to short-time Fourier transform (STFT) to obtain the time-frequency distribution characteristics of the current phase angle. The short-time Fourier transform is expressed as: in, As the window function, choose the Hanning window: in, is the length of the window function. After STFT transformation, the time-frequency distribution of the current phase angle is obtained. , which describes the frequency characteristics of the phase angle at different time points. The phase angle spectrum is extracted from the time-frequency distribution characteristics, especially the high-frequency resonant phase angle characteristics. Since the high-frequency components contain important fault information under different working conditions of the system, it is necessary to pay attention to the phase angle characteristics within a specific frequency range. The high-frequency resonant phase angle characteristics are expressed as: in, Specific sensitive frequency points are selected, which are between 100Hz and 5kHz. The phase angle changes in these frequency ranges contain the characteristic modes caused by faults. In order to analyze the evolution of the current vector under different working conditions, the d-axis and q-axis currents are normalized and mapped to construct the current vector trajectory mapping function, which is defined as follows: Through this normalized mapping, the current vector under different load conditions is converted into a trajectory on the unit circle, thereby eliminating the impact of load changes and making the phase angle characteristics under different working conditions more comparable. Based on the current vector trajectory mapping function, the partial derivatives of each system parameter with respect to the phase angle are calculated to obtain the phase angle sensitivity matrix. Since the current phase angle is affected by parameters such as motor resistance, inductance, and magnetic flux, the sensitivity matrix is ​​defined as follows: in, is the stator resistance, and are the d-axis and q-axis inductances, is the permanent magnet flux linkage, and are the voltage and current deviation parameters, Indicates the phase angle Stator resistance sensitivity, Indicates the phase angle Inductance of d-axis sensitivity, Indicates the phase angle Inductance of q axis sensitivity, Indicates the phase angle Permanent magnet flux sensitivity, Indicates the phase angle Voltage deviation sensitivity, Indicates the phase angle Current deviation The sensitivity matrix is ​​used to evaluate the impact of different parameter changes on the current phase angle, thereby helping to identify the fault type. After calculating the phase angle sensitivity matrix, the standardized eigenvectors are combined with the high-frequency resonant phase angle characteristics according to different scaling factors to finally construct an enhanced fault feature matrix. The standardized eigenvector is usually composed of features such as voltage fluctuation ratio difference, motor speed, and current amplitude, which can be expressed as: in, is the corresponding scaling factor, which is used to eliminate the scale differences between different physical quantities. The enhanced fault feature matrix is ​​expressed as: in, : corresponds to the first element in the phase angle sensitivity matrix , is based on Determine the sensitive frequency point, that is, the frequency at which the phase angle is most sensitive to the change in stator resistance, is the high frequency resonance phase angle characteristic value at this frequency point, is the corresponding scaling factor, when the phase angle is more sensitive to stator resistance changes (| |The larger the value), The smaller the value, the The larger the weight in the enhanced fault feature matrix, the more effectively the stator resistance fault is reflected. : corresponds to the second element in the phase angle sensitivity matrix , is based on Determine the sensitive frequency point, that is, the frequency at which the phase angle is most sensitive to the change of d-axis inductance, is the high frequency resonance phase angle characteristic value at this frequency point, is the corresponding scaling factor. When the phase angle is more sensitive to the change of d-axis inductance (| |The larger the value), The smaller the value, the The larger the weight in the enhanced fault feature matrix, the more effectively the d-axis inductance fault is reflected. : corresponds to the third element in the phase angle sensitivity matrix , is based on Determine the sensitive frequency point, that is, the frequency at which the phase angle is most sensitive to the change of q-axis inductance, is the high frequency resonance phase angle characteristic value at this frequency point, is the corresponding scaling factor. When the phase angle is more sensitive to the change of q-axis inductance (| |The larger the value), The smaller the value, the The larger the weight in the enhanced fault feature matrix, the more effectively the q-axis inductance fault is reflected. : corresponds to the fourth element in the phase angle sensitivity matrix , is based on The determined sensitive frequency point is the frequency at which the phase angle is most sensitive to the change in the permanent magnet flux. is the high frequency resonance phase angle characteristic value at this frequency point, is the corresponding scaling factor. When the phase angle is more sensitive to the change of permanent magnet flux linkage (| |The larger the value), The smaller the value, the The larger the weight in the enhanced fault feature matrix, the more effectively it reflects the permanent magnet flux linkage fault. : corresponds to the fifth element in the phase angle sensitivity matrix , is based on Determine the sensitive frequency point, that is, the frequency at which the phase angle is most sensitive to the voltage deviation change, is the high frequency resonance phase angle characteristic value at this frequency point, is the corresponding scaling factor. When the phase angle is more sensitive to the voltage deviation change (| |The larger the value), The smaller the value, the The larger the weight in the enhanced fault feature matrix, the more effectively the voltage deviation fault is reflected. : corresponds to the sixth element in the phase angle sensitivity matrix , is based on Determine the sensitive frequency point, that is, the frequency at which the phase angle is most sensitive to the change of current deviation, is the high frequency resonance phase angle characteristic value at this frequency point, is the corresponding scaling factor. When the phase angle is more sensitive to the current deviation change (| |The larger the value), The smaller the value, the The larger the weight in the enhanced fault feature matrix, the more effectively it reflects the current deviation fault. Through this enhanced fault feature matrix, the basic characteristics of voltage and current are combined, and the time-frequency information of the current phase angle is integrated to improve the accuracy of fault feature extraction and enhance the ability to identify different types of faults. For example, in a typical servo motor fault scenario, if the stator resistance Drift occurs, resulting in a slight shift in the current phase angle. The phase angle feature extracted by STFT shows an obvious amplitude change in a specific frequency range. At the same time, the sensitivity matrix The components will also increase, causing some key components in the enhanced fault feature matrix to change.

