Double-layer fault diagnosis method and device based on data driving, equipment and medium
Through the data-driven double-layer fault diagnosis method, the feature extraction and decision tree model are used, combined with filtering and dimensionality reduction processing, the false alarm and efficiency problems in the fault diagnosis of lithium battery energy storage system are solved, achieving high accuracy and rapid faults position.
Patent Information
- Application Number
- CN202410929845.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-08
AI Technical Summary
Due to its strong nonlinearity and sensitivity of environmental factors, lithium battery energy storage systems lead to false alarms in fault diagnosis and low diagnostic accuracy. The existing algorithms cannot take into account both diagnostic accuracy and computing speed, resulting in poor efficiency.
Using a data-driven double-layer fault diagnosis method, it uses operational data for feature extraction, uses decision tree model and hyperplane function for fault classification and diagnosis, and combines filtering, dimensionality reduction and periodic segmentation to improve data reliability and analysis efficiency. .
It improves the accuracy and efficiency of lithium battery fault diagnosis, takes into account the diagnosis accuracy and computing speed, reduces the false alarm rate, and enhances the accuracy and interpretability of fault location.
Smart Images

Figure CN118916791A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a data-driven double-layer fault diagnosis method, device, equipment and medium. Background Art
[0002] With the rapid development of the global economy, energy shortages and environmental pollution are becoming increasingly prominent. Gradually changing the energy consumption structure and vigorously developing renewable energy have gradually become a global consensus. Although renewable energy power generation has obvious advantages and broad development prospects, renewable energy grid-connected power generation faces many problems. Efficient and intelligent energy storage systems provide an effective means for the transformation and upgrading of traditional power systems and for solving many problems caused by the access of high-penetration renewable energy to the power grid. Among them, new energy storage systems represented by lithium-ion batteries are the main energy storage systems currently used. Lithium battery energy storage systems are composed of many battery cells and are complex nonlinear systems with strong nonlinearity, sensitivity to environmental factors such as temperature, performance degradation over time, and sudden failures.
[0003] According to statistics, there have been more than 54 energy storage fire accidents worldwide so far, with more than 17 in 2020 alone. From the perspective of accident cause classification, accidents caused by battery cell failures account for as much as 64%, far higher than failures caused by other reasons. Lithium batteries are strongly nonlinear systems, and as time changes, the internal characteristics of the battery will also change accordingly. It is very difficult to establish an accurate internal model, resulting in false alarms in fault diagnosis and poor diagnostic accuracy. At the same time, existing algorithms cannot take into account the relationship between diagnostic accuracy and computing speed, and the diagnostic efficiency does not meet the requirements. Summary of the invention
[0004] In view of this, the present invention provides a data-driven two-layer fault diagnosis method, device, equipment and medium to solve the problem that the internal characteristics of the battery will change accordingly with the change of time, and it is very difficult to establish an accurate internal model, resulting in false alarm problems in fault diagnosis and poor diagnostic accuracy. At the same time, the existing algorithm cannot take into account the relationship between diagnostic accuracy and computing speed, and the diagnostic efficiency does not meet the requirements.
[0005] In the first aspect, the present invention provides a data-driven two-layer fault diagnosis method, the method comprising: obtaining operating data of a device to be diagnosed, performing feature extraction on the operating data, and obtaining a feature vector; obtaining fault classification data based on the feature vector and a preset fault analysis classifier; the fault analysis classifier is obtained by training based on negative sample data with fault markers and positive sample data with normal markers; constructing a hyperplane function based on the fault classification data, and obtaining a fault diagnosis result of the device to be diagnosed based on the hyperplane function.
[0006] The data-driven two-layer fault diagnosis method provided in this embodiment first extracts features from the operating data to obtain feature vectors, and converts the original data into feature vectors, which can remove noise and redundant information, thereby simplifying the data set and making data processing and analysis more efficient. In machine learning and data analysis, using feature vectors as input can help improve the performance of the model. By learning and optimizing feature vectors, the model can better capture the patterns and relationships of the data and improve the accuracy of prediction and classification. Secondly, the fault classification data is obtained by using feature vectors and preset fault analysis classifiers. The decision tree model is a powerful machine learning algorithm as a fault analysis classifier, which can be classified according to feature vectors. Compared with traditional fault detection methods, the fault classification method based on feature vectors and decision tree models has higher accuracy. Finally, by constructing a hyperplane function based on fault classification data and calculating the optimal solution of the hyperplane function, the fault diagnosis result of the device to be diagnosed is obtained. The hyperplane function can determine the type of fault according to the input feature vector, which helps to quickly locate the cause of the fault, further improves the efficiency of fault diagnosis, and takes into account the relationship between the diagnosis accuracy and the operation speed.
