Fault prediction method and device of servo system, equipment and storage medium

Abstract

The invention discloses a fault prediction method and device for a servo system, equipment and a storage medium. When the method provided by the embodiment of the invention is executed, firstly, real-time operation data of the target servo system are collected according to the state reference variable; then, observation vectors are constructed by using the data, and corresponding estimation vectors are generated in combination with a memory matrix of the target servo system. And calculating the estimation vector and the observation vector to obtain a residual error value, and when the residual error exceeds a preset threshold value, obtaining new real-time operation data again so as to continuously monitor the system state. Through accumulative calculation of a sliding window, the residual error is comprehensively evaluated, and once the residual error exceeds a set threshold value, a fault prompt is sent out in time. According to the invention, the hardware cost of fault prediction can be significantly reduced.

Description

Technical Field

[0001] This application relates to the technical field of fault prediction and diagnosis, and particularly to a fault prediction method, device, equipment and storage medium for a servo system. Background Art

[0002] The permanent magnet synchronous motor mechanical servo system combines the motor system and various mechanical structures (such as gears, lead screws, etc.), which makes the system more complex and precise compared to the traditional permanent magnet synchronous motor system. With the introduction of mechanical structures such as gears and lead screws, the non-linear factors inside the system increase significantly, thus bringing more potential fault risks. Currently, relying on installing multiple sensors to monitor the potential fault problems brought by mechanical structures is an effective method, but it also comes with high costs.

[0003] Therefore, how to effectively reduce the monitoring cost of the permanent magnet synchronous motor mechanical servo system is a technical problem that needs to be urgently solved by those skilled in the art. Summary of the Invention

[0004] Based on the above problems, this application provides a fault prediction method, device, equipment and storage medium for a servo system, which can effectively reduce the monitoring cost of the permanent magnet synchronous motor mechanical servo system.

[0005] The embodiments of this application disclose the following technical solutions:

[0006] A fault prediction method for a servo system, the method includes:

[0007] Obtain a piece of real-time operation data of the target servo system from the target servo system according to the state reference variable;

[0008] Construct an observation vector according to the real-time operation data, and construct an estimation vector according to the memory matrix of the target servo system and the observation vector;

[0009] Subtract the estimation vector from the observation vector and take the modulus value to obtain a residual;

[0010] When the residual is greater than a preset threshold, obtain new real-time operation data again, so as to use the new real-time operation data to execute the steps of constructing the observation vector according to the real-time operation data and subsequent steps, and perform residual accumulation calculation on the residual through a sliding window to obtain a residual sum;

[0011] When the residual sum is greater than the preset threshold, send a system fault reminder.

[0012] In a possible implementation manner, the method further includes:

[0013] When the residual is less than a preset threshold, it is determined whether the memory matrix contains the observation vector;

[0014] If the memory matrix does not contain the observation vector, insert the observation vector into the memory matrix to update the memory matrix;

[0015] Re-obtain new real-time operation data to execute the steps of constructing the observation vector and subsequent steps according to the new real-time operation data and the updated memory matrix, and perform residual accumulation calculation on the residual through a sliding window to obtain the residual sum.

[0016] In a possible implementation manner, the method further includes:

[0017] If the memory matrix does not contain the observation vector, re-obtain new real-time operation data to execute the steps of constructing the observation vector and subsequent steps according to the new real-time operation data, and perform residual accumulation calculation on the residual through a sliding window to obtain the residual sum.

[0018] In a possible implementation manner, the method further includes:

[0019] Perform a redundancy removal operation on the updated memory matrix;

[0020] If the number of vector columns of the memory matrix after the redundancy removal operation is greater than the column number threshold, calculate the excess number of columns according to the number of vector columns of the memory matrix and the column number threshold;

[0021] According to the excess number of columns, randomly delete column vectors from the memory matrix to obtain a new memory matrix.

[0022] In a possible implementation manner, the first construction process of the memory matrix includes:

[0023] Obtain N pieces of historical normal operation data of the target servo system from the target servo system according to the state reference variable; one piece of historical normal operation data corresponds to the state reference variable of the servo system at a certain normal operation historical moment; N is a positive integer;

[0024] Perform preprocessing on the N pieces of historical normal operation data to obtain N pieces of preprocessed data;

[0025] Extract M pieces of representative historical normal operation data from the N pieces of preprocessed data and organize them into a memory matrix; M is a positive integer and M is less than N.

