Servo motor fault diagnosis method and system based on neural network

Abstract

The invention relates to the technical field of servo motor fault diagnosis, in particular to a servo motor fault diagnosis method and system based on a neural network. The method comprises four steps of data acquisition and preprocessing, feature analysis and mode construction, real-time monitoring and anomaly detection, and fault diagnosis and feedback optimization. Servo motor operation data are synchronously collected through multiple sensors, features are extracted through a neural network model, and a fault diagnosis model is constructed. The state of the servo motor is monitored in real time, abnormity is recognized, the diagnosis result is optimized based on historical data, and the fault prediction accuracy is improved. The system comprises a data acquisition module, a mode training module, a monitoring and diagnosis module and a feedback optimization module which work cooperatively to realize intelligent fault diagnosis. The fault detection precision can be improved, the downtime of equipment is reduced, and the industrial production efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of servo motor operation status monitoring and fault diagnosis, and particularly to a fault diagnosis method and system based on neural network, which can be widely applied to fields such as intelligent manufacturing, automated production, and robot control. Background Art

[0002] At present, servo motors have been widely used in fields such as industrial automation, robot control, and numerical control machine tools. Due to their characteristics of high precision and high dynamic response, the health status of servo motors has an important impact on the overall operation performance of the system. However, in a complex industrial environment, servo motors are easily affected by factors such as electromagnetic interference, load fluctuations, and mechanical wear, resulting in abnormal operation states and further causing faults. Therefore, how to accurately monitor the status of servo motors and effectively give early warnings before faults occur is a key issue in current industrial equipment maintenance.

[0003] In recent years, neural network technology has been introduced into the field of servo motor fault diagnosis to improve the intelligence level of fault identification. By constructing a deep learning model, features in multi-dimensional data such as current, voltage, vibration, and temperature can be automatically extracted and pattern recognition can be performed. However, most existing studies focus on the depth and complexity of the model, lacking a systematic solution for data preprocessing, feature extraction, and real-time optimization. In addition, during the fault diagnosis process, some models have the problem of overfitting, resulting in insufficient generalization ability and difficulty in adapting to servo motors under different working conditions. Summary of the Invention

[0004] In order to achieve the above-mentioned invention purposes, the present invention provides the following technical solutions: A servo motor fault diagnosis method based on neural network, including the following steps:

[0005] S1. Data acquisition and preprocessing, obtaining current, speed, temperature, and vibration data during the operation of the servo motor, and normalizing, denoising, and extracting features from the data;

[0006] S2. Feature analysis and pattern construction, based on the preprocessed data, using a neural network model for training to generate a fault diagnosis feature library and constructing a reference operation pattern;

[0007] S3. Real-time monitoring and anomaly detection, inputting the current operation status data of the servo motor into the trained neural network model to determine whether there is an anomaly, and if there is an anomaly, locating the specific fault category;

[0008] S4. Fault diagnosis and feedback optimization, according to the anomaly detection results, combining with the historical fault library for comprehensive evaluation, providing fault cause analysis and optimization suggestions, and iteratively updating the model.

[0009] Preferably, in step S1, the data acquisition and preprocessing steps specifically include:

[0010] Synchronously collect the operating data of the servo motor through a current sensor, a speed encoder, a temperature sensor, and a vibration sensor, and perform timestamp marking to ensure the temporal continuity of the data;

[0011] Adopt the wavelet denoising method to denoise the collected data, remove environmental noise, random interference, and sensor errors, and improve the data quality;

[0012] According to the operating state of the servo motor, extract key characteristic parameters, including the root mean square value of current, spectral energy, temperature change rate, and vibration acceleration, and normalize the data to enhance the comparability between different characteristic dimensions;

[0013] Use the principal component analysis method to reduce the dimension of the extracted features, reduce data redundancy, and improve the model calculation efficiency.

