Electrical control detection system

Abstract

The invention discloses an electrical control detection system, and relates to the technical field of electrical control, a data acquisition module acquires current, voltage, temperature, control signal state and other electrical data, and a data processing module performs feature extraction and classification on the data by using a spatial domain filtering technology and a support vector machine, so that the operation state of the system can be accurately judged; and the detection efficiency and accuracy are improved. The fault diagnosis module adopts an isolation forest algorithm to carry out anomaly detection, can more intelligently identify the abnormal state of the system, carries out predictive maintenance analysis through the electrical maintenance module combining a convolutional neural network and a long-short-term memory network, can predict the maintenance time and fault causes of electrical equipment, and provides an optimal maintenance strategy. And the maintenance delay and cost are reduced. The communication module guarantees data communication, and the man-machine interaction module is convenient for checking a detection result; according to the invention, many problems of a traditional detection means are effectively solved, and the comprehensiveness, systematicness and intelligent degree of detection and maintenance of the electrical control system are improved.

Description

Technical Field

[0001] The present invention relates to the field of electrical control technology, and more particularly to an electrical control detection system. Background Art

[0002] At present, electrical control systems play a vital role in modern industrial production and the operation of various electrical equipment. Its stability, reliability and accuracy directly affect the safety and efficiency of the entire production process. However, traditional electrical control detection often has many limitations. For example, the detection means are single, and most of them can only perform simple monitoring of electrical parameters, making it difficult to conduct a comprehensive and in-depth evaluation of complex control logic and overall system performance; the detection process lacks systematicity and automation, and requires a lot of manual intervention, which not only leads to low detection efficiency, but also easily leads to inaccurate detection results due to human factors; in addition, the ability to diagnose and locate faults is insufficient. When the system is abnormal, it is difficult to quickly and accurately determine the source of the fault, thereby delaying maintenance time and increasing production costs.

[0003] Therefore, how to achieve accurate positioning while improving detection efficiency is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0004] In view of this, the present invention provides an electrical control detection system to solve the problems existing in the background technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] An electrical control detection system includes: a data acquisition module, a data processing module, a control logic detection module, a fault diagnosis module, and a communication module;

[0007] The data acquisition module is used to collect electrical data in the electrical control system, and the electrical data includes current, voltage, temperature and status information of control signals;

[0008] The data processing module receives the electrical data collected by the data collection module for processing;

[0009] The fault diagnosis module determines the fault type and the location of the fault according to the abnormal data characteristics provided by the data processing module and the pre-established fault diagnosis model;

[0010] The communication module is responsible for data communication between various modules.

[0011] Optionally, the data acquisition module is connected to various monitoring points of the electrical control system to ensure that the required electrical parameters and control signals can be accurately collected; the data acquisition module includes a sensor module and a signal conditioning circuit; the sensor in the sensor module is selected according to different monitoring parameters; the analog signal collected by the sensor module is amplified, filtered and analog-to-digital converted by the signal conditioning circuit, and then transmitted to the data processing module.

[0012] Optionally, the data processing module includes a filtering module and a data analysis module;

[0013] The filtering module uses spatial domain filtering technology to extract features from electrical data, maximizes the variance of one type of signal according to the label, and minimizes the variance of another type of signal, finds the best spatial projection direction to perform differential projection on the input electrical data, and calculates the normalized variance as a feature vector;

[0014] The data analysis module uses a support vector machine to perform binary classification on the feature vector to obtain a basic classification model; the basic classification model is used to classify the current operating status of the electrical control system to determine whether the system is in a normal operating state. If the system is operating normally, data collection and analysis will continue; if there is an abnormality in the system, the currently collected feature vector will be transmitted to the fault diagnosis module as an abnormal data feature.

