Circuit breaker mechanical wear prediction system based on vibration characteristics

Abstract

The invention provides a circuit breaker mechanical wear prediction system based on vibration characteristics, and compared with the prior art, the circuit breaker mechanical wear prediction system further comprises a vibration signal acquisition module, a signal processing module, a characteristic extraction module, a characteristic screening and fusion module, a control and evaluation module, a correction module and a communication module. According to the system, original vibration signals of a circuit breaker in the opening and closing process are collected, multiple vibration features are extracted after filtering, denoising and time alignment processing, then an optimal feature combination is obtained through a feature screening and fusion algorithm, and a mechanical wear prediction model is constructed by adopting a random forest. And introducing an environmental health factor to correct a prediction result, and finally outputting a corrected wear degree prediction value RHI. The system has the advantages of high prediction precision, strong environmental adaptability, flexible deployment and the like, solves the problems of unstable performance and large prediction error of a traditional method under complex working conditions, and is suitable for online monitoring and predictive maintenance of circuit breaker equipment.

Description

Technical Field

[0001] The present invention relates to the field of low-voltage electrical appliances, and in particular to a circuit breaker mechanical wear prediction system based on vibration characteristics. Background Art

[0002] In power systems, circuit breakers are critical electrical devices widely used for connecting, carrying, and disconnecting circuits. Their primary function is to interrupt current in both normal and fault conditions, protecting electrical equipment from damage caused by overloads, short circuits, and other faults. Circuit breakers typically achieve on-off control through the mechanical contact and separation of moving and stationary contacts. When a circuit fault occurs, the circuit breaker rapidly interrupts the current within a specified timeframe, preventing damage to the electrical system and even fires caused by overcurrent, thermal effects, or arcing. During frequent closing and opening operations, the moving and stationary contacts are subjected not only to repeated mechanical impact but also to electrical stresses such as arc erosion and thermal expansion and contraction. This causes mechanical wear and electrical corrosion, gradually weakening contact pressure and compromising the circuit breaker's switching reliability and long-term stable performance. Therefore, accurately assessing the mechanical wear of circuit breakers is crucial for ensuring the safe operation of power systems.

[0003] This experimental team has been browsing and studying a large amount of relevant records and materials for a long time in relation to the relevant technologies. At the same time, relying on relevant resources and conducting a large number of relevant experiments, after a large number of searches, it was found that there are existing technologies such as CN110567697B, CN109239591A, CN115508695B, and CN114963953A disclosed in the prior art. For example, a high-precision online detection method for the wear rate of circuit breaker contacts disclosed in the prior art includes the collection of the number of circuit breaker on-off lifespans under different ambient temperature parameters and different voltage values, and curve fitting is performed on the collected data to obtain the temperature influence factor and the voltage influence factor; through the preprocessing and curve fitting of the circuit breaker on-off lifespan monitoring data Line fitting is used to realize contact wear rate detection in a continuous large current domain, and the accuracy of curve fitting is improved by the curve fitting error evaluation method; the data curve fitting function of the circuit breaker contact wear rate requires further discretization processing, and the discretization processing comprehensively considers the storage space and system computing capacity of the system processor; finally, the circuit breaker controller system adopts binary table lookup technology to realize the circuit breaker controller's fast online table lookup operation of the contact wear rate, which can realize the contact wear rate detection of the circuit breaker in a continuous large current domain. The high-precision, continuous, and wide-range contact wear rate online detection technology can meet the requirements of low-voltage intelligent power distribution technology for online detection of equipment health status and ultimately meet the requirements of intelligent equipment operation and maintenance.

[0004] The present invention is made in order to solve the common problems in this field, such as insufficient response to environmental conditions, delayed prediction and lack of trend identification ability, high data processing complexity and poor real-time performance, etc. Summary of the Invention

[0005] The purpose of the present invention is to address the deficiencies in the current art and to propose a circuit breaker mechanical wear prediction system based on vibration characteristics.