[0014] Preferably, in step S2, the feature analysis and pattern construction steps specifically include:

[0015] Adopt an autoencoder neural network to reduce the dimension of the normalized data and extract key features;

[0016] Learn the temporal data characteristics of the servo motor through a long short-term memory network, construct a normal operation mode, and generate a multi-scale fault feature library;

[0017] Set the fault judgment criteria, mark the feature vectors that exceed the normal operation range, and calculate the fault probability. The fault probability calculation adopts a multi-scale dynamic weight adjustment formula, as follows:

[0018]

[0019] Where:

[0020] P i is the probability of the occurrence of fault type i;

[0021] α i is the learned feature weight, used to adjust the importance of a single fault feature;

[0022] f i is the activation value of the feature vector extracted from the neural network corresponding to fault type i;

[0023] β is the global adjustment factor, which balances the influence of neural network features and external supplementary information;

[0024] ω k is the dynamic adjustment weight, which is adaptively updated according to the operating state of the servo motor;

[0025] g k It is an external supplementary feature (including historical fault trends, operating environment parameters, and operation and maintenance records) for enhancing the adaptability of the model;

[0026] According to the fault feature library, a complete fault mode data set is generated and cross-validated with various working condition data to improve the generalization ability of the model.

[0027] Preferably, in step S3, the real-time monitoring and anomaly detection step specifically includes:

[0028] The sliding window method is used to segment the newly collected data, and the corresponding feature parameters are calculated for each window to ensure time series continuity;

[0029] The calculated feature vectors are input into the trained neural network model for feature matching, and the fault probability distribution of the current state is calculated;

[0030] An alarm threshold is set. When the fault probability exceeds the set threshold, an anomaly alarm is triggered and the anomaly type and timestamp are recorded;

[0031] If the anomaly persists, further perform fault mode matching, and comprehensively judge the fault type and cause by means of fuzzy classification;

[0032] Combined with the real-time data trend, short-term prediction of the abnormal state is carried out to give early warning of possible faults.

[0033] Preferably, in step S4, the fault diagnosis and feedback optimization step specifically includes:

[0034] According to the historical fault library data, the Bayesian update method is used to optimize the fault classification model to improve the accuracy of fault recognition;

[0035] Combined with the operation and maintenance records, false alarms and missed alarms are marked, and the parameters of the neural network model are dynamically adjusted to improve the diagnostic effect;

[0036] A fault analysis report is generated, including the fault occurrence time, the change trend of feature parameters, the possible fault type and maintenance suggestions, and is pushed to the operation and maintenance personnel;

[0037] In subsequent operations, the fault recognition strategy is continuously optimized according to the actual operation and maintenance situation, and the model is adaptively trained by means of reinforcement learning to achieve self-learning and evolution.

[0038] The present invention also provides a servo motor fault diagnosis system based on a neural network, including the following modules:

[0039] A data acquisition module for obtaining the operation state data of the servo motor and performing preliminary processing;

[0040] A pattern training module for analyzing the collected data, training a neural network model, and constructing a fault feature library;

[0041] A monitoring and diagnosis module for real-time monitoring of the servo motor status, detecting abnormalities, and performing fault diagnosis;

[0042] A feedback optimization module for optimizing the diagnosis model based on historical fault data to improve the diagnosis accuracy.

[0043] Furthermore, the data acquisition module includes:

[0044] A sensor unit including a current sensor, a speed encoder, a temperature sensor, and a vibration sensor;

[0045] A data processing unit including signal filtering, feature extraction, and data normalization functions to ensure data quality;

[0046] A data storage unit for storing real-time data and historical data and providing a data query interface.

[0047] Furthermore, the pattern training module includes:

[0048] A data screening unit for receiving and screening the servo motor operation data, removing outliers and noise, and performing normalization processing;

[0049] A feature extraction unit for extracting key features such as current, vibration, and temperature and performing dimensionality reduction optimization;

[0050] A model training unit for training a fault diagnosis model based on a neural network algorithm and adjusting the weight parameter α i To optimize the diagnosis accuracy;

[0051] A model verification unit for verifying the trained model using test data and optimizing the training parameters based on the evaluation results.