[0026] In a possible implementation manner, constructing an estimation vector according to the memory matrix of the target servo system and the observation vector includes:

[0027] Using the estimation vector calculation formula, the estimation vector is calculated by combining the observation vector and the memory matrix to obtain the estimation vector;

[0028] Among them, the estimation vector calculation formula is X est is the estimation vector; D is the memory matrix; X obs is the observation vector.

[0029] In a possible implementation manner, the state reference variables include the position, speed, quadrature axis current, and direct axis current of the servo system.

[0030] A fault prediction device for a servo system, the device includes:

[0031] A real-time operation data acquisition unit, configured to acquire a piece of real-time operation data of the target servo system from the target servo system according to the state reference variable;

[0032] An observation vector construction unit, configured to construct an observation vector according to the real-time operation data;

[0033] An estimation vector construction unit, configured to construct an estimation vector according to the memory matrix of the target servo system and the observation vector;

[0034] A residual calculation unit, configured to subtract the estimation vector from the observation vector and take the modulus value to obtain a residual;

[0035] A first integration unit, when the residual is greater than a preset threshold, is configured to re-acquire new real-time operation data, so as to use the new real-time operation data to execute the steps of constructing the observation vector according to the real-time operation data and subsequent steps, and perform residual accumulation calculation on the residual through a sliding window to obtain a residual sum;

[0036] A fault reminder unit, when the residual sum is greater than the preset threshold, is configured to issue a system fault reminder.

[0037] A fault prediction device for a servo system, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the fault prediction method for the servo system as described above is implemented.

[0038] A computer-readable storage medium, in which instructions are stored. When the instructions run on a terminal device, the terminal device is caused to execute the fault prediction method for the servo system as described above.

[0039] Compared with the prior art, the present application has the following beneficial effects:

[0040] The present application provides a fault prediction method, device, equipment and storage medium for a servo system. Specifically, when implementing the fault prediction method for the servo system provided in the embodiments of the present application, first, a piece of real-time operation data can be obtained from the target servo system according to the state reference variable, which is used as a reflection of the system state. An observation vector is constructed based on the real-time operation data, and at the same time, an estimated vector is constructed by using the memory matrix of the target servo system and the observation vector. Then, the difference between the estimated vector and the observation vector is calculated and its modulus value is taken to obtain the residual. The residual reflects the deviation between the estimated value and the actual observed value. If the deviation is large, it may indicate system abnormality. Therefore, when the residual is greater than the preset threshold, new real-time operation data is re-obtained, and the observation vector, estimated vector and residual are recalculated through the same steps. In addition, the sliding window technique is used to cumulatively calculate the residual to obtain the sum of residuals, which is used to further evaluate the health state of the system. If the sum of residuals exceeds the preset threshold, a fault reminder is triggered to indicate the risk of possible system failure. Through the construction and analysis of real-time operation data and estimated vectors, the present application avoids the investment in a large number of hardware sensors, reduces the number of sensors and maintenance costs. At the same time, through real-time data acquisition based on state reference variables, as well as differential and modulus value calculations of the data, fault monitoring can be carried out from more dimensions and detailed levels, so as to capture system abnormalities more accurately, especially in the case of non-linear and dynamic changes. In addition, through the cumulative calculation of residuals in the sliding window, the system can adapt to complex non-linear behaviors. As the time window changes, potential faults caused by non-linear changes in mechanical components can be detected more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] To more clearly illustrate the technical solutions in the embodiments or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0042] Figure 1 It is a flowchart of a method for a fault prediction method of a servo system provided by an embodiment of the present application;

[0043] Figure 2 It is a flowchart of a method for the first construction method of a memory matrix provided by an embodiment of the present application;

[0044] Figure 3 It is a flowchart of another fault prediction method for a servo system provided by an embodiment of the present application;

[0045] Figure 4 It is a flowchart of a method for updating a memory matrix provided by an embodiment of the present application;

[0046] Figure 5 This is a schematic structural diagram of a fault prediction device for a servo system provided by an embodiment of the present application. Specific embodiments

[0047] To facilitate the understanding of the technical solutions provided by the embodiments of the present application, the background technology related to the embodiments of the present application will be described first below.

[0048] The permanent magnet synchronous motor mechanical servo system combines the motor system and mechanical structures (such as gears, lead screws, etc.), making the system more complex than the traditional pure motor drive system. The traditional permanent magnet synchronous motor system usually only focuses on the operating characteristics of the motor itself, while the mechanical servo system not only includes the motor but also involves the mechanical parts connecting the motor and the load, such as gears, lead screws, bearings, etc. These mechanical structures play a role in transmitting and converting torque, and at the same time make the working characteristics of the system more diverse and complex.