[0006] In order to overcome the deficiencies of the prior art, the present invention adopts the following technical solutions: A circuit breaker mechanical wear prediction system based on vibration characteristics, comprising a vibration signal acquisition module, a signal processing module, a feature extraction module, a feature screening and fusion module, a control and evaluation module, a correction module, and a communication module. The vibration signal acquisition module is used to collect the original vibration signal of the circuit breaker during the opening and closing process and transmit it to the signal processing module; The signal processing module is used to perform bandpass filtering, denoising and time synchronization alignment on the original vibration signal to obtain a processed signal, and transmit the processed signal to the feature extraction module; The feature extraction module is used to extract feature data from the processed signal and transmit the feature data to the feature screening and fusion module; The feature screening and fusion module is used to screen and fuse features related to the mechanical wear state of the circuit breaker, and transmit the fused feature data to the control and evaluation module; The control and evaluation module is used to predict the mechanical wear of the circuit breaker based on the fused feature data and output a preliminary prediction value; The correction module is used to output a final corrected wear prediction value RHI based on the preliminary prediction value; The communication module is used to send the wear level prediction value RHI of the circuit breaker to the user end.

[0007] Furthermore, the vibration signal acquisition module acquires the vibration signal through an acceleration sensor installed near the circuit breaker operating mechanism or contact mechanism, and the acceleration sensor signal is preprocessed and output through an analog-to-digital conversion chip ADC.

[0008] Furthermore, the characteristic data includes the numerical values ​​of six characteristics of the processed signal, and the six characteristics are respectively numbered 1-6: peak value P, root mean square value RMS, kurtosis P, dominant frequency fdom, spectrum energy Espec and wavelet packet energy Ewave.

[0009] Furthermore, the feature screening and fusion module includes a feature screening unit and a feature fusion unit, wherein the feature screening unit implements the following steps: S101: Perform a large number of test trainings, using the values ​​of the six features in one test training and the actual mechanical wear of the circuit breaker as a sample, and store all samples obtained in the test training as a data set. S102: Generate all non-empty feature subsets containing 1 to 6 features in combination from the six features, generating a total of M = 63 feature subsets. All feature subsets are numbered S1, S2, ..., SM, where the mth feature subset is denoted as Sm, m∈[1,M], Sm contains k features selected from the six features, and the value range of k is 1 to 6. S103: From the historical data, obtain the data of each feature subset respectively, and use the data of the feature subset as the input feature, and use the corresponding actual wear degree value as the target variable. M random forest regression models are obtained by training the input features and target variables. The random forest regression model trained with the mth feature subset Sm is recorded as the mth random forest regression model. S104: Using K test circuit breakers for comparative testing, use the trained M random forest regression models to predict the wear degree values ​​of the K test circuit breakers respectively, and evaluate the prediction performance of each random forest regression model. S105: Measure the prediction accuracy of M random forest regression models: ; Among them, RS m is the prediction deviation of the mth random forest regression model, K is the total number of test circuit breakers, i is the i-th test circuit breaker, y ture，i is the actual wear value of the i-th test circuit breaker, y m，i is the wear degree value predicted by the mth random forest regression model for the i-th test circuit breaker, S106: Select the feature subset with the smallest prediction deviation value as the optimal feature subset S, and use the random forest regression model obtained by training with the optimal feature subset S as the input feature as the target module.

[0010] Furthermore, the calculation method of the wear degree value DOW of the circuit breaker is: , Among them, Vfir is the current volume of the internal mechanical components of the circuit breaker, Vorig is the initial volume of the internal components of the circuit breaker, and Vfir and Vorig are obtained through industrial CT and non-contact three-dimensional imaging calculations. The internal mechanical components include contact assemblies, closing springs, connecting rod mechanisms, transmission slides and opening springs, which are key components that are prone to structural wear during long-term mechanical operation.