[0028] The method further includes enhancing the dead zone characteristics and low-speed diagnostic capability of the servo system based on the d-axis and q-axis voltage and current response data after constructing the standardized feature vector and enhanced fault feature matrix corresponding to the fault type, including: applying a high-pass filter to the d-axis and q-axis voltage and current response data for spectrum separation to obtain a high-frequency component signal containing the driver switching characteristics; extracting the switch resonance characteristics of the servo driver during the dead zone based on the high-frequency component signal to obtain a dead zone characteristic parameter model; performing correlation analysis between the dead zone characteristic parameter model and the motor speed signal to obtain a speed-related dead zone compensation factor; dynamically weighting the standardized feature vector according to the dead zone compensation factor to obtain a modified feature vector with enhanced diagnostic capability for low-speed working conditions; establishing a feature mapping relationship between low-speed and rated speed working conditions based on the modified feature vector and the enhanced fault feature matrix to obtain a speed-independent feature conversion matrix; using the feature conversion matrix to construct a diagnostic feature set suitable for the full speed range of the servo system; performing conditional screening and principal component analysis on the diagnostic feature set to obtain the most discriminative core feature subset; calculating the fault separation index under different working conditions based on the core feature subset to obtain an enhanced feature input for low-speed working conditions for parameter estimation.

[0029] In a specific embodiment, the process of executing step S4 may specifically include the following steps: Calculate the deviation of motor resistance, inductance, flux parameters and voltage and current deviation relative to the rated value to obtain the relative deviation vector of system parameters; Apply exponential function transformation to the relative deviation vector of system parameters to obtain parameter health index; Based on the standardized eigenvector and the enhanced fault feature matrix, the observation vector and the observation matrix are constructed to obtain the mathematical model of recursive parameter estimation; Apply the least square method to the mathematical model of recursive parameter estimation to obtain real-time estimates of system parameters; Based on the prediction error in the parameter estimation process, the adaptive forgetting factor is calculated to obtain the updated system parameter estimation value; According to the updated estimated values ​​of system parameters, the eigenvalues ​​of the system characteristic equation are calculated and the minimum absolute value of the real part is obtained to obtain the system stability margin for fault warning.