[0049] After adding mechanical components such as gears and lead screws, the power transmission method in the system becomes more complex. The traditional permanent magnet synchronous motor mainly focuses on factors such as the electromagnetic torque, speed, and rotation angle of the motor, while in the mechanical servo system, factors such as the friction, elastic deformation, lubrication state, and transmission error of the mechanical transmission components will directly affect the operation performance of the motor. These factors usually manifest as the non-linear characteristics of the system. For example, the clearance, elastic deformation during the gear meshing process, and the friction force caused by the load change will all make the performance of the motor drive system more complex and unpredictable. These non-linear factors brought by these mechanical parts will trigger a series of potential fault problems. For example, the wear of the gear may lead to poor meshing, thereby increasing vibration and noise, and even may cause drive failure; the error or wear of the lead screw may affect the transmission accuracy, resulting in an increase in position error and affecting the system control accuracy. These non-linear factors make the fault mode more complex and increase the demand for monitoring the health status of the system.

[0050] For the mechanical structure part in the permanent magnet synchronous motor mechanical servo system, many modern mechanical servo systems adopt a multi-sensor monitoring method for fault diagnosis. These sensors are usually arranged on key components such as gears and lead screws to monitor signals such as temperature and sound in real time and capture abnormal changes in mechanical components. However, this method brings significant cost pressure.

[0051] To solve this problem, an embodiment of the present application provides a method, device, equipment, and storage medium for fault prediction of a servo system. First, real-time operation data is obtained from the target servo system according to state reference variables, then an observation vector is constructed based on the real-time operation data, and an estimated vector is generated in combination with the memory matrix of the target servo system. Next, the difference between the estimated vector and the observation vector is taken and the modulus value is obtained as the residual. If the residual is greater than a preset threshold, new real-time operation data is re-obtained and the above steps are repeated. At the same time, the residual is cumulatively calculated through a sliding window to obtain the sum of residuals. When the sum of residuals is greater than the preset threshold, a system fault reminder is issued. The present application analyzes real-time operation data and the memory matrix to construct an observation vector and an estimated vector, thereby calculating the residual and performing fault detection. This method reduces the dependence on a large number of sensors and greatly reduces the hardware cost in the prediction process.

[0052] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present application.

[0053] See Figure 1 , which is a flowchart of a method for fault prediction of a servo system provided by an embodiment of the present application. As Figure 1 shown, the method for fault prediction of the servo system may include steps S101 - S105:

[0054] S101: Obtain one piece of real-time operation data of the target servo system from the target servo system according to state reference variables.

[0055] According to the expression "obtain one piece of real-time operation data of the target servo system from the target servo system according to state reference variables", it means that by monitoring and recording the key parameters (i.e., state reference variables) in the servo system, the current operating state information of the system is obtained.

[0056] Specifically, the state reference variables include: Position, Speed, quadrature-axis current (Iq), and direct-axis current (Id) of the servo system. These variables are key data used to describe the working state of the servo system. By collecting these variables in real time, the performance of the system at a specific moment can be reflected.

[0057] Position and speed: The servo system is usually equipped with position sensors (such as encoders) and speed sensors for real-time measurement and feedback of the current position and motion speed of the motor or actuator. Through these sensors, the real-time values of position and speed can be obtained.

[0058] Quadrature axis current and direct axis current: A servo motor generally has a quadrature axis current and a direct axis current, and these two currents reflect the load condition and driving force of the motor. The quadrature axis current is usually related to the torque of the motor, while the direct axis current reflects the load condition of the motor in another dimension. Through a current sensor, these two current values can be monitored in real time.

[0059] These state reference variables are fed back to the servo control system in real time through sensors. The control system uses a data acquisition module or real-time data stream to read the numerical values of these variables from the sensors. Then, these numerical values are aggregated and transmitted to the control system or monitoring system, thus forming a real-time operating data stream that reflects the current working state of the target servo system.

[0060] S102: Construct an observation vector based on the real-time operating data, and construct an estimation vector based on the memory matrix of the target servo system and the observation vector.

[0061] During the fault prediction process of the servo system, it is also necessary to construct an observation vector based on the real-time operating data of the target servo system, and then use the memory matrix of the target servo system and this observation vector to construct an estimation vector.