[0011] Furthermore, the feature fusion unit implements the following operation steps: S201: Standardize the values ​​of each feature in the optimal feature subset, and express the value of the jth standardized feature in the optimal feature subset as : ; Among them, f j is the value of the jth feature in the optimal feature subset in the data set, and the unit depends on the feature type; μ j is the mean of the jth feature in the data set, and its unit is the same as f j Consistent, σ j is the standard deviation of the jth feature in the data set, and the unit is the same as f j consistent; is dimensionless; S202: Using the data set as training input, the fuzzy C-means clustering algorithm is used to analyze the standardized features and obtain the feature membership matrix U=[u ij ], based on which the fusion weight of each feature in the optimal feature subset is calculated, and The fusion weight is expressed as w j : ; Among them, Q refers to the total number of samples used for fuzzy cluster analysis in the data set, in units of pieces; u qj It represents the fuzzy membership of the qth sample in the dataset on the jth feature in the optimal feature subset, which is calculated by the fuzzy C-means clustering algorithm. Its value range is [0,1] and the unit is dimensionless. w j It reflects the matching degree between the feature and the cluster center in different samples. The larger the value, the higher the consistency or discrimination ability of the feature in most samples. Therefore, its weight in the final fusion feature vector should also be increased accordingly. j The unit of is dimensionless; S203: Based on the fusion weight w obtained in step S202 j , each standardized eigenvalue f in the optimal feature subset S j′ According to the corresponding fusion weight w j Perform weighted summation to obtain the fused eigenvalue F: ; Where J is the total number of features in the optimal feature subset, F is the fused feature value, and F is dimensionless, serving as the input variable of the control and evaluation modules; S204: Obtain the fusion feature values ​​corresponding to all samples from the data set, and express the fusion feature of the Lth sample as F L , the fusion feature set X is expressed as the fusion feature of all samples in the dataset, X={F1,F2…FQ}, and the actual mechanical wear of all circuit breakers in the sample in the data set is taken as the wear set Y, Y={R1,R2…R Q}, where R L is the actual mechanical wear of the circuit breaker in the Lth sample.

[0012] The beneficial effects achieved by the present invention are: 1. Improved prediction accuracy: By constructing a nonlinear mapping model between fusion features and mechanical wear degree and using a random forest algorithm for training, the accuracy and stability of circuit breaker wear state prediction are effectively improved, overcoming the strong feature dependence and poor generalization capabilities of traditional methods.

[0013] 2. Enhanced environmental adaptability: An environmental health factor E, constructed based on factors such as operating time, humidity, and dust exposure, is introduced, and the initial prediction results are corrected in combination with wear trend parameters, ensuring that the system still has reliable prediction performance under complex operating conditions such as high humidity and high dust.

[0014] 3. Trend recognition capability and deployment flexibility: By introducing wear change acceleration and environmental response coefficient, the system can not only assess the current wear level but also perceive its development trend. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but rather the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0016] Figure 1 This is a modular schematic diagram of the short-circuit mechanical wear prediction module based on vibration characteristics of the present invention.

[0017] Figure 2 Schematic diagram of the modularization of the feature screening and fusion module of the present invention.

[0018] Figure 3 Schematic diagram of the surface experiment of the present invention for predicting the response of the RHI value under different environmental and wear trend conditions. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with its embodiments; it should be pointed out that the specific embodiments described herein are only used to explain the present invention and are not used to limit this case. For those skilled in the art, after reviewing the following detailed description, other systems, methods and / or features of this embodiment will become apparent. In addition, the terms describing the positional relationship in the accompanying drawings are only for illustrative purposes and cannot be understood as limiting this patent. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to the specific circumstances.

[0020] Example 1: Combined with the attached Figure 1 , Attachment Figure 2 , Attachment Figure 3 This embodiment constructs a circuit breaker mechanical wear prediction system based on vibration characteristics, including a vibration signal acquisition module, a signal processing module, a feature extraction module, a feature screening and fusion module, a control and evaluation module, a correction module, and a communication module.

[0021] The vibration signal acquisition module acquires the vibration signal through an acceleration sensor installed near the circuit breaker operating mechanism or contact mechanism, and the acceleration sensor signal is pre-processed and output through an analog-to-digital conversion chip ADC.

[0022] The bandpass filter is used to filter out frequency components below 1kHz and above 10kHz in the original vibration signal. The denoising process adopts the wavelet denoising method to remove high-frequency noise or irrelevant interference in the signal. The time synchronization alignment is achieved by the dynamic time warping algorithm to synchronize the vibration signals from different acceleration sensors. The signal processing module is implemented by configuring a digital signal processor DSP or an edge computing gateway.

[0023] The feature extraction module is implemented through a signal analysis program developed based on MATLAB or Python. The feature extraction module, feature screening and fusion module, control and evaluation module and correction module can be deployed in the same server, or deployed in independent edge computing devices or cloud servers according to the actual system architecture and resource configuration requirements. The modules exchange data through a local area network or industrial communication protocol, and have good system scalability and module independence.