[0030] Specifically, define the key electrical parameter set of the system and calculate its relative deviation. Set the motor stator resistance , d-axis and q-axis inductance , permanent magnet flux , voltage deviation and current deviation The rated values ​​are , in actual operation, these parameters deviate from their rated values ​​due to aging, temperature changes or external disturbances, so their relative deviation vectors are calculated: in, Relative deviation vector representing the system, each element is the normalized deviation value of the corresponding electrical parameter. The normalization method ensures that different physical quantities are in the same numerical range, making subsequent calculations more stable. Based on the relative deviation vector, combined with the standardized eigenvector and the enhanced fault feature matrix, the observation vector and observation matrix are constructed, and the mathematical model of recursive parameter estimation is established. Standardized eigenvector It is mainly composed of fault characteristic components, such as voltage fluctuation ratio difference, motor speed, current amplitude, etc., while the enhanced fault characteristic matrix The time-frequency information of the current phase angle and the current vector trajectory mapping characteristics are combined. By combining the standardized feature vector and the enhanced fault feature matrix, the observation vector of the system is defined: in, is the observation vector, which consists of standardized features and enhanced fault features, is the observation matrix, describing the relationship between system parameters and measurement data, is the system parameter vector to be estimated, and is the observation noise. By constructing this observation model, the recursive least squares method is used to estimate the system parameters in real time. In the recursive parameter estimation process, the least squares method is used to estimate the real-time value of the system parameters in an optimal way. is the parameter estimate at the current moment, then its recursive update equation is expressed as: in, is the gain matrix, calculated as follows: in, is the measurement noise covariance matrix, is the parameter estimate at the current moment, is the parameter estimate for the next moment. In the parameter estimation process, in order to avoid the excessive influence of old data on the new estimate, an adaptive forgetting factor is introduced. To dynamically balance the weights of new and old data. The calculation of the forgetting factor is based on the prediction error The changing trend of the adaptive forgetting factor is defined as: in, is a tuning parameter, and its typical value range is 0.01 to 0.1. When the prediction error is large, becomes smaller, which increases the weight of new data; when the prediction error is small, The influence of historical data increases. This mechanism ensures that parameter estimation is adjusted quickly when the system state changes dramatically, and remains stable when the system is running stably. After obtaining the real-time estimated values ​​of the system parameters, the eigenvalues ​​of the system characteristic equation are calculated, and the stability margin of the system is calculated based on this. For the servo drive system, the eigenvalues ​​of its dynamic equations are By the Jacobian matrix The eigenvalues ​​of are calculated as: System stability margin It is determined by the minimum value of the real part of the eigenvalue, that is: when If the value is too small, it indicates that the system is prone to instability under external disturbances, resulting in oscillation or performance degradation. Therefore, the stability margin As an important quantitative indicator, it is used to evaluate the health status of the system. In order to obtain the final parameter health index, the relative deviation vector of the system parameters is transformed by an exponential function, thereby mapping it to a health index range between 0 and 1. The parameter health function is defined as follows: in, is the second norm of the relative deviation vector of system parameters, is a scaling factor that determines the rate at which health decreases. When the parameter deviation is small, Close to 1, indicating that the system is in a healthy state; when the parameter deviation increases, Gradually approaching 0, it means that the system status is deteriorating.

[0031] In a specific embodiment, the process of executing step S5 may specifically include the following steps: The normalized feature vector, enhanced fault feature matrix and parameter health index are combined into a unified vector to obtain the fault feature space representation; The fault feature space representation is input into the support vector machine classifier and mapped to the high-dimensional feature space through the radial basis kernel function to obtain the optimized classification hyperplane; Based on the optimized classification hyperplane, the multi-category fault classification function value and fault confidence index are calculated for the input samples to obtain the candidate fault type and confidence score; The fault severity levels are divided according to the candidate fault types and parameter health indicators, and the target fault diagnosis results are output in combination with the system stability margin and the fault severity level.