[0062] It should be noted that the observation vector is a mathematical representation extracted from the real-time operating data, which usually includes various dynamic characteristics or measurement data of the system. These data can include key variables such as the position, speed, quadrature axis current (Iq), and direct axis current (Id) of the servo system, reflecting the actual operating state of the system at a certain moment. Denote these 4 variables observed at a certain moment i as the observation vector: X (i) =[x position , x speed , x Iq , x Id T . The purpose of constructing the observation vector is to extract key information related to faults, so that subsequent estimations can accurately reflect the behavior of the system.

[0063] The estimation vector is calculated based on the memory matrix of the target servo system and the observation vector. The memory matrix is a mathematical tool that describes the historical behavior of the system, usually constructed through the state data of the system over a period of time in the past (for example, historical normal operating data). The memory matrix can help the system understand and predict its future operating behavior.

[0064] This process usually combines the current observation information with the past historical states to estimate the current "predicted state" or future behavior of the system. In this way, the estimation vector provides the system with a "speculation" on the current and future states based on historical data. ​

[0065] In short, the observation vector is constructed based on current real-time data, and the estimation vector is calculated by combining historical data and current observation information with the help of the memory matrix. The combined use of both can help the system perform more accurate fault diagnosis and prediction during the prediction process.

[0066] See Figure 2 , Figure 2 which is a flowchart of a method for initially constructing a memory matrix provided by an embodiment of the present application. Specifically, the initial construction process of the memory matrix can be implemented through steps A1 - A3:

[0067] A1: Obtain N pieces of historical normal operation data of the target servo system from the target servo system according to the state reference variable.

[0068] By collecting the state reference variables of the target servo system over a period of time in the past. These variables can be, for example, position, speed, acceleration, etc., depending on the design and monitoring requirements of the system. Each piece of historical normal operation data corresponds to the state variables at a certain normal operation historical moment. From the target servo system

[0069] A2: Preprocess the N pieces of historical normal operation data to obtain N pieces of preprocessed data.

[0070] After obtaining the N pieces of historical normal operation data, it is necessary to preprocess the collected N pieces of historical data to ensure the quality and consistency of the data.

[0071] In a possible implementation manner, the preprocessing includes: data cleaning (such as removing duplicate data, handling missing values, and removing outliers), standardization, and normalization, etc.

[0072] Through preprocessing, N pieces of preprocessed data are obtained, and these data will be used for the next feature extraction.

[0073] A3: Extract M pieces of representative historical normal operation data from the N pieces of preprocessed data and organize them into a memory matrix.

[0074] After obtaining the N pieces of preprocessed data, the most representative M pieces of data can be selected from the N pieces of preprocessed data to form a memory matrix.

[0075] Specifically, first select M pieces of data that are of great significance for system fault prediction or state estimation through certain criteria (such as: representativeness of data distribution, time series characteristics, etc.). Then organize these M pieces of representative data into a matrix form to obtain the memory matrix D. Usually, the state variables of each piece of data are used as the rows or columns of the matrix to form a structured memory matrix D that describes the historical state:

[0076]

[0077] It should be noted that M is a positive integer and less than N (N∶M can be set to 10:1, but this application does not specifically limit the ratio of N and M, and users can adjust the ratio of N and M according to actual needs), which means selecting M representative data from the original N pieces of data. The memory matrix is the result of this selection process and will be used for fault prediction and state estimation in subsequent steps.

[0078] The construction process of the memory matrix is carried out through the above three steps. First, historical data is obtained from the target servo system; then the data is preprocessed to ensure its quality; finally, the most representative data is extracted from the preprocessed data through feature selection to construct the memory matrix. The memory matrix contains the key information of the system at historical moments and provides necessary support for subsequent fault prediction and state estimation.

[0079] In a possible implementation, constructing an estimation vector according to the memory matrix and the observation vector of the target servo system includes:

[0080] Using the estimation vector calculation formula, combining the observation vector and the memory matrix to calculate the estimation vector to obtain the estimation vector;

[0081] Among them, the estimation vector calculation formula is X est is the estimation vector; D is the memory matrix; X obs is the observation vector.

[0082] It should be noted that for any matrix A and matrix B (A and B have the same number of rows), n is the total number of rows of matrix A or matrix B, and i is the number of rows of the matrix.

[0083] S103: Subtract the observation vector from the estimation vector and take the modulus value to obtain the residual.

[0084] "Subtract the observation vector from the estimation vector and take the modulus value to obtain the residual" means comparing the difference between the estimation vector and the observation vector. The estimation vector is the expected state inferred from the memory matrix and real-time data, while the observation vector is the actual state directly obtained from the system. By calculating the difference between these two vectors (i.e., subtraction), a residual vector representing the difference between the two can be obtained. Then, by taking the modulus value of this residual vector, the overall difference degree between the two vectors can be quantified.