[0015] Optionally, the specific process of extracting features from electrical data using the spatial domain filtering technology is as follows:

[0016] First, the normalized covariance matrix of the electrical data is calculated, where R1 and R2 correspond to the covariance matrices of the normal operating state and the fault state respectively; then the mixed space covariance matrix R is calculated:

[0017] Perform eigenvalue decomposition on the mixed space covariance matrix R: R = UλU T , where U is the eigenvector matrix of matrix λ, and λ is the diagonal matrix of the corresponding eigenvalues; then the eigenvalues ​​are arranged in descending order to obtain the whitened matrix P:

[0018] Transform R1 and R2 to obtain: S1 = PR1P T and S2 = PR2P T ; Then do principal component decomposition on S1 and S2 to get: and Prove that the eigenvector matrix of matrix S1 is equal to the eigenvector matrix of matrix S2, that is, B1 = B2 = B;

[0019] At the same time, the sum of the diagonal matrices λ1 and λ2 of the two eigenvalues ​​is the unit matrix, that is, the sum of the eigenvalues ​​of the two types of matrices is always 1, then the eigenvector corresponding to the maximum eigenvalue of S1 makes S2 have the minimum eigenvalue, and vice versa; the transformation of the whitening matrix to the eigenvector corresponding to the maximum eigenvalue in λ1 and λ2 is optimal for separating the variances in the two signal matrices, then the optimal spatial projection matrix W is: W = B T P;

[0020] The original signal X is spatially filtered to obtain a filtered signal Z=WX, and then the normalized variance of the filtered signal is calculated as the feature vector A.

[0021] Optionally, the process of using a support vector machine to perform binary classification on the extracted feature vector is:

[0022] Construct the optimal classification surface objective function; suppose the equation of the classification surface is x·ω+b=0, then the sample set (x i ,y i ), i = 1,…,n; x∈R d ,y i ∈{+1,-1} satisfies: y i (ω·x i +b)-1≥0; satisfy the above formula and make The smallest classification line is called the optimal classification line. In the case of linear inseparability, a relaxation term ξ is added i ≥0, that is, y i (ω·x i +b)-1-ξ i ≥0; thus minimizing the objective function:

[0023]

[0024] Among them, C is the error penalty factor, that is, the generalized optimal classification surface is obtained by comprehensively considering the minimum misclassified samples and the maximum classification interval;

[0025] The electrical data set after filtering is used as the training set to extract features and train the basic classification model to obtain the optimal classification surface parameters. A copy is extracted as the test set to verify the classification accuracy of the basic classification model.

[0026] Optionally, the fault diagnosis module includes an abnormality detection module;

[0027] The anomaly detection not only identifies anomalies based on preset thresholds, but also uses a machine learning algorithm to automatically learn and identify the normal and abnormal states of the electrical control system, and identify state deviations. The anomaly detection specifically uses an isolation forest algorithm to automatically learn and identify the normal and abnormal states of the electrical control system. The isolation forest algorithm isolates observations by randomly selecting features and randomly selecting cut values ​​of the features.

[0028] Optionally, the isolation forest algorithm is expressed as follows:

[0029] Let G be the data set, g be a sample in G, each tree in the isolation forest algorithm is an isolation tree, and for each tree, the anomaly score S(g, n)

[0030]

[0031] Among them, E(h(g)) is the expected value of the path length required to isolate sample g in the tree, n is the number of samples, and E(h(g)) also represents the average path length of point g on the random tree in the random forest, which is used to measure the difficulty of isolating point g. The shorter the path length, the easier it is to isolate g, that is, the closer it is to an outlier. c(n) is a normalization factor of the path length, which is used to convert E(h(g)) into an anomaly score between 0 and 1. The closer the anomaly score s(g, n) is to 1, the closer the sample g is to an outlier.