[0032] Specifically, the standardized feature vector, enhanced fault feature matrix and parameter health index are combined into a unified vector to construct a fault feature space representation. The standardized feature vector is composed of fault feature components, including key information such as voltage fluctuation ratio, motor speed, current amplitude, etc., while the enhanced fault feature matrix integrates the time-frequency analysis results of the current phase angle and the current vector trajectory mapping characteristics. These features can provide a complete state description of the servo drive system under different fault modes. In addition, the parameter health index is used to measure the overall health status of the system, and its calculation is based on the deviation and stability margin evaluation of the system parameters. Therefore, the fault feature space is represented as: in, is the fault feature space representation, is the standardized feature vector, is the enhanced fault feature matrix, is the parameter health index. The fault feature space is represented by As input, it is input to the support vector machine (SVM) classifier and mapped to the high-dimensional feature space through the radial basis kernel function (RBF kernel) to obtain the optimized classification hyperplane. The expression of the radial basis kernel function is: in, are two eigenvectors and The kernel function calculation value between is the kernel function parameter, which determines the scale range of the mapping. The kernel function is used to nonlinearly map the original fault feature space to a high-dimensional space, making the classification hyperplanes between different fault categories more linearly separable. Based on the kernel function calculation results, the classification hyperplane is solved by optimizing the problem: in, is the Lagrange multiplier, For sample The category label of is the total number of samples, is the penalty coefficient of the slack variable, which is used to control the generalization ability of the classifier. After the optimization is completed, the decision function of the classification hyperplane is expressed as: in, is the bias term. This optimized classification hyperplane can effectively distinguish different types of fault categories and provide a highly robust classification boundary. Based on the optimized classification hyperplane, the multi-category fault classification function value is calculated for the input sample, and the fault confidence index is calculated to obtain the candidate fault type and confidence score. For each input sample , the fault classification function value is expressed as: in, is the weight of the support vector, is the eigenvector after kernel function mapping, and For Category By calculating the classification function values ​​of all categories, the candidate fault types are obtained, and the fault confidence scores are calculated at the same time: in, is a temperature parameter used to adjust the degree of nonlinear mapping of classification confidence. The range is [0,1]. The larger the value, the more likely the sample belongs to the fault category. After obtaining the candidate fault type and the corresponding confidence score, combined with the parameter health index The fault severity level is divided into different levels, and the target fault diagnosis result is finally obtained. The fault severity level is mainly divided based on the drift degree of key system parameters and the system stability margin. as follows: ; Among them, when the health High confidence in fault classification When the system is in normal state or a minor fault state, no emergency treatment is required; Decline and When it is lower than 0.75, it indicates that the system may be in a serious fault state and needs timely maintenance; If it further drops below 0.5 and the confidence score is also low, it indicates that the system has a high risk of failure and immediate measures should be taken to repair it.

[0033] The above describes the fault diagnosis method of the servo drive system in the embodiment of the present invention. The following describes the fault diagnosis system of the servo drive system in the embodiment of the present invention. Figure 2 , an embodiment of the fault diagnosis system of the servo drive system in the embodiment of the present invention includes: A modeling module is used to model the servo drive system in a dq coordinate system and obtain a system parameter monitoring matrix; An injection module is used to inject positive virtual signals and negative virtual signals into the observer of the servo drive system based on the system parameter monitoring matrix to obtain d-axis and q-axis voltage and current response data; A construction module, used for constructing a standardized feature vector corresponding to a fault type and an enhanced fault feature matrix according to d-axis and q-axis voltage and current response data; A parameter estimation module is used to calculate the system stability margin and parameter health index of the servo drive system based on the standardized feature vector and the enhanced fault feature matrix; The fault identification module is used to input the standardized feature vector, enhanced fault feature matrix and parameter health index into the support vector machine classifier for fault identification, obtain the fault severity level, and output the target fault diagnosis result according to the system stability margin and the fault severity level.

[0034] Through the collaboration of the above components and the design of positive and negative virtual signal injection strategies, it is found that the fault characteristics are only related to the ratio of the q-axis and d-axis voltage fluctuations, making the diagnosis results completely independent of the position error and the nonlinear characteristics of the driver, greatly improving the stability and reliability of the diagnosis. The high-frequency sampling mechanism combined with the resonance characteristic analysis can capture the characteristic information of the system in the high-frequency domain, and realize high-precision fault detection at low-speed operation (less than 2% of the rated speed), breaking through the limitation of poor diagnostic performance of traditional methods under low-speed conditions. The time-frequency analysis method based on the current phase angle realizes the comprehensive diagnosis of all parameters such as motor resistance, inductance, flux parameters and voltage and current deviations, with a more comprehensive diagnosis range and stronger adaptability. By constructing parameter health indicators and system stability margin evaluation mechanisms, not only can the fault type be identified, but also the fault severity can be quantified. The combination of standardized feature vectors and enhanced fault feature matrices significantly improves the accuracy of fault diagnosis under different load conditions. It has been verified that the diagnostic accuracy rate under various fault conditions is more than 99%. The algorithm combining recursive parameter estimation with support vector machine classification makes the fault diagnosis process adaptive and sensitive to parameter drift, so that potential problems can be discovered in time at the early stage of fault.