[0085] The smaller the residual, the closer the estimation vector is to the actual observed value; the larger the residual, the greater the deviation between the estimation vector and the actual observed value.

[0086] S104: When the residual is greater than a preset threshold, obtain new real-time operation data again to execute the steps of constructing an observation vector based on the real-time operation data and subsequent steps using the new real-time operation data, and perform cumulative calculation of the residual through a sliding window to obtain a residual sum.

[0087] When the residual is greater than the preset threshold, it indicates that the current state of the system deviates too much from the prediction, and it is necessary to consider that the system may be abnormal. At this time, the system will obtain new real-time data again, update the observation vector with the new data, and continue the subsequent steps. By using the sliding window method, the system will perform cumulative calculation on the residual to obtain a residual sum, so as to better evaluate the overall state of the system and determine whether more intervention measures need to be taken.

[0088] Specifically, the residual is the difference between the actual value and the estimated value, which can reflect the deviation between the current state of the system and the predicted state. If the residual is large, it means that the system may have an abnormality or a fault. To determine whether this abnormality actually exists, the system will set a preset threshold. If the calculated residual is greater than this threshold, it means that the current system state deviates from the normal range and there may be a fault or an abnormality. When the residual is greater than the preset threshold, the system will consider that the current state information is inaccurate or abnormal, so it is necessary to obtain new real-time operation data again. Once the new real-time data is obtained, the system will reconstruct the observation vector according to these data. Then, the system will use these new observation vectors to perform subsequent processing steps, including updating the estimated vector, recalculating the residual, etc. To process time series data in a dynamic system, the sliding window technique is usually used. The sliding window method can avoid the influence of too much historical information on the current calculation by only considering the historical data within a certain period of time. In this process, the system will perform cumulative calculation on the residual according to the sliding window to obtain a residual sum. This residual sum is the cumulative value of all residuals in the past period of time and is used to reflect the overall trend of the system state. If the residual sum continues to increase, it may indicate that the abnormal situation of the system persists or gradually worsens, and further measures may need to be taken, such as alarming or performing fault troubleshooting, etc.

[0089] Exemplarily, assume that the window size is 3, that is, the residual sum is calculated within the range of every three data points.

[0090] Assume that the residual sequence is (e_1, e_2, e_3, e_4, e_5).

[0091] At the beginning, the sliding window includes (null, e_1, e_2), and at this time, the residual sum is not calculated.

[0092] After the sliding window slides one step backward, it includes (e_1, e_2, e_3), and the residual sum is calculated: (e_1 + e_2 + e_3).

[0093] After the sliding window slides one more step backward, it includes (e_2, e_3, e_4), and calculate the residual sum: (e_2 + e_3 + e_4).

[0094] And so on. Each time it slides one step, a new residual sum is recalculated.

[0095] If the residual sum at a certain moment is greater than the set threshold, it is considered that the system has a fault and a warning is issued.

[0096] It should be noted that when there is a data gap in the sliding window, the residual sum calculation is not performed.

[0097] Generally speaking, the cumulative residual sum of the sliding window effectively detects anomalies in the system operation by calculating the sum of residuals of multiple consecutive data points within a certain time period.

[0098] See Figure 3 , Figure 3 which is the flowchart of another method for fault prediction of a servo system provided by an embodiment of the present application, and can be specifically implemented through steps B1 - B3.

[0099] B1: When the residual is less than the preset threshold, determine whether the memory matrix contains the observation vector.

[0100] If the residual is very small, it means that the system prediction is relatively accurate. Next, it is necessary to consider whether to incorporate the current observed data into the memory matrix. If the observation vector is not in the memory matrix, it needs to be added to update the model or for future calculations.

[0101] B2: If the memory matrix does not contain the observation vector, insert the observation vector into the memory matrix to update the memory matrix.

[0102] If the memory matrix does not contain the observation vector (which may be judged by a certain similarity calculation or matching mechanism), then insert the current observation vector into the memory matrix. Update the memory matrix to make it contain the latest observed data. This strengthens the system's memory by adding new observed data and gradually improves the accuracy of the estimated vector prediction.

[0103] B3: Re - obtain new real - time operation data to execute the steps of constructing the observation vector based on the new real - time operation data and the subsequent steps using the new real - time operation data and the updated memory matrix, and calculate the residual sum by cumulative calculation of residuals through a sliding window.