[0032] Optionally, the fault diagnosis module includes an electrical maintenance module. The electrical maintenance module is based on a prediction model, which not only predicts when the electrical equipment needs maintenance, but also provides the cause of the fault and the best maintenance strategy. According to the actual operation of the electrical control system and historical maintenance records, a customized maintenance plan is proposed, and a deep learning model combining a convolutional neural network and a long short-term memory network is used to perform predictive maintenance analysis, as follows:

[0033] Convolutional neural network part: contains multiple convolutional layers, each layer has a specific number of convolutional kernels, which are used to extract features at different levels. The convolutional layer uses electrical equipment sensor data as input, including torque, temperature, and vibration. Each convolutional layer uses convolutional kernels of different sizes to capture features of different scales. Each convolutional layer is followed by a pooling layer to reduce the spatial dimension of the features and reduce the amount of calculation. The activation function of the convolutional layer uses the ReLU function;

[0034] Long short-term memory network LSTM layer: The LSTM layer includes multiple LSTM units, each of which has a specific number of neurons. The LSTM layer is a multi-layer structure to enhance the learning ability of the model. The activation function of the LSTM layer uses the tanh function.

[0035] Optionally, the prediction model parameters and calculations are as follows:

[0036] Convolutional layer: For each convolutional layer, the calculation is expressed as: F i =ReLU(W i *X+b i ), where * represents the convolution operation, W i and b i are the weight and bias of the convolution kernel, which are adjusted according to the characteristics of different sensor data to optimize feature extraction. i is a feature map, which represents the key features extracted from the raw electrical equipment data;

[0037] LSTM layer: The calculation of the LSTM unit in the LSTM layer includes the forget gate, input gate, unit state update and output gate. The calculation of each part involves the weight matrix, bias term and activation function. The output H of the LSTM unit t Expressed as: H t =O t ⊙tanh(C t ), where O t is the activation value of the output gate, C t is the unit state;

[0038] The final output consists of one or more fully connected layers, which convert the output of the LSTM unit into the final prediction result, which includes the probability that the electrical control system needs maintenance and the possibility of the fault type.

[0039] Optionally, a human-computer interaction module is also included to display the test results, system operating status, fault type and maintenance information.

[0040] It can be known from the above technical solutions that, compared with the prior art, the present invention discloses an electrical control detection system, which collects electrical data such as current, voltage, temperature and control signal status through the data acquisition module, and the data processing module uses spatial filtering technology and support vector machine to extract and classify the data, which can accurately judge the system operation status and improve the detection efficiency and accuracy. The fault diagnosis module uses the isolation forest algorithm for anomaly detection, which can more intelligently identify the abnormal state of the system. It also performs predictive maintenance analysis through the electrical maintenance module combining convolutional neural network and long short-term memory network, which can predict the maintenance time of electrical equipment, the cause of the fault and provide the best maintenance strategy to reduce maintenance delays and costs. The communication module ensures data communication, and the human-computer interaction module facilitates viewing of information such as detection results; the present invention as a whole effectively solves many problems of traditional detection methods, and improves the comprehensiveness, systematicness and intelligence of electrical control system detection and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0042] Figure 1 This is a schematic diagram of the system structure provided by the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] The embodiment of the present invention discloses an electrical control detection system, such as Figure 1 As shown, it includes: a data acquisition module, a data processing module, a fault diagnosis module, and a communication module;

[0045] A data acquisition module is used to collect electrical data in the electrical control system, including current, voltage, temperature and status information of control signals;

[0046] A data processing module receives and processes the electrical data collected by the data acquisition module;

[0047] The fault diagnosis module determines the fault type and location of the fault based on the abnormal data features provided by the data processing module and the pre-established fault diagnosis model;

[0048] The communication module is responsible for data communication between modules.

[0049] In a specific embodiment, the data acquisition module is connected to each monitoring point of the electrical control system to ensure that the required electrical parameters and control signals can be accurately collected; the data acquisition module includes a sensor module and a signal conditioning circuit; the sensor in the sensor module is selected according to different monitoring parameters; the analog signal collected by the sensor module is amplified, filtered and converted into analog to digital by the signal conditioning circuit, and then transmitted to the data processing module. The specific design of the signal conditioning circuit is a prior art and will not be described in detail here.