[0024] The characteristic data includes the numerical values ​​of six characteristics of the processed signal, and the six characteristics are numbered 1-6 respectively: peak value P, root mean square value RMS, kurtosis P, dominant frequency fdom, spectrum energy Espec and wavelet packet energy Ewave.

[0025] The calculation method of the wear degree value DOW of the circuit breaker is: , Among them, Vfir is the current volume of the internal mechanical components of the circuit breaker, Vorig is the initial volume of the internal components of the circuit breaker, and Vfir and Vorig are obtained through industrial CT and non-contact three-dimensional imaging calculations. The internal mechanical components include contact assemblies, closing springs, connecting rod mechanisms, transmission slides and opening springs, which are key components that are prone to structural wear during long-term mechanical operation.

[0026] The feature screening and fusion module includes a feature screening unit and a feature fusion unit, wherein the feature screening unit implements the following steps: S101: Perform a large number of experimental trainings, using the values ​​of the six features in one experimental training and the actual mechanical wear of the circuit breaker as a sample, and storing all samples obtained in the experimental training as a data set; S102: Generate all non-empty feature subsets containing 1 to 6 features from the six features in a combination manner, generating a total of M = 63 feature subsets. All feature subsets are numbered S1, S2, ..., SM, where the mth feature subset is denoted as Sm, m∈[1,M], Sm contains k features selected from the six features, and the value of k ranges from 1 to 6; S103: Obtain data from each feature subset from the historical data, use the feature subset data as input features, and use the corresponding actual wear degree value as the target variable. Train M random forest regression models based on the input features and target variables. The random forest regression model trained with the mth feature subset Sm is recorded as the mth random forest regression model. S104: Using K test circuit breakers for comparative testing, use the trained M random forest regression models to predict the wear degree values ​​of the K test circuit breakers respectively, and evaluate the prediction performance of each random forest regression model; S105: Measure the prediction accuracy of M random forest regression models: ; Among them, RS m is the prediction deviation of the mth random forest regression model, K is the total number of test circuit breakers, i is the i-th test circuit breaker, y ture，i is the actual wear value of the i-th test circuit breaker, y m，i is the wear degree value predicted by the mth random forest regression model for the i-th test circuit breaker; S106: Select the feature subset with the smallest prediction deviation value as the optimal feature subset S, and use the random forest regression model obtained by training with the optimal feature subset S as the input feature as the target module.

[0027] The feature fusion unit implements the following operation steps: S201: Standardize the values ​​of each feature in the optimal feature subset, and express the value of the jth standardized feature in the optimal feature subset as : ; Among them, f j is the value of the jth feature in the optimal feature subset in the data set, and the unit depends on the feature type; μ j is the mean of the jth feature in the data set, and its unit is the same as f j Consistent, σ j is the standard deviation of the jth feature in the data set, and the unit is the same as f j consistent; is dimensionless; S202: Using the data set as training input, the fuzzy C-means clustering algorithm is used to analyze the standardized features and obtain the feature membership matrix U=[u ij ], based on which the fusion weight of each feature in the optimal feature subset is calculated, and The fusion weight is expressed as w j : ; Among them, Q refers to the total number of samples used for fuzzy cluster analysis in the data set, in units of pieces; u qj It represents the fuzzy membership of the qth sample in the dataset to the jth feature in the optimal feature subset, which is calculated by the fuzzy C-means clustering algorithm. Its value range is [0,1] and the unit is dimensionless. w j It reflects the matching degree between the feature and the cluster center in different samples. The larger the value, the higher the consistency or discrimination ability of the feature in most samples. Therefore, its weight in the final fusion feature vector should also be increased accordingly. j The unit of is dimensionless; S203: Based on the fusion weight w obtained in step S202 j , each standardized eigenvalue f in the optimal feature subset S j′ According to the corresponding fusion weight w j Perform weighted summation to obtain the fused eigenvalue F: ; J is the total number of features in the optimal feature subset, F is the fused feature value, and F is dimensionless, which serves as the input variable of the control and evaluation modules.