[0104] Obtain new real-time operation data and use this data to interact with the updated memory matrix for the next calculation. This means that new observation data will be obtained each time the residual is less than the preset threshold and an updated memory matrix is obtained, and then the residual will be recalculated. At the same time, the cumulative sum of the residuals is continuously calculated by means of a sliding window. The sliding window is usually a window of a fixed size. As new data points arrive, the window slides forward, discarding old data and adding new data. The sum of residuals represents the cumulative value of the difference between prediction and actual over a period of time, which helps to track changes in system performance.

[0105] In a possible implementation, the method further includes:

[0106] If the observation vector is not included in the memory matrix, obtain new real-time operation data again to use the new real-time operation data to perform the steps of constructing the observation vector according to the real-time operation data and subsequent steps, and perform cumulative calculation of the residuals through a sliding window to obtain the sum of residuals.

[0107] If the observation vector is not included in the memory matrix (possibly judged by a certain similarity calculation or matching mechanism), obtain new real-time operation data and use this data to interact with the memory matrix used last time for the next calculation. This means that new observation data will be obtained each time the residual is less than the preset threshold and there is no need to update the memory matrix, and then the residual will be recalculated. At the same time, the cumulative sum of the residuals is continuously calculated by means of a sliding window. The sliding window is usually a window of a fixed size. As new data points arrive, the window slides forward, discarding old data and adding new data. The sum of residuals represents the cumulative value of the difference between prediction and actual over a period of time, which helps to track changes in system performance.

[0108] See Figure 4 , Figure 4 which is the flowchart of a method for updating a memory matrix provided by an embodiment of the present application, and can be specifically implemented through steps C1 - C3:

[0109] C1: Perform a redundancy removal operation on the updated memory matrix.

[0110] The purpose of the redundancy removal operation is to ensure that the memory matrix does not contain duplicate or overly similar row vectors. This is achieved by a certain method (such as distance metric or similarity metric) to judge which rows in the matrix are redundant and remove them. The memory matrix after redundancy removal can retain more information, while removing useless or overly repeated data, reducing the computational burden.

[0111] Specifically, metrics such as Euclidean distance and cosine similarity can be used to determine the similarity between two row vectors. If the similarity of certain vectors exceeds a preset threshold, these row vectors can be considered redundant, and thus a part of them can be deleted.

[0112] C2: If the number of vector columns of the memory matrix after the redundancy removal operation is greater than the column number threshold, calculate the excess number of columns according to the number of vector columns of the memory matrix and the column number threshold.

[0113] Control the size of the memory matrix. After the redundancy removal operation, if the number of columns of the matrix still exceeds the preset threshold, it indicates that the matrix may contain too many columns, affecting the system efficiency or resource consumption.

[0114] For example, if the set threshold is 10 columns and the memory matrix after the redundancy removal operation has 15 columns, then 5 columns are redundant and need further processing.

[0115] C3: According to the excess number of columns, randomly delete column vectors from the memory matrix to obtain a new memory matrix.

[0116] On the premise of ensuring system efficiency, further reduce the scale of the memory matrix. Randomly delete the excess rows to bring the number of rows of the matrix back to a reasonable range. Randomly select the excess columns from the memory matrix after redundancy removal and delete them. This process can be carried out by random sampling to avoid the system being biased towards deleting certain specific types of vectors. After deleting the excess rows, a new and more concise memory matrix is obtained, and the number of columns returns below the set threshold.

[0117] This mechanism helps prevent the memory matrix from being too large, thereby improving the calculation efficiency, avoiding memory overflow or system performance degradation, and also avoiding having too much similar information in the memory.

[0118] S105: When the residual sum is greater than the preset threshold, send a system fault reminder.

[0119] The mechanism of sending a system fault reminder when the residual is greater than the preset threshold aims to timely identify and handle the deviations and anomalies in the system to ensure the stability and performance of the system. In this way, the system can give an early warning when potential faults occur and take corresponding repair measures to avoid more serious problems.