[0050] In a specific embodiment, the data processing module includes a filtering module and a data analysis module;

[0051] The filtering module uses spatial filtering technology to extract features from electrical data, maximizes the variance of one type of signal according to the label, and minimizes the variance of another type of signal, finds the best spatial projection direction to perform differential projection on the input electrical data, and calculates the normalized variance as a feature vector, which is input into the basic classification model;

[0052] The specific process of extracting features from electrical data using spatial domain filtering technology is as follows:

[0053] Assuming that m sets of data are collected, each set of data contains current I, voltage V, temperature T and control signal C, then a data matrix X can be constructed with a dimension of m×4, where each row represents a set of electrical data collected at the same time.

[0054] Define labels according to different states of electrical data. The normal operation and fault state of electrical equipment are divided into two types of labels, marked as y i ∈{+1, -1}(i=1, 2, ..., m), where +1 indicates a normal operating state and -1 indicates a fault state.

[0055] The data is divided into two categories according to the labels, and the covariance matrix of each category of data is calculated separately. Let the data matrix of normal operation state be X1 (the dimension is m1×4, m1 is the number of data groups of normal operation state), and the data matrix of fault state be X2 (the dimension is m2×4, m2 is the number of data groups of fault state).

[0056] Calculate the covariance matrices R1 and R2 of X1 and X2

[0057] The covariance matrix is ​​calculated as (where n is the number of data sets, is the mean vector of the data). For R1, n=m1, X=X1; for R2, n=m2, X=X2.

[0058] Among them, R1 and R2 correspond to the covariance matrices of normal operation and fault state respectively; then calculate the mixed space covariance matrix R:

[0059] Perform eigenvalue decomposition on the mixed space covariance matrix R: R = UλU T , where U is the eigenvector matrix of matrix λ, and λ is the diagonal matrix of the corresponding eigenvalues; then the eigenvalues ​​are arranged in descending order to obtain the whitened matrix P:

[0060] Transform R1 and R2 to obtain: S1 = PR1P T and S2 = PR2P T ; Then do principal component decomposition on S1 and S2 to get: and Prove that the eigenvector matrix of matrix S1 is equal to the eigenvector matrix of matrix S2, that is, B1 = B2 = B;

[0061] At the same time, the sum of the two eigenvalue diagonal matrices λ1 and λ2 is the unit matrix, that is, the sum of the eigenvalues ​​of the two types of matrices is always 1. Then the eigenvector corresponding to the maximum eigenvalue of S1 makes S2 have the minimum eigenvalue, and vice versa; the minimum eigenvalue, and vice versa. i The eigenvalues ​​in are arranged in descending order, and the corresponding eigenvalues ​​in λ2 are arranged in ascending order. Based on this, it can be inferred that λ1 and λ2 have the following form:

[0062] λ1=diag(I1 σ M 0),λ2=diag(0 σ M I2)

[0063] The transformation of the whitening matrix to the eigenvector corresponding to the largest eigenvalue in λ1 and λ2 is optimal for separating the variances in the two signal matrices, so the optimal spatial projection matrix W is: W = B T P;

[0064] The original signal X is spatially filtered to obtain the filtered signal Z = WX, and then the normalized variance of the filtered signal is calculated as the feature vector A. Input to the base classification model.

[0065] The data analysis module uses a support vector machine to perform binary classification on the feature vector to obtain a basic classification model; the basic classification model is used to classify the current operating status of the electrical control system to determine whether the system is in a normal operating state. If the system is operating normally, data collection and analysis will continue; if there is an abnormality in the system, the currently collected feature vector will be transmitted to the fault diagnosis module as an abnormal data feature.

[0066] The process of using support vector machine to classify the extracted feature vector into two categories is as follows:

[0067] SVM maps the input feature vector to a high-dimensional feature space through appropriate nonlinear mapping, so that the data (belonging to two categories) can always be separated by a hyperplane, that is, the optimal classification surface. The so-called optimal classification surface requires that the classification surface can not only correctly separate the two types of data, but also maximize the classification interval.