[0052] Furthermore, the monitoring and diagnosis module includes:

[0053] A fault detection unit for performing feature matching and anomaly recognition on real-time input data based on the trained neural network model;

[0054] A fault classification unit for performing fault mode matching based on the detection results and calculating the probability of fault occurrence;

[0055] An alarm unit for triggering an alarm and pushing a diagnosis report to the operation and maintenance personnel when the fault probability exceeds the threshold.

[0056] Furthermore, the feedback optimization module includes:

[0057] Self-learning unit, which optimizes the neural network model based on historical fault data to improve the accuracy of fault diagnosis;

[0058] Maintenance advice unit, which generates maintenance guidance advice based on the fault classification results to optimize the maintenance process;

[0059] Model update unit, which periodically performs iterative training on the fault model to adapt to the evolution of fault characteristics under different working conditions.

[0060] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0061] Improve the accuracy of fault diagnosis: The present invention adopts a neural network combined with an adaptive feature extraction method, which can accurately identify the operating state and potential faults of the servo motor, reduce false alarms and missed alarms compared with traditional methods, and improve the diagnosis accuracy.

[0062] Enhance the adaptability and generalization ability of the model: Through the dynamic parameter optimization and real-time data feedback mechanism, the model can adapt to the operating state of the servo motor under different working conditions, improve the adaptability to complex industrial environments, and effectively reduce the operation and maintenance costs. Brief Description of the Drawings

[0063] Figure 1 It is a schematic diagram of the method step flow provided by this application;

[0064] Figure 2 It is a schematic diagram of the system module provided by this application;

[0065] Figure 3 It is a schematic diagram of the data acquisition module of this application;

[0066] Figure 4 It is a schematic diagram of the mode training module of this application;

[0067] Figure 5 It is a schematic diagram of the monitoring and diagnosis module of this application;

[0068] Figure 6 It is a schematic diagram of the feedback optimization module of this application; Detailed Description of the Invention

[0069] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below 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 invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0070] Refer to Figure 1, an embodiment of the present invention provides a servo motor fault diagnosis method based on a neural network, including the following steps:

[0071] Step 1: Data acquisition and preprocessing. Obtain the current, speed, temperature, and vibration data of the servo motor during operation, and perform normalization, denoising, and feature extraction on the data.

[0072] In Step 1, a current sensor, a speed encoder, a temperature sensor, and a vibration sensor are used to synchronously collect the operation data of the servo motor, and the time sequence continuity of the data is ensured through timestamp marking. To improve the data quality, first, the collected raw data is processed using the wavelet denoising method to remove environmental noise, random interference, and sensor errors. Subsequently, according to the operation state of the servo motor, key feature parameters are extracted from the data, including the root mean square value of the current, spectral energy, temperature change rate, and vibration acceleration, and all the data is normalized to enhance the comparability between different feature dimensions. Finally, the principal component analysis method is used to perform dimensionality reduction on the extracted features, effectively reducing the data redundancy and improving the model calculation efficiency, thereby providing high-quality input data for the subsequent fault diagnosis model.

[0073] Step 2: Feature analysis and pattern construction. Based on the preprocessed data, a neural network model is used for training to generate a fault diagnosis feature library and construct a benchmark operation pattern.

[0074] In Step 2, first, an autoencoder neural network is used to perform dimensionality reduction on the normalized operation data of the servo motor to extract key features and reduce data redundancy. Subsequently, a long short-term memory network is used to learn the time sequence data characteristics of the servo motor, construct a normal operation pattern, and generate a multi-scale fault feature library containing multiple fault types. On this basis, a fault judgment criterion is set, the feature vectors outside the normal operation range are marked, and the fault probability is calculated based on the multi-scale dynamic weight adjustment formula. Among them, the fault probability is jointly determined by the activation value of the feature extracted by the neural network, the adaptive dynamic weight, and the external supplementary features to enhance the adaptability of the model to different operation states. The formula is as follows:

[0075]

[0076] Where:

[0077] P i is the probability of the occurrence of fault type i;

[0078] α i is the learned feature weight used to adjust the importance of a single fault feature;

[0079] f i is the activation value of the feature vector corresponding to fault type i extracted from the neural network;

[0080] β is the global adjustment factor to balance the influence of neural network features and external supplementary information;

[0081] ω k is the dynamic adjustment weight, which is adaptively updated according to the operating state of the servo motor;

[0082] g k is the external supplementary feature (including historical fault trends, operating environment parameters, and operation and maintenance records), which is used to enhance the adaptability of the model.