[0007] In an optional implementation, feature extraction is performed on the operating data, including: filtering the operating data to obtain filtered data; performing dimensionality reduction processing on the filtered data to obtain reduced dimensionality data; and performing feature extraction on the reduced dimensionality data to obtain a feature vector.
[0008] The data-driven double-layer fault diagnosis method provided in this embodiment solves the problem that the lithium battery operation data will inevitably be affected by interference, sensor failure and other factors during the collection process, generating abnormal fault data and affecting the accuracy of fault diagnosis by filtering and reducing the operating data. The present invention uses discrete cosine transform and inverse transform to process the data set, removes abnormal data, and increases data reliability. Through dimensionality reduction, high-dimensional data can be converted into low-dimensional data, making the data easier to process and understand. This helps to simplify the data analysis process and improve efficiency.
[0009] In an optional embodiment, the filtered data is subjected to dimensionality reduction processing to obtain reduced-dimensional data, including: constructing a variable matrix based on the filtered data, and calculating the covariance matrix of the variable matrix; extracting the eigenvalues and the first eigenvector of the covariance matrix; performing principal component analysis based on the eigenvalues and the first eigenvector to obtain principal component values and corresponding contribution rates to obtain reduced-dimensional data.
[0010] In an optional embodiment, feature extraction is performed on the reduced-dimensional data, including: periodically segmenting the reduced-dimensional data based on a feature extraction method to construct a data feature set; setting labels for the data feature set to obtain a machine learning data set; sampling the machine learning data set based on a preset number of times to obtain a data sampling set; and performing set processing on multiple data sampling sets to obtain a feature vector.
[0011] The data-driven two-layer fault diagnosis method provided in this embodiment can divide the dimension reduction data into different time periods through period segmentation and feature extraction, and extract a data feature set reflecting the data characteristics from each time period. These data feature sets contain rich information, which helps to understand the characteristics and laws of the data more comprehensively. Setting labels for the data feature sets can make the machine learning data set have clear classification or identification, which is convenient for subsequent model training and learning. By corresponding the labels to the data feature sets, the performance and accuracy of the model can be better evaluated. By sampling the machine learning data set through a sampling method, multiple data sampling sets can be obtained. These data sampling sets can be used for different model training and analysis tasks, which improves the utilization and flexibility of the data. By performing set processing on multiple data sampling sets, more comprehensive and diversified feature vectors can be obtained. These feature vectors can capture more patterns and features of the data, which helps to improve the accuracy and robustness of fault diagnosis.
[0012] In an optional implementation, fault classification data is obtained based on feature vectors and preset fault analysis classifiers, including: randomly extracting a data sampling set, obtaining a Gini index based on the random extraction result; splitting based on the Gini index to generate a decision tree; and integrating the decision trees using a majority voting method to obtain a random forest. Fault classification data is obtained based on feature vectors and preset fault analysis classifiers, including: inputting target operation data into a binary classifier of a random forest for classification prediction, obtaining a classifier decision result for each data; locating the target operation data based on the classifier decision result to obtain fault data, and the fault data includes a fault data type and a fault type.
[0013] The data-driven two-layer fault diagnosis method provided in this embodiment generates a decision tree by splitting based on the Gini index. The random forest consists of multiple decision trees, and each decision tree is trained based on a randomly extracted data set. Since each decision tree only uses part of the data for training, the risk of overfitting can be reduced. At the same time, the classification accuracy can be further improved by integrating the decision trees through the majority voting method. By inputting the target operating data into the binary classifier of the random forest for classification prediction, obtaining the classifier decision results of each data, and using the binary classifier of the random forest for fault location, the fault diagnosis process can be simplified. Compared with the traditional fault detection method, the fault location method based on the classifier decision result has higher accuracy and efficiency. Fault location based on the classifier decision result can more accurately identify fault data.
[0014] In an optional embodiment, a hyperplane function is constructed based on the fault classification data, and a fault diagnosis result of the device to be diagnosed is obtained based on the hyperplane function, including: inputting the feature vector into a preset vector machine model to obtain a hyperplane function corresponding to the fault classification data; and calculating the optimal solution of the hyperplane to obtain a fault diagnosis result corresponding to the fault classification data.