[0068] Construct the optimal classification surface objective function; suppose the equation of the classification surface is x·ω+b=0, then the sample set (x i ,y i ), i = 1,…,n; x∈R d ,y i ∈{+1,-1} satisfies:

[0069] y i (ω·x i +b)-1≥0;

[0070] Satisfy the above formula and make The smallest classification line is called the optimal classification line. After solving the optimal problem using the Lagrange multiplier method, the optimal decision function is expressed as:

[0071]

[0072] Where N is the number of support vectors, a i is the Lagrange multiplier.

[0073] In the case of linear inseparability, add a relaxation term ξ i ≥0, that is:

[0074] y i (ω·x i +b)-1-ξ i ≥0;

[0075] So as to minimize the objective function:

[0076]

[0077] Among them, C is the error penalty factor, that is, the generalized optimal classification surface is obtained by comprehensively considering the minimum misclassified samples and the maximum classification interval;

[0078] The electrical data set after filtering is used as the training set to extract features and train the basic classification model to obtain the optimal classification surface parameters. A copy is extracted as the test set to verify the classification accuracy of the basic classification model.

[0079] In a specific embodiment, the fault diagnosis module includes an anomaly detection module;

[0080] Anomaly detection not only identifies anomalies based on preset thresholds, but also uses machine learning algorithms to automatically learn and identify the normal and abnormal states of the electrical control system, and identify state deviations. Anomaly detection specifically uses the isolation forest algorithm to automatically learn and identify the normal and abnormal states of the electrical control system. The isolation forest algorithm isolates observations by randomly selecting features and randomly selecting the segmentation value of the feature. The isolation forest algorithm is expressed as follows:

[0081] Let G be the data set, g be a sample in G, each tree in the isolation forest algorithm is an isolation tree, and for each tree, the anomaly score S(g, n)

[0082]

[0083] Among them, E(h(g)) is the expected value of the path length required to isolate sample g in the tree, n is the number of samples, and E(h(g)) also represents the average path length of point g on the random tree in the random forest, which is used to measure the difficulty of isolating point g. The shorter the path length, the easier it is to isolate g, that is, the closer it is to an outlier. c(n) is a normalization factor of the path length, which is used to convert E(h(g)) into an anomaly score between 0 and 1. The closer the anomaly score s(g, n) is to 1, the closer the sample g is to an outlier.

[0084] Sample quantity n: In the present invention, the sample quantity is determined according to the amount of monitoring data of the electrical control system. If each electrical control system generates hundreds of records every day, daily or weekly data can be selected as a sample set.

[0085] Number of trees: The number of trees should be large enough to ensure the stability and reliability of the results. In the scenario of electrical control system detection, it can be set between 100 and 500 trees to obtain better anomaly detection performance.

[0086] Subsample size: The subsample size used by each tree can be determined based on the characteristics of the electrical equipment and the diversity of the data. The size can be set to a small proportion of the training data (256 or 512) to ensure diversity among the trees.

[0087] Path length E(h(g)): Path length is the number of steps to reach a leaf node during the isolation process. In the scenario of electrical equipment monitoring data, path length reflects the degree of deviation of a data point from the normal operating mode.

[0088] Normalization factor c(n): The normalization factor is used to convert the path length into an anomaly score between 0 and 1. This factor depends on the number of samples and the characteristics of the data and is adjusted according to the specific electrical equipment monitoring scenario.

[0089] In a specific embodiment, the fault diagnosis module includes an electrical maintenance module. The electrical maintenance module is based on a prediction model, which not only predicts when the electrical equipment needs maintenance, but also provides the cause of the fault and the best maintenance strategy. According to the actual operation of the electrical control system and historical maintenance records, a customized maintenance plan is proposed, and a deep learning model of a convolutional neural network and a long short-term memory network is combined to perform predictive maintenance analysis, as follows:

[0090] Convolutional neural network part: contains multiple convolutional layers, each layer has a specific number of convolutional kernels, which are used to extract features at different levels. The convolutional layer uses electrical equipment sensor data as input, including torque, temperature, and vibration. Each convolutional layer uses convolutional kernels of different sizes to capture features of different scales. Each convolutional layer is followed by a pooling layer to reduce the spatial dimension of the features and reduce the amount of calculation. The activation function of the convolutional layer uses the ReLU function;

[0091] Long short-term memory network LSTM layer: The LSTM layer includes multiple LSTM units, each of which has a specific number of neurons. The LSTM layer is a multi-layer structure to enhance the learning ability of the model. The activation function of the LSTM layer uses the tanh function.