[0083] Finally, based on the fault feature library, a complete fault mode data set is generated, and cross-validation is performed in combination with various working condition data to optimize the fault identification ability of the model and improve the generalization performance.

[0084] Step 3: Real-time monitoring and anomaly detection. Input the current operating state data of the servo motor into the trained neural network model to determine whether there is an anomaly. If there is an anomaly, locate the specific fault category.

[0085] In Step 3, first, the sliding window method is used to segment the newly collected operating data of the servo motor, and the corresponding characteristic parameters are calculated within each window to ensure the continuity and effectiveness of the time-series data. Subsequently, the calculated feature vectors are input into the trained neural network model for feature matching, and the fault probability distribution of the current operating state is calculated. According to the set alarm threshold, when the fault probability exceeds the preset value, the system immediately triggers an anomaly alarm, records the anomaly type and the corresponding timestamp for subsequent analysis. If the anomaly persists, the system further performs fault mode matching and comprehensively judges the fault type and possible fault causes using the fuzzy classification method. In addition, combined with real-time data trend analysis, the system can perform short-term prediction on the abnormal state, early warning of possible faults, and ensure the stability and safety of the servo motor operation.

[0086] Step 4: Fault diagnosis and feedback optimization. According to the anomaly detection results, conduct a comprehensive evaluation in combination with the historical fault library, provide fault cause analysis and optimization suggestions, and perform iterative updates on the model.

[0087] In step four, first, based on the historical fault database data, the Bayesian update method is adopted to optimize the fault classification model to improve the accuracy of fault identification. Subsequently, combined with the operation and maintenance records, the misreport and missed report situations that have occurred are marked, and the parameters of the neural network model are dynamically adjusted to further improve the diagnostic effect. When the system detects a fault, a fault analysis report is automatically generated. The report content includes the fault occurrence time, the change trend of characteristic parameters, possible fault types, and corresponding maintenance suggestions, and is pushed to the operation and maintenance personnel in real time for quick response and handling. At the same time, during the subsequent operation process, the system continuously optimizes the fault identification strategy according to the actual operation and maintenance situation, and uses the reinforcement learning method to adaptively train the model, enabling it to continuously self-learn and evolve to meet the fault diagnosis requirements under different working conditions and improve the stability and intelligent level of long-term operation.

[0088] Reference Figures 2 - 6 , an embodiment of the present invention provides a servo motor fault diagnosis system based on a neural network, including the following modules:

[0089] The data acquisition module is used to obtain the servo motor operation state data and perform preliminary processing.

[0090] Among them, the data acquisition module includes:

[0091] The sensor unit includes a current sensor, a speed encoder, a temperature sensor, and a vibration sensor;

[0092] The data processing unit includes signal filtering, feature extraction, and data normalization processing functions to ensure data quality;

[0093] The data storage unit is used to store real-time data and historical data and provide a data query interface.

[0094] The mode training module is used to analyze the collected data and train the neural network model to build a fault feature library.

[0095] Among them, the mode training module includes:

[0096] The data screening unit is used to receive and screen the servo motor operation data, remove outliers and noise, and perform normalization processing;

[0097] The feature extraction unit is used to extract key features such as current, vibration, and temperature, and perform dimensionality reduction optimization;

[0098] The model training unit is used to train the fault diagnosis model based on the neural network algorithm and adjust the weight parameter α i to optimize the diagnostic accuracy;

[0099] The model verification unit is used to verify the trained model using test data and optimize the training parameters based on the evaluation results.