[0028] S204: Obtain the fusion feature values ​​corresponding to all samples from the data set, and express the fusion feature of the Lth sample as F L, the fusion feature set X is expressed as the fusion feature of all samples in the dataset, X={F1,F2…F Q}, and the actual mechanical wear of all circuit breakers in the sample in the data set is taken as the wear set Y, Y={R1,R2…R Q}, where R L is the actual mechanical wear of the circuit breaker in the Lth sample.

[0029] The present invention provides a circuit breaker mechanical wear prediction system based on vibration characteristics, which has significant advantages such as high precision, strong robustness and engineering feasibility, and realizes full-process closed-loop control from vibration signal acquisition, signal processing, feature extraction, feature selection and fusion, health assessment, wear correction to remote communication.

[0030] The present invention adopts a random forest regression method based on multi-feature combination to model the mechanical wear state of the circuit breaker. Combining a large amount of experimental data with the optimal feature subset screening mechanism, the model generalization ability and prediction accuracy are greatly improved. At the same time, the fuzzy C-means clustering algorithm is introduced to weight the fused features, combined with the feature representativeness strength calculation, to effectively improve the discriminative power after feature fusion.

[0031] Example 2: Combined with the attached Figure 1 , Attachment Figure 2 and attached Figure 3 In addition to the contents of the above embodiments, the control and evaluation module implements the following steps:

[0032] The correction module implements the following steps: S401: To improve the accuracy of health assessment, the environmental health factor E is introduced to adjust the original prediction results: ; Where TU is the operating time of the circuit breaker in hours, TL is the expected life of the circuit breaker in hours, HA is the average daily humidity of the environment in which the circuit breaker is currently located in %, DDU is the dust exposure concentration in milligrams per cubic meter, HR is the rated allowable humidity of the circuit breaker, DRA is the rated allowable dust concentration of the circuit breaker, A deg is the wear change acceleration, e is the base of the natural logarithm, γ1 is the time load correction factor, γ2 is the humidity exposure correction factor, γ3 is the dust concentration correction factor, S402: Based on the predicted value Rr and the environmental correction factor E, the final corrected wear prediction value RHI is obtained. ; Among them, A deg The calculation formula is as follows: ; R -1tis the wear prediction value measured at a moment that lags behind the current moment by one monitoring cycle, R -2t is the wear prediction value measured at a moment that lags behind the current moment by two monitoring cycles. Δt represents the time interval between the two monitoring cycles. α is the environmental correction response coefficient, which is dimensionless and represents the response degree of the environmental factor E to the wear prediction value. β is the wear change trend response coefficient, which is measured in hours and is used to measure the influence of the wear change acceleration on the final prediction value.

[0033] The calculation formula of the performance evaluation factor E is used to comprehensively quantify the impact of the current operating environment and historical operating conditions of the circuit breaker on its wear trend. By modeling the correction factors of the actual usage time, humidity exposure level and dust exposure degree of the circuit breaker, combining the time load correction coefficient γ1, humidity correction coefficient γ2 and dust concentration correction coefficient γ3, and using the exponential form of the natural logarithm base e to perform nonlinear synthesis of these influences, the environmental health factor E is finally obtained. The performance evaluation factor E is used to improve the accuracy of mechanical wear prediction and enhance the robustness of the system under complex environmental conditions.

[0034] The γ1, γ2, and γ3 coefficients are obtained by those skilled in the art through a large number of experimental measurements, data modeling, and error correction methods to accurately quantify the impact of environmental and operating status factors on the circuit breaker health index assessment, specifically including: γ1 simulates the operating conditions of the circuit breaker under different environmental humidity conditions, records the corresponding mechanical wear process and health index changes, constructs a humidity-degradation relationship model, and compares it with the rated humidity HRA. The coefficient of humidity's influence on the health degradation rate is extracted. After data fitting and residual analysis, the coefficient is calibrated to enable γ1 to reflect the sensitivity of humidity changes to circuit breaker performance. γ2 changes the dust particle concentration in the test environment to observe the impact of foreign matter accumulation in the circuit breaker's motion mechanism on the mechanical response and wear. A nonlinear correlation model between the pollution level and the health index is constructed using the rated dust concentration DRA. The value of γ2 is then determined by fitting, so that γ2 can reasonably reflect the interference intensity of different dust conditions on system performance. γ3 monitors the health index change trajectories of different circuit breakers in continuous operation cycles over a long period of time, calculates the wear decrease rate between two adjacent detections as an indicator of mechanical wear change acceleration, compares it with the actual rising trend of failure risk, fits its influencing factors, and extracts the value of the dynamic change correction coefficient γ3. γ3 is used to adjust the sensitivity of the health index calculation to the rapid aging trend.