[0092] In a specific embodiment, the prediction model parameters and calculations are as follows:

[0093] Convolutional layer: For each convolutional layer, the calculation is expressed as: F i =ReLU(W i *X+b i ), where * represents the convolution operation, W i and b i are the weight and bias of the convolution kernel, which are adjusted according to the characteristics of different sensor data to optimize feature extraction. i is a feature map, which represents the key features extracted from the raw electrical equipment data;

[0094] LSTM layer: The calculation of the LSTM unit in the LSTM layer includes the forget gate, input gate, unit state update and output gate. The calculation of each part involves the weight matrix, bias term and activation function. The output H of the LSTM unit t Expressed as: H t =O t ⊙tanh(C t ), where O t is the activation value of the output gate, C t is the unit state;

[0095] The final output consists of one or more fully connected layers, which convert the output of the LSTM unit into the final prediction result, which includes the probability that the electrical control system needs maintenance and the possibility of the fault type.

[0096] CNN part: the number and size of convolutional layers: according to the complexity and feature dimension of the electrical equipment data, multiple convolutional layers can be set, each with a different number of convolution kernels. For electrical equipment containing multiple sensor data, more convolutional layers are needed to effectively extract features.

[0097] Convolution kernel size: Select an appropriate convolution kernel size, such as 3x3 or 5x5, based on the characteristics of the electrical equipment data to effectively capture local features in the data.

[0098] Pooling layer: Pooling layer is added after convolutional layer to reduce feature dimension and computational complexity while keeping key information.

[0099] LSTM part:

[0100] Number of LSTM units: Select an appropriate number of LSTM units based on the length and complexity of the time series data. For long time series and complex data patterns, more units are required to capture long-term dependencies.

[0101] Number of LSTM layers: The multi-layer LSTM structure enhances the model's ability to learn time series data. The number of layers depends on the time dependency and complexity of the data.

[0102] Fully connected layer and output layer: The number of neurons in the fully connected layer is determined according to the target to be predicted (maintenance requirements, potential fault types).

[0103] Output layer: Designed for specific prediction tasks, such as binary classification, multi-classification, or regression, depending on the specific type of maintenance requirements.

[0104] In addition, the module has self-learning capabilities, and the accuracy of its analysis and prediction will continue to improve over time and with the accumulation of data.

[0105] In a specific embodiment, a human-computer interaction module is also included to display the detection results, system operation status, fault type and maintenance information. Specific functions include but are not limited to:

[0106] Data visualization: Displays real-time data from data acquisition modules, such as voltage, current, temperature, etc., in the form of charts, dashboards, and trend lines.

[0107] Abnormal alerts and notifications: When the fault diagnosis module detects an abnormality or predicts maintenance needs, the human-machine interaction module will display an alert and provide detailed fault information and recommended actions.

[0108] Maintenance schedule planning: Based on predictive maintenance analysis, the human-computer interaction module provides a maintenance schedule planning tool to help users arrange maintenance activities according to system recommendations.

[0109] Historical Data Analysis: Provides access to historical monitoring data and maintenance records to facilitate long-term performance analysis and trend identification.

[0110] Interactive control: allows users to directly adjust electrical equipment parameters or apply maintenance measures recommended by the system through the human-computer interaction module.

[0111] The human-machine interaction module is designed to enhance the user experience, enabling electrical equipment operators and maintenance engineers to easily monitor the status of electrical equipment, quickly respond to system alarms, and effectively plan maintenance activities, thereby improving the efficiency and reliability of electrical equipment operation.