[0035] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0036] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0037] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault diagnosis method for a servo drive system, characterized in that: The method comprises: Model the servo drive system in the dq coordinate system and obtain the system parameter monitoring matrix; Based on the system parameter monitoring matrix, positive virtual signals and negative virtual signals are respectively injected into the observer of the servo drive system to obtain d-axis and q-axis voltage and current response data; According to the d-axis and q-axis voltage and current response data, construct a standardized feature vector corresponding to the fault type and an enhanced fault feature matrix; Based on the standardized feature vector and the enhanced fault feature matrix, calculating the system stability margin and parameter health index of the servo drive system; The standardized feature vector, the enhanced fault feature matrix and the parameter health index are input into a support vector machine classifier for fault identification to obtain a fault severity level, and a target fault diagnosis result is output according to the system stability margin and the fault severity level.

2. The fault diagnosis method of the servo drive system according to claim 1, characterized in that: The servo drive system is modeled in the dq coordinate system to obtain a system parameter monitoring matrix, including: The voltage equation in the dq coordinate system is established for the electrical part of the servo drive system, and the characteristic equation including the stator resistance, d-axis inductance, q-axis inductance and permanent magnet flux is obtained; According to the characteristic equation, a virtual parameter set including motor resistance, inductance, flux parameters and voltage and current deviations is designed to obtain a virtual interference signal for injection; Injecting the virtual interference signal into the characteristic equation to obtain an enhanced system model including a parameter disturbance amount, and constructing an observer structure based on the enhanced system model to obtain a dynamic response equation for estimating a d-axis current and a q-axis current; The partial derivatives of each parameter of the enhanced system model with respect to the system dynamic characteristics are calculated to obtain a system parameter monitoring matrix. At the same time, a rank analysis is performed on the system parameter monitoring matrix to obtain a virtual signal injection model that meets the system observability conditions.

3. The fault diagnosis method of the servo drive system according to claim 1, characterized in that: The method of injecting positive virtual signals and negative virtual signals into the observer of the servo drive system based on the system parameter monitoring matrix to obtain d-axis and q-axis voltage and current response data includes: According to the system parameter monitoring matrix, a positive disturbance vector and a negative disturbance vector including motor resistance, inductance, magnetic flux and voltage and current deviation parameters are set to obtain a positive virtual signal δP+ and a negative virtual signal δP-; Applying amplitude constraints to the positive virtual signal δP+ and the negative virtual signal δP- to ensure that the system stability is not affected, and obtaining the positive virtual signal δP+ and the negative virtual signal δP- after the amplitude constraints; The system response is calculated based on the positive virtual signal δP+ after amplitude constraint, and the d-axis and q-axis forward voltage equations including parameter disturbances are obtained; Based on the amplitude-constrained negative virtual signal δP-, the system response is calculated and the negative voltage equations of the d-axis and q-axis including parameter disturbances are obtained. An alternating injection timing and a smooth transition function are designed for the positive virtual signal δP+ after amplitude constraint and the negative virtual signal δP- after amplitude constraint to obtain a time-continuous injection signal, and the voltage and current data of the system in different time periods before and after the injection of the time-continuous injection signal are recorded by the observer of the servo drive system to obtain the d-axis and q-axis voltage and current response data.

4. The fault diagnosis method of the servo drive system according to claim 1, characterized in that: The step of constructing a standardized feature vector and an enhanced fault feature matrix corresponding to the fault type according to the d-axis and q-axis voltage and current response data includes: Based on the d-axis and q-axis voltage and current response data, calculating the d-axis and q-axis voltage deviations caused by the injection of positive virtual signals and negative virtual signals to obtain a voltage deviation sequence; Performing a sliding average filter on the voltage deviation sequence to obtain a filtered voltage deviation value, and calculating a ratio of a d-axis voltage deviation to a q-axis voltage deviation based on the filtered voltage deviation value to obtain a voltage fluctuation ratio parameter under positive and negative injection conditions; Applying threshold judgment processing to the voltage fluctuation ratio parameter, and calculating the voltage fluctuation ratio difference under positive and negative injection conditions, to obtain key fault characteristic parameters; Calculating characteristic components based on the key fault characteristic parameters, motor speed, d-axis current and q-axis current to obtain a four-dimensional fault characteristic vector; Normalizing each component of the fault feature vector by a corresponding scaling factor to obtain a standardized feature vector corresponding to the fault type; The d-axis and q-axis voltage and current response data are used to analyze the time-frequency characteristics of the current phase angle of the servo drive system to obtain an enhanced fault feature matrix.