[0120] Based on the content of S101 - S105, first, according to the state reference variable of the target servo system, a piece of real - time operation data of the servo system is obtained. Then, these real - time data are used to construct an observation vector, and an estimation vector is constructed in combination with the memory matrix of the target servo system. Then, by taking the difference between the estimation vector and the observation vector, a residual is obtained, and its modulus value is taken. The residual reflects the difference between the current operating state and the expected state of the system. If the calculated residual is greater than a preset threshold, new real - time operation data are obtained again. These new data will be used to reconstruct the observation vector and the estimation vector, and the subsequent residual calculation process continues. To ensure the real - time performance and accuracy of fault prediction, the sliding window technique is used to cumulatively calculate the residual to obtain the sum of residuals. When the sum of residuals accumulated within the sliding window exceeds the preset threshold, it indicates that there may be a fault risk in the system. At this time, a fault reminder is issued to prompt the maintenance personnel to conduct inspections or repairs. This application can improve the reliability and efficiency of the system while effectively reducing the cost of fault prediction.

[0121] See Figure 5 , Figure 5 is a schematic structural diagram of a fault prediction device for a servo system provided by an embodiment of the present application. As Figure 5 shown, the fault prediction device for the servo system includes:

[0122] A real - time operation data acquisition unit 501, configured to obtain a piece of real - time operation data of the target servo system from the target servo system according to the state reference variable;

[0123] An observation vector construction unit 502, configured to construct an observation vector according to the real - time operation data;

[0124] An estimation vector construction unit 503, configured to construct an estimation vector according to the memory matrix of the target servo system and the observation vector;

[0125] A residual calculation unit 504, configured to subtract the estimation vector from the observation vector and take the modulus value to obtain a residual;

[0126] A first integration unit 505, when the residual is greater than the preset threshold, is configured to obtain new real - time operation data again, so as to use the new real - time operation data to execute the steps of constructing the observation vector according to the real - time operation data and subsequent steps, and perform cumulative calculation of the residual through a sliding window to obtain the sum of residuals;

[0127] A fault reminder unit 506, when the sum of residuals is greater than the preset threshold, is configured to issue a system fault reminder.

[0128] In a possible implementation manner, the device further includes:

[0129] A judgment unit, which is used to judge whether the observation vector is included in the memory matrix when the residual is less than a preset threshold;

[0130] An observation vector insertion unit, if the observation vector is not included in the memory matrix, is used to insert the observation vector into the memory matrix to update the memory matrix;

[0131] A second integration unit, which is used to re-obtain new real-time operation data, so as to use the new real-time operation data and the updated memory matrix to execute the steps of constructing the observation vector according to the real-time operation data and subsequent steps, and perform residual accumulation calculation on the residual through a sliding window to obtain the residual sum.

[0132] In a possible implementation manner, the device further includes:

[0133] A third integration unit, if the observation vector is not included in the memory matrix, is used to re-obtain new real-time operation data, so as to use the new real-time operation data to execute the steps of constructing the observation vector according to the real-time operation data and subsequent steps, and perform residual accumulation calculation on the residual through a sliding window to obtain the residual sum.

[0134] In a possible implementation manner, the device further includes:

[0135] A redundancy removal unit, which is used to perform redundancy removal operations on the updated memory matrix;

[0136] An extra column number calculation unit, if the number of vector columns of the memory matrix after redundancy removal operation is greater than the column number threshold, is used to calculate the extra column number according to the number of vector columns of the memory matrix and the column number threshold;

[0137] A random deletion unit, which is used to randomly delete column vectors from the memory matrix according to the extra column number to obtain a new memory matrix.

[0138] In a possible implementation manner, the device further includes:

[0139] A historical normal operation data acquisition unit, which is used to acquire N pieces of historical normal operation data of the target servo system from the target servo system according to the state reference variable; one piece of historical normal operation data corresponds to the state reference variable of the servo system at a certain normal operation historical moment; N is a positive integer;

[0140] A preprocessing unit, which is used to preprocess the N pieces of historical normal operation data to obtain N pieces of preprocessed data;

[0141] A memory matrix construction unit, which is used to extract M pieces of representative historical normal operation data from the N pieces of preprocessed data and organize them into a memory matrix; M is a positive integer, and M is less than N.

[0142] In a possible implementation, the estimation vector construction unit 503 is specifically configured to:

[0143] Calculate the estimation vector by using the estimation vector calculation formula and combining the observation vector and the memory matrix to obtain the estimation vector;

[0144] Wherein, the estimation vector calculation formula is X est is the estimation vector; D is the memory matrix; X obs is the observation vector.

[0145] In a possible implementation, the state reference variables include the position, speed, quadrature axis current, and direct axis current of the servo system.

[0146] In addition, an embodiment of the present application further provides a fault prediction device for a servo system, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the fault prediction method for the servo system as described above is implemented.

[0147] In addition, an embodiment of the present application further provides a computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a terminal device, the terminal device is caused to execute the fault prediction method for the servo system as described above.