[0015] The data-driven two-layer fault diagnosis method provided in this embodiment can quickly detect and identify abnormal data by classifying fault data based on a hyperplane. When abnormal data appears, the hyperplane model can quickly make a judgment, and the fault diagnosis results based on the hyperplane have good interpretability. By understanding and analyzing the hyperplane, the nature and mode of the fault can be better understood, and the credibility and transparency of the fault diagnosis results can be improved.
[0016] In an optional implementation, the hyperplane divides the short-circuit fault data, aging fault data and short-circuit fault data of the device to be diagnosed, and the expression of the hyperplane is: T *x i +b=0where x i is the training sample input, w is the weight vector, and b is the threshold.
[0017] In the data-driven two-layer fault diagnosis method provided in this embodiment, the weight vector and threshold in the hyperplane expression are adjustable. By adjusting these parameters, the performance of the model can be optimized and adapted to different data sets and tasks.
[0018] In the second aspect, the present invention provides a data-driven two-layer fault diagnosis device, which includes: a feature extraction module, which is used to obtain the operating data of the device to be diagnosed, perform feature extraction on the operating data, and obtain a feature vector; a fault analysis module, which is used to obtain fault classification data based on the feature vector and a preset fault analysis classifier; the fault analysis classifier is trained based on negative sample data with fault identification and positive sample data with normal identification; a fault diagnosis module, which is used to construct a hyperplane function based on the fault classification data, calculate the optimal solution of the hyperplane function, and obtain the fault diagnosis result of the device to be diagnosed.
[0019] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the data-driven two-layer fault diagnosis method described in the first aspect or any corresponding embodiment thereof by executing the computer instructions.
[0020] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the data-driven two-layer fault diagnosis method described in the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0022] Figure 1 is a flow chart of a data-driven dual-layer fault diagnosis method according to an embodiment of the present invention;
[0023] Figure 2 is a flow chart of another data-driven double-layer fault diagnosis method according to an embodiment of the present invention;
[0024] Figure 3 is a flow chart of a battery fault detection method according to an embodiment of the present invention;
[0025] Figure 4 is a structural block diagram of a data-driven double-layer fault diagnosis device according to an embodiment of the present invention;
[0026] Figure 5 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without creative work are within the scope of protection of the present invention.
[0028] With the rapid development of the global economy, energy shortages and environmental pollution are becoming increasingly prominent. Gradually changing the energy consumption structure and vigorously developing renewable energy have gradually become a global consensus. Although renewable energy power generation has obvious advantages and broad development prospects, renewable energy grid-connected power generation faces many problems. Efficient and intelligent energy storage systems provide an effective means for the transformation and upgrading of traditional power systems and for solving many problems caused by the access of high-penetration renewable energy to the power grid. Among them, new energy storage systems represented by lithium-ion batteries are the main energy storage systems currently used. Lithium battery energy storage systems are composed of many battery cells and are complex nonlinear systems with strong nonlinearity, sensitivity to environmental factors such as temperature, performance degradation over time, and sudden failures.
[0029] According to statistics, there have been more than 54 energy storage fire accidents worldwide so far, with more than 17 in 2020 alone. From the perspective of accident cause classification, accidents caused by battery cell failures account for as much as 64%, far higher than failures caused by other reasons. Lithium batteries are strongly nonlinear systems, and as time changes, the internal characteristics of the battery will also change accordingly. It is very difficult to establish an accurate internal model, resulting in false alarms in fault diagnosis and poor diagnostic accuracy. At the same time, existing algorithms cannot take into account the relationship between diagnostic accuracy and computing speed, and the diagnostic efficiency does not meet the requirements.
[0030] The data-driven two-layer fault diagnosis method provided in this embodiment first extracts features from the operating data to obtain feature vectors, and converts the original data into feature vectors, which can remove noise and redundant information, thereby simplifying the data set and making data processing and analysis more efficient. In machine learning and data analysis, using feature vectors as input can help improve the performance of the model. By learning and optimizing feature vectors, the model can better capture the patterns and relationships of the data and improve the accuracy of prediction and classification. Secondly, the fault classification data is obtained by using feature vectors and preset fault analysis classifiers. The decision tree model is a powerful machine learning algorithm as a fault analysis classifier, which can be classified according to feature vectors. Compared with traditional fault detection methods, the fault classification method based on feature vectors and decision tree models has higher accuracy. Finally, by constructing a hyperplane function based on fault classification data and calculating the optimal solution of the hyperplane function, the fault diagnosis result of the device to be diagnosed is obtained. The hyperplane function can determine the type of fault according to the input feature vector, which helps to quickly locate the cause of the fault, further improves the efficiency of fault diagnosis, and takes into account the relationship between the diagnosis accuracy and the operation speed.