[0112] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between 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, and the relevant parts can be referred to the method part.

[0113] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An electrical control detection system, characterized in that: include: Data acquisition module, data processing module, control logic detection module, fault diagnosis module, communication module; The data acquisition module is used to collect electrical data in the electrical control system, and the electrical data includes current, voltage, temperature and status information of control signals; The data processing module receives the electrical data collected by the data collection module for processing; The fault diagnosis module determines the fault type and the location of the fault according to the abnormal data characteristics provided by the data processing module and the pre-established fault diagnosis model; The communication module is responsible for data communication between various modules.

2. An electrical control detection system according to claim 1, characterized in that: The data acquisition module is connected to various monitoring points of the electrical control system to ensure that the required electrical parameters and control signals can be accurately collected; the data acquisition module includes a sensor module and a signal conditioning circuit; the sensor in the sensor module is selected according to different monitoring parameters; the analog signal collected by the sensor module is amplified, filtered and analog-to-digital converted by the signal conditioning circuit, and then transmitted to the data processing module.

3. An electrical control detection system according to claim 1, characterized in that: The data processing module includes a filtering module and a data analysis module; The filtering module uses spatial domain filtering technology to extract features from electrical data, maximizes the variance of one type of signal according to the label, and minimizes the variance of another type of signal, finds the best spatial projection direction to perform differential projection on the input electrical data, and calculates the normalized variance as a feature vector; The data analysis module uses a support vector machine to perform binary classification on the feature vector to obtain a basic classification model; the basic classification model is used to classify the current operating status of the electrical control system to determine whether the system is in a normal operating state. If the system is operating normally, data collection and analysis will continue; if there is an abnormality in the system, the currently collected feature vector will be transmitted to the fault diagnosis module as an abnormal data feature.

4. An electrical control detection system according to claim 3, characterized in that: The specific process of extracting features from electrical data using spatial domain filtering technology is as follows: First, the normalized covariance matrix of the electrical data is calculated, where R1 and R2 correspond to the covariance matrices of the normal operating state and the fault state respectively; then the mixed space covariance matrix R is calculated: Perform eigenvalue decomposition on the mixed space covariance matrix R: R = UλU T , where U is the eigenvector matrix of matrix λ, and λ is the diagonal matrix of the corresponding eigenvalues; then the eigenvalues ​​are arranged in descending order to obtain the whitened matrix P: Transform R1 and R2 to obtain: S1 = PR1P T and S2 = PR2P T ; Then do principal component decomposition on S1 and S2 to get: and Prove that the eigenvector matrix of matrix S1 is equal to the eigenvector matrix of matrix S2, that is, B1 = B2 = B; At the same time, the sum of the diagonal matrices λ1 and λ2 of the two eigenvalues ​​is the unit matrix, that is, the sum of the eigenvalues ​​of the two types of matrices is always 1, then the eigenvector corresponding to the maximum eigenvalue of S1 makes S2 have the minimum eigenvalue, and vice versa; the transformation of the whitening matrix to the eigenvector corresponding to the maximum eigenvalue in λ1 and λ2 is optimal for separating the variances in the two signal matrices, then the optimal spatial projection matrix W is: W = B T P; The original signal X is spatially filtered to obtain a filtered signal Z=WX, and then the normalized variance of the filtered signal is calculated as the feature vector A.

5. An electrical control detection system according to claim 3, characterized in that: The process of using support vector machine to classify the extracted feature vector into two categories is as follows: Construct the optimal classification surface objective function; suppose the equation of the classification surface is x·ω+b=0, then the sample set (x i ,y i ), i = 1,…,n; x∈R d ,y i ∈{+1,-1} satisfies: y i (ω·x i +b)-1≥0; satisfy the above formula and make The smallest classification line is called the optimal classification line. In the case of linear inseparability, a relaxation term ξ is added i ≥0, that is, y i (ω·x i +b)-1-ξ i ≥0; thus minimizing the objective function: Among them, C is the error penalty factor, that is, the generalized optimal classification surface is obtained by comprehensively considering the minimum misclassified samples and the maximum classification interval; The electrical data set after filtering is used as the training set to extract features and train the basic classification model to obtain the optimal classification surface parameters. A copy is extracted as the test set to verify the classification accuracy of the basic classification model.