[0148] The above has introduced in detail a fault prediction method, device, equipment, and storage medium for a servo system provided by the present application. The embodiments in the specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For the relevant parts, refer to the description of the method part. It should be noted that for those of ordinary skill in the art in the technical field of the present application, without departing from the principle of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

[0149] It should be understood that in this application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that there can be three relationships. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (one) of the following" or a similar expression means any combination of these items, including any combination of single items (ones) or plural items (ones). For example, at least one (one) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0150] It should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent in such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

Claims

1. A fault prediction method for a servo system, characterized in that: The method comprises: Acquire a piece of real-time operation data of the target servo system from the target servo system according to the state reference variable; constructing an observation vector according to the real-time operation data, and constructing an estimation vector according to a memory matrix of the target servo system and the observation vector; Subtract the estimated vector from the observed vector and take the modulus value to obtain a residual; When the residual is greater than a preset threshold, new real-time operation data is acquired again, so as to use the new real-time operation data to execute the step of constructing the observation vector according to the real-time operation data and subsequent steps, and to perform residual accumulation calculation on the residual through a sliding window to obtain a residual sum; When the residual sum is greater than the preset threshold, a system fault reminder is issued.

2. The method according to claim 1, characterized in that The method further comprises: When the residual is less than a preset threshold, determining whether the memory matrix contains the observation vector; If the memory matrix does not contain the observation vector, inserting the observation vector into the memory matrix to update the memory matrix; New real-time operation data is reacquired to execute the construction of the observation vector according to the real-time operation data and subsequent steps using the new real-time operation data and the updated memory matrix, and residual accumulation calculation is performed on the residuals through a sliding window to obtain the residual sum.

3. The method according to claim 2, characterized in that The method further comprises: If the memory matrix does not contain the observation vector, new real-time operation data is reacquired to use the new real-time operation data to execute the construction of the observation vector according to the real-time operation data and subsequent steps, and the residuals are accumulated through a sliding window to obtain the residual sum.

4. The method according to claim 2, characterized in that: The method further comprises: Perform redundancy removal operation on the updated memory matrix; If the number of vector columns of the memory matrix after the redundancy removal operation is greater than the column number threshold, the number of redundant columns is calculated according to the number of vector columns of the memory matrix and the column number threshold; According to the redundant number of columns, column vectors are randomly deleted from the memory matrix to obtain a new memory matrix.

5. The method according to claim 1, characterized in that The first construction process of the memory matrix includes: Acquire N pieces of historical normal operation data of the target servo system from the target servo system according to the state reference variable; one piece of historical normal operation data corresponds to the state reference variable of the servo system at a certain normal operation historical moment; N is a positive integer; Preprocessing the N pieces of historical normal operation data to obtain N pieces of preprocessed data; M pieces of representative historical normal operation data are extracted from the N pieces of pre-processed data and organized into a memory matrix; M is a positive integer and M is less than N.

6. The method according to claim 1, characterized in that Constructing an estimation vector according to a memory matrix of the target servo system and the observation vector, comprising: Using an estimation vector calculation formula, combining the observation vector and the memory matrix to perform estimation vector calculation to obtain the estimation vector; Among them, the estimated vector calculation formula is: X est is the estimated vector; D is the memory matrix; X obs is the observation vector.

7. The method according to claim 1, characterized in that The state reference variables include the position, speed, quadrature-axis current and direct-axis current of the servo system.

8. A fault prediction device for a servo system, characterized in that: The device comprises: A real-time operation data acquisition unit, used for acquiring a piece of real-time operation data of the target servo system from the target servo system according to a state reference variable; An observation vector construction unit, used to construct an observation vector according to the real-time operation data; An estimation vector construction unit, configured to construct an estimation vector according to a memory matrix of the target servo system and the observation vector; A residual calculation unit, used for performing a difference between the estimated vector and the observed vector and taking a modulus value to obtain a residual; A first integration unit, when the residual is greater than a preset threshold, is used to reacquire new real-time operation data, so as to use the new real-time operation data to execute the step of constructing the observation vector according to the real-time operation data and subsequent steps, and to perform residual accumulation calculation on the residual through a sliding window to obtain a residual sum; A fault reminder unit is used to issue a system fault reminder when the residual sum is greater than the preset threshold.

9. A servo system fault prediction device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the fault prediction method for the servo system according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a terminal device, the terminal device executes the fault prediction method for a servo system according to any one of claims 1 to 7.

Images