[0035] α is determined in advance by a technician in this field by comparing multiple circuit breaker samples that have been exposed to different retention environments but have known wear progress within the same time period, analyzing the impact of the performance evaluation factor E on the actual wear degree prediction deviation, and using the minimum mean square error method to fit and obtain its optimal coefficient value.

[0036] The value of β is established by mapping the wear rate change trend obtained through continuous monitoring between different samples and the actual wear deviation. The optimal value of β is fitted using the regression method, so that β can be dynamically adjusted within a preset range according to the type and operating conditions of the circuit breaker, thereby improving the response sensitivity of the wear trend change to the evaluation results.

[0037] The present invention constructs a nonlinear mapping model between fusion characteristics and mechanical wear degree through the control and evaluation module, combined with the environmental health factor E and wear change acceleration A introduced in the correction module. deg , dynamically correcting prediction results, enabling high-precision and highly adaptable prediction of circuit breaker mechanical wear. This not only effectively improves prediction accuracy under multiple operating conditions and in high-noise environments, but also provides the ability to perceive wear trends. This addresses existing issues such as poor model generalization, weak environmental adaptability, and delayed predictions, offering excellent engineering feasibility and forward-looking operational value.

[0038] Although the present invention has been described above with reference to various embodiments, it will be appreciated that many changes and modifications may be made without departing from the scope of the present invention. That is, the methods, systems, and devices discussed above are examples. Various configurations may omit, replace, or add various processes or components as appropriate. For example, in alternative configurations, the methods may be performed in an order different from that described, and / or various components may be added, omitted, and / or combined. Moreover, the features described with respect to certain configurations may be combined in various other configurations, such as different aspects and elements of the configurations may be combined in a similar manner. In addition, as technology develops, the elements therein may be updated, i.e., many elements are examples and do not limit the scope of the present disclosure or claims. It will also be appreciated that, after reading the contents of the present invention, a technician may make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A circuit breaker mechanical wear prediction system based on vibration characteristics, characterized in that: The circuit breaker mechanical wear prediction system based on vibration characteristics includes a vibration signal acquisition module, a signal processing module, a feature extraction module, a feature screening and fusion module, a control and evaluation module, a correction module and a communication module. The vibration signal acquisition module is used to collect the original vibration signal of the circuit breaker during the opening and closing process and transmit it to the signal processing module; The signal processing module is used to perform bandpass filtering, denoising and time synchronization alignment on the original vibration signal to obtain a processed signal, and transmit the processed signal to the feature extraction module; The feature extraction module is used to extract feature data from the processed signal and transmit the feature data to the feature screening and fusion module; The feature screening and fusion module is used to screen and fuse features related to the mechanical wear state of the circuit breaker, and transmit the fused feature data to the control and evaluation module; The control and evaluation module is used to predict the mechanical wear of the circuit breaker based on the fused feature data and output a preliminary prediction value; The correction module is used to output a final corrected wear prediction value RHI based on the preliminary prediction value; The communication module is used to send the wear level prediction value RHI of the circuit breaker to the user end.

2. The circuit breaker mechanical wear prediction system according to claim 1, characterized in that: The vibration signal acquisition module acquires the vibration signal through an acceleration sensor installed near the circuit breaker operating mechanism or contact mechanism, and the acceleration sensor signal is pre-processed and output through an analog-to-digital conversion chip ADC.

3. The circuit breaker mechanical wear prediction system according to claim 1, wherein: The characteristic data includes the numerical values ​​of six characteristics of the processed signal, and the six characteristics are numbered 1-6 respectively: peak value P, root mean square value RMS, kurtosis P, dominant frequency fdom, spectrum energy Espec and wavelet packet energy Ewave.