[0100] A monitoring and diagnosis module for real-time monitoring of the servo motor status, detecting abnormalities and performing fault diagnosis.

[0101] Among them, the monitoring and diagnosis module includes:

[0102] A fault detection unit that performs feature matching and anomaly recognition on real-time input data based on a trained neural network model;

[0103] A fault classification unit that performs fault mode matching based on the detection results and calculates the probability of fault occurrence;

[0104] An alarm unit that triggers an alarm and pushes a diagnostic report to the operation and maintenance personnel when the fault probability exceeds the threshold.

[0105] A feedback optimization module for optimizing the diagnostic model based on historical fault data to improve the diagnostic accuracy.

[0106] Among them, the feedback optimization module includes:

[0107] A self-learning unit that optimizes the neural network model based on historical fault data to improve the fault diagnosis accuracy;

[0108] A maintenance advice unit that generates maintenance guidance suggestions based on the fault classification results to optimize the maintenance process;

[0109] A model update unit that periodically performs iterative training on the fault model to adapt to the evolution of fault characteristics under different working conditions.

[0110] It should be noted that, without conflict, the embodiments and the features and technical solutions in the embodiments of the present invention can be combined with each other.

[0111] Obviously, the above-described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. The drawings show the preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing specific embodiments, or perform equivalent replacements on some of the technical features. Any equivalent structure made by using the specification and drawings of the present invention, directly or indirectly applied to other related technical fields, is equally within the scope of the patent protection of the present invention.

Claims

1. A servo motor fault diagnosis method based on a neural network, characterized in that, It includes the following steps: S1. Data acquisition and preprocessing: Obtain the current, speed, temperature, and vibration data of the servo motor during operation, and perform normalization, denoising, and feature extraction on the data; S2. Feature analysis and pattern construction: Based on the preprocessed data, use a neural network model for training to generate a fault diagnosis feature library and construct a reference operation pattern; S3. Real-time monitoring and anomaly detection: Input the current operation status data of the servo motor into the trained neural network model to determine whether there is an anomaly. If there is an anomaly, locate the specific fault category; S4. Fault diagnosis and feedback optimization: According to the anomaly detection results, conduct a comprehensive evaluation in combination with the historical fault library, provide fault cause analysis and optimization suggestions, and perform iterative updates on the model.

2. The servo motor fault diagnosis method based on a neural network according to claim 1, characterized in that The data acquisition and preprocessing step specifically includes: Synchronously collect the operation data of the servo motor through a current sensor, a speed encoder, a temperature sensor, and a vibration sensor, and perform timestamp marking to ensure the temporal continuity of the data; Adopt the wavelet denoising method to perform noise reduction processing on the collected data, remove environmental noise, random interference, and sensor errors, and improve the data quality; According to the operation status of the servo motor, extract key feature parameters, including the root mean square value of the current, spectral energy, temperature change rate, and vibration acceleration, and perform normalization processing on the data to enhance the comparability between different feature dimensions; Use the principal component analysis method to reduce the dimension of the extracted features, reduce data redundancy, and improve the model calculation efficiency.

3. A servo motor fault diagnosis method based on a neural network according to claim 1, characterized in that The feature analysis and pattern construction step specifically includes: Adopt an autoencoder neural network to perform feature dimension reduction on the normalized data and extract key features; Learn the temporal data characteristics of the servo motor through a long short-term memory network, construct a normal operation pattern, and generate a multi-scale fault feature library; Set a fault judgment criterion, mark the feature vectors that exceed the normal operation range, and calculate the fault probability. The calculation of the fault probability adopts a multi-scale dynamic weight adjustment formula as follows: Where: P i is the probability of occurrence of fault type i; α i is the feature weight for learning, which is used to adjust the importance of a single fault feature; f i is the activation value corresponding to the fault type i of the feature vector extracted in the neural network; β is the global adjustment factor, which balances the influence of neural network features and external supplementary information; ω k For dynamically adjusting the weight, it is adaptively updated according to the operating state of the servo motor; g k It is an external supplementary feature (including historical fault trends, operating environment parameters, and operation and maintenance records) used to enhance the adaptability of the model; Generate a complete fault pattern data set based on the fault feature library, and perform cross-validation in combination with various working condition data to improve the generalization ability of the model.