6. An electrical control detection system according to claim 1, characterized in that: The fault diagnosis module includes an abnormality detection module; The anomaly detection not only identifies anomalies based on preset thresholds, but also uses a machine learning algorithm to automatically learn and identify the normal and abnormal states of the electrical control system, and identify state deviations. The anomaly detection specifically uses an isolation forest algorithm to automatically learn and identify the normal and abnormal states of the electrical control system. The isolation forest algorithm isolates observations by randomly selecting features and randomly selecting cut values ​​of the features.

7. An electrical control detection system according to claim 6, characterized in that: The isolation forest algorithm is expressed as follows: Let G be the data set, g be a sample in G, each tree in the isolation forest algorithm is an isolation tree, and for each tree, the anomaly score S(g,n) Among them, E(h(g)) is the expected value of the path length required to isolate sample g in the tree, n is the number of samples, and E(h(g)) also represents the average path length of point g on the random tree in the random forest, which is used to measure the difficulty of isolating point g. The shorter the path length, the easier it is to isolate g, that is, the closer it is to an outlier. c(n) is a normalization factor of the path length, which is used to convert E(h(g)) into an anomaly score between 0 and 1. The closer the anomaly score s(g,n) is to 1, the closer the sample g is to an outlier.

8. An electrical control detection system according to claim 1, characterized in that: The fault diagnosis module includes an electrical maintenance module. The electrical maintenance module is based on a prediction model. It not only predicts when electrical equipment needs maintenance, but also provides the cause of the fault and the best maintenance strategy. According to the actual operation of the electrical control system and historical maintenance records, a customized maintenance plan is proposed. The deep learning model of the convolutional neural network and the long short-term memory network is combined to perform predictive maintenance analysis, as follows: Convolutional neural network part: contains multiple convolutional layers, each layer has a specific number of convolutional kernels, which are used to extract features at different levels. The convolutional layer uses electrical equipment sensor data as input, including torque, temperature, and vibration. Each convolutional layer uses convolutional kernels of different sizes to capture features of different scales. Each convolutional layer is followed by a pooling layer to reduce the spatial dimension of the features and reduce the amount of calculation. The activation function of the convolutional layer uses the ReLU function; Long short-term memory network LSTM layer: The LSTM layer includes multiple LSTM units, each of which has a specific number of neurons. The LSTM layer is a multi-layer structure to enhance the learning ability of the model. The activation function of the LSTM layer uses the tanh function.

9. An electrical control detection system according to claim 8, characterized in that: The prediction model parameters and calculations are as follows: Convolutional layer: For each convolutional layer, the calculation is expressed as: F i =ReLU(W i *X+b i ), where * represents the convolution operation, W i and b i are the weight and bias of the convolution kernel, which are adjusted according to the characteristics of different sensor data to optimize feature extraction. i is a feature map, which represents the key features extracted from the raw electrical equipment data; LSTM layer: The calculation of the LSTM unit in the LSTM layer includes the forget gate, input gate, unit state update and output gate. The calculation of each part involves the weight matrix, bias term and activation function. The output H of the LSTM unit t Expressed as: H t =O t ⊙tanh(C t ), where O t is the activation value of the output gate, C t is the unit state; The final output consists of one or more fully connected layers, which convert the output of the LSTM unit into the final prediction result, which includes the probability that the electrical control system needs maintenance and the possibility of the fault type.

10. An electrical control detection system according to claim 1, characterized in that: It also includes a human-computer interaction module for displaying test results, system operating status, fault type and maintenance information.

Images