4. The circuit breaker mechanical wear prediction system according to claim 1, wherein: The feature screening and fusion module includes a feature screening unit and a feature fusion unit, wherein the feature screening unit implements the following steps: S101: Perform a large number of test trainings, using the values ​​of the six features in one test training and the actual mechanical wear of the circuit breaker as a sample, and store all samples obtained in the test training as a data set. S102: Generate all non-empty feature subsets containing 1 to 6 features in combination from the six features, generating a total of M = 63 feature subsets. All feature subsets are numbered S1, S2, ..., SM, where the mth feature subset is denoted as Sm, m∈[1,M], Sm contains k features selected from the six features, and the value range of k is 1 to 6. S103: From the historical data, obtain the data of each feature subset respectively, and use the data of the feature subset as the input feature, and use the corresponding actual wear degree value as the target variable. M random forest regression models are obtained by training the input features and target variables. The random forest regression model trained with the mth feature subset Sm is recorded as the mth random forest regression model. S104: Using K test circuit breakers for comparative testing, use the trained M random forest regression models to predict the wear degree values ​​of the K test circuit breakers respectively, and evaluate the prediction performance of each random forest regression model. S105: Measure the prediction accuracy of M random forest regression models: ； Among them, RS m is the prediction deviation of the mth random forest regression model, K is the total number of test circuit breakers, i is the i-th test circuit breaker, y ture，i is the actual wear value of the i-th test circuit breaker, y m，i is the wear degree value predicted by the mth random forest regression model for the i-th test circuit breaker, S106: Select the feature subset with the smallest prediction deviation value as the optimal feature subset S, and use the random forest regression model obtained by training with the optimal feature subset S as the input feature as the target module.

5. The circuit breaker mechanical wear prediction system according to claim 4, characterized in that: The calculation method of the wear degree value DOW of the circuit breaker is: , Among them, Vfir is the current volume of the internal mechanical components of the circuit breaker, Vorig is the initial volume of the internal components of the circuit breaker, and Vfir and Vorig are obtained through industrial CT and non-contact three-dimensional imaging calculations. The internal mechanical components include contact assemblies, closing springs, connecting rod mechanisms, transmission slides and opening springs, which are key components that are prone to structural wear during long-term mechanical operation.

6. The circuit breaker mechanical wear prediction system according to claim 4, characterized in that: The feature fusion unit implements the following operation steps: S201: Standardize the values ​​of each feature in the optimal feature subset, and express the value of the jth standardized feature in the optimal feature subset as : ， Among them, f j is the value of the jth feature in the optimal feature subset in the data set, and the unit depends on the feature type; μ j is the mean of the jth feature in the data set, and its unit is the same as f j Consistent, σ j is the standard deviation of the jth feature in the data set, and the unit is the same as f j consistent; is dimensionless; S202: Using the data set as training input, the fuzzy C-means clustering algorithm is used to analyze the standardized features and obtain the feature membership matrix U=[u ij ], based on which the fusion weight of each feature in the optimal feature subset is calculated, and The fusion weight is expressed as w j : ； Among them, Q refers to the total number of samples used for fuzzy cluster analysis in the data set, in units of pieces; u qj It represents the fuzzy membership of the qth sample in the dataset on the jth feature in the optimal feature subset, which is calculated by the fuzzy C-means clustering algorithm. Its value range is [0,1] and the unit is dimensionless. w j It reflects the matching degree between the feature and the cluster center in different samples. The larger the value, the higher the consistency or discrimination ability of the feature in most samples. Therefore, its weight in the final fusion feature vector should also be increased accordingly. j The unit of is dimensionless; S203: Based on the fusion weight w obtained in step S202 j , each standardized eigenvalue f in the optimal feature subset S j′ According to the corresponding fusion weight w j Perform weighted summation to obtain the fused eigenvalue F: ； Where J is the total number of features in the optimal feature subset, F is the fused feature value, and F is dimensionless, serving as the input variable of the control and evaluation modules; S204: Obtain the fusion feature values ​​corresponding to all samples from the data set, and express the fusion feature of the Lth sample as F L , the fusion feature set X is expressed as the fusion feature of all samples in the dataset, X={F1,F2…F Q }, and the actual mechanical wear of all circuit breakers in the sample in the data set is taken as the wear set Y, Y={R1,R2…R Q }, where R L is the actual mechanical wear of the circuit breaker in the Lth sample.

Images