5. The fault diagnosis method of the servo drive system according to claim 4, characterized in that: The method uses the d-axis and q-axis voltage and current response data to analyze the time-frequency characteristics of the current phase angle of the servo drive system to obtain an enhanced fault feature matrix, including: Calculating the amplitude and phase angle of the current vector according to the d-axis and q-axis voltage and current response data to obtain a current phase characteristic reflecting the system operation state; Performing a short-time Fourier transform on the current phase characteristic to obtain a time-frequency distribution characteristic of the current phase angle; Extracting a phase angle spectrum from the time-frequency distribution characteristics to obtain a high-frequency resonant phase angle characteristic, and performing normalized mapping processing on the d-axis and q-axis currents according to the amplitude of the current vector to obtain a current vector trajectory mapping function under different working conditions; Calculating partial derivatives of each system parameter with respect to the phase angle based on the current vector trajectory mapping function to obtain a phase angle sensitivity matrix representing parameter sensitivity; The phase angle sensitivity matrix is ​​used to determine the sensitive frequency points of the high-frequency resonance phase angle characteristics and their corresponding scaling factors, and the high-frequency resonance phase angle characteristics are scaled at the determined sensitive frequency points to obtain a scaled phase angle feature vector, and the scaled phase angle feature vector is combined with the standardized feature vector to obtain an enhanced fault feature matrix.

6. The fault diagnosis method of the servo drive system according to claim 1, characterized in that: The calculating the system stability margin and parameter health index of the servo drive system based on the standardized feature vector and the enhanced fault feature matrix includes: Calculate the deviations of motor resistance, inductance, flux parameters and voltage and current deviations relative to the rated values ​​to obtain the relative deviation vector of system parameters; Applying an exponential function transformation to the relative deviation vector of the system parameter to obtain a parameter health index; constructing an observation vector and an observation matrix based on the standardized feature vector and the enhanced fault feature matrix to obtain a mathematical model for recursive parameter estimation; Applying the least square method to the mathematical model of the recursive parameter estimation to obtain real-time estimated values ​​of the system parameters; Based on the prediction error in the parameter estimation process, an adaptive forgetting factor is calculated to obtain an updated system parameter estimation value; According to the updated estimated values ​​of the system parameters, the characteristic values ​​of the system characteristic equation are calculated and the minimum absolute value of the real part thereof is obtained to obtain the system stability margin for fault early warning.

7. The fault diagnosis method of the servo drive system according to claim 1, characterized in that: The step of inputting the standardized feature vector, the enhanced fault feature matrix and the parameter health index into a support vector machine classifier for fault identification to obtain a fault severity level, and outputting a target fault diagnosis result according to the system stability margin and the fault severity level includes: Combining the standardized feature vector, the enhanced fault feature matrix and the parameter health index into a unified vector to obtain a fault feature space representation; Input the fault feature space representation into a support vector machine classifier, map it to a high-dimensional feature space through a radial basis kernel function, and obtain an optimized classification hyperplane; Calculating multi-category fault classification function values ​​and fault confidence indicators for input samples based on the optimized classification hyperplane to obtain candidate fault types and confidence scores; The fault severity levels are divided according to the candidate fault types and the parameter health index, and a target fault diagnosis result is output in combination with the system stability margin and the fault severity level.

8. A fault diagnosis system for a servo drive system, characterized in that: Used to perform the fault diagnosis method of the servo drive system according to any one of claims 1 to 7, the fault diagnosis system of the servo drive system comprising: A modeling module is used to model the servo drive system in a dq coordinate system and obtain a system parameter monitoring matrix; An injection module, used for injecting positive virtual signals and negative virtual signals into the observer of the servo drive system respectively based on the system parameter monitoring matrix to obtain d-axis and q-axis voltage and current response data; A construction module, used to construct a standardized feature vector corresponding to the fault type and an enhanced fault feature matrix according to the d-axis and q-axis voltage and current response data; A parameter estimation module, used for calculating the system stability margin and parameter health index of the servo drive system based on the standardized feature vector and the enhanced fault feature matrix; A fault identification module is used to input the standardized feature vector, the enhanced fault feature matrix and the parameter health index into a support vector machine classifier for fault identification, obtain a fault severity level, and output a target fault diagnosis result according to the system stability margin and the fault severity level.

Images