4. A servo motor fault diagnosis method based on a neural network according to claim 1, characterized in that The real-time monitoring and anomaly detection step specifically includes: Adopt the sliding window method to segment the newly collected data, calculate the corresponding feature parameters for each window to ensure temporal continuity; Input the calculated feature vectors into the trained neural network model for feature matching, and calculate the fault probability distribution of the current state; Set an alarm threshold. When the fault probability exceeds the set threshold, trigger an anomaly alarm and record the anomaly type and timestamp; If the anomaly persists, further perform fault pattern matching, and comprehensively judge the fault type and cause through a fuzzy classification method; Combine the real-time data trend to perform short-term prediction on the abnormal state and give early warnings of possible faults in advance.

5. A servo motor fault diagnosis method based on a neural network according to claim 1, characterized in that, The fault diagnosis and feedback optimization step specifically includes: According to the data in the historical fault library, use the Bayesian update method to optimize the fault classification model and improve the fault recognition accuracy; Combined with the operation and maintenance records, annotate the false alarms and missed alarms, and dynamically adjust the neural network model parameters to improve the diagnostic effect; Generate a fault analysis report, including the fault occurrence time, the change trend of characteristic parameters, possible fault types, and maintenance suggestions, and push it to the operation and maintenance personnel; During subsequent operation, continuously optimize the fault identification strategy according to the actual operation and maintenance situation, and use the reinforcement learning method for model adaptive training to achieve self-learning and evolution.

6. A servo motor fault diagnosis system based on a neural network, characterized in that, It includes the following modules: Data acquisition module, used to obtain the operation status data of the servo motor and perform preliminary processing; Mode training module, used to analyze the collected data and train the neural network model to build a fault feature library; Monitoring and diagnosis module, used to monitor the status of the servo motor in real time, detect abnormalities and perform fault diagnosis; Feedback optimization module, used to optimize the diagnostic model based on historical fault data to improve the diagnostic accuracy.

7. The servo motor fault diagnosis system based on a neural network according to claim 6, wherein, The data acquisition module includes: Sensor unit, including current sensor, speed encoder, temperature sensor, and vibration sensor; Data processing unit, including signal filtering, feature extraction, and data normalization processing functions to ensure data quality; Data storage unit, used to store real-time data and historical data, and provide a data query interface.

8. A servo motor fault diagnosis system based on a neural network, characterized in that, The mode training module includes: Data screening unit, used to receive and screen the operation data of the servo motor, remove outliers and noise, and perform normalization processing; Feature extraction unit, used to extract key features such as current, vibration, and temperature, and perform dimensionality reduction optimization; A model training unit for training a fault diagnosis model based on a neural network algorithm and adjusting the weight parameter α i to optimize the diagnosis accuracy; Model verification unit, used to verify the trained model with test data and optimize the training parameters based on the evaluation results.

9. The servo motor fault diagnosis system based on a neural network according to claim 6, characterized in that, The monitoring and diagnosis module includes: Fault detection unit, based on the trained neural network model, perform feature matching and anomaly identification on real-time input data; Fault classification unit, perform fault mode matching according to the detection results and calculate the probability of fault occurrence; Alarm unit, when the fault probability exceeds the threshold, trigger an alarm and push the diagnostic report to the operation and maintenance personnel.

10. The servo motor fault diagnosis system based on neural network according to claim 6, characterized in that, The feedback optimization module includes: Self-learning unit, optimize the neural network model based on historical fault data to improve the fault diagnosis accuracy; Maintenance suggestion unit, generate maintenance guidance suggestions according to the fault classification results to optimize the maintenance process; Model update unit, periodically perform iterative training on the fault model to adapt to the evolution of fault characteristics under different working conditions.

Images