[0031] According to an embodiment of the present invention, a data-driven two-layer fault diagnosis method embodiment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0032] In this embodiment, a data-driven two-layer fault diagnosis method is provided. Figure 1 is a flow chart of a data-driven dual-layer fault diagnosis method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:
[0033] Step S101, obtaining operation data of the device to be diagnosed, performing feature extraction on the operation data, and obtaining a feature vector.
[0034] Specifically, the operation data may include the state of charge (SOC), voltage, temperature data, etc. of the battery during operation. Feature extraction is performed on the operation data to obtain a feature vector, which is used to characterize the data characteristics of the operation data. Feature extraction of the operation data is an important step in data processing. The lithium battery operation data is preprocessed to remove abnormal data, improve data accuracy, reduce data dimensions, and increase calculation speed. The operation data is trained and learned using machine learning algorithms to obtain feature representations of the data, such as support vector machines (SVMs), neural networks, etc. The extracted features should have clear meanings and physical meanings to facilitate subsequent analysis and understanding.
[0035] Step S102, obtaining fault classification data based on the feature vector and a preset fault analysis classifier; the fault analysis classifier is obtained by training based on negative sample data with fault identification and positive sample data with normal identification.
[0036] Specifically, a fault analysis classifier is a tool for identifying and diagnosing a variety of faults. It uses an algorithm to classify faults and can classify faults online. The advantages of this classifier are that it is simple, can classify faults online, and has strong fault classification capabilities. When implementing a fault analysis classifier, first collect data samples of various faults and use these data samples to train the classifier. During the training process, the algorithm learns how to identify and distinguish different types of faults. When new fault data appears, the classifier can quickly classify the fault. During the training process, negative sample data is usually data from a faulty device or system, and the fault mark indicates that this type of data belongs to negative sample data, while positive sample data is data from a normally operating device or system, and the normal mark indicates that this type of data belongs to positive sample data. By learning from this data, the fault analysis classifier can learn to identify similar features in new data and determine whether the device or system is operating normally or whether a fault has occurred based on these features.
[0037] Step S103, constructing a hyperplane function based on the fault classification data, and obtaining a fault diagnosis result of the device to be diagnosed based on the optimal solution of the hyperplane function.
[0038] Specifically, constructing a hyperplane function based on fault classification data can further optimize the accuracy of fault diagnosis. The hyperplane function is a classifier that divides the fault state of the device to be diagnosed into two states: normal or faulty, according to the input feature vector. In order to calculate the optimal solution of the hyperplane function, some optimization algorithms need to be adopted, such as the support vector machine (SVM) algorithm. The SVM algorithm is a binary classifier that divides the input samples into two states by constructing a hyperplane with the maximum interval. The goal of this algorithm is to find a hyperplane that maximizes the interval while satisfying the constraints. When calculating the optimal solution, the data sampling set needs to be divided into a training set and a validation set. The training set is used to train the hyperplane function, while the validation set is used to evaluate the performance of the model. The cross-validation method is used to find the optimal hyperplane parameters by training and validating the model multiple times. Once the optimal solution is found, the hyperplane function can be used to diagnose the fault of the device to be diagnosed. Specifically, the various characteristic indicators of the device to be diagnosed can be used as input vectors and substituted into the hyperplane function for classification. According to the classification results, it can be judged whether the device to be diagnosed is operating normally or whether a fault has occurred.
[0039] The data-driven two-layer fault diagnosis method provided in this embodiment first extracts features from the operating data to obtain feature vectors, and converts the original data into feature vectors, which can remove noise and redundant information, thereby simplifying the data set and making data processing and analysis more efficient. In machine learning and data analysis, using feature vectors as input can help improve the performance of the model. By learning and optimizing feature vectors, the model can better capture the patterns and relationships of the data and improve the accuracy of prediction and classification. Secondly, the fault classification data is obtained by using feature vectors and preset fault analysis classifiers. The decision tree model is a powerful machine learning algorithm as a fault analysis classifier, which can be classified according to feature vectors. Compared with traditional fault detection methods, the fault classification method based on feature vectors and decision tree models has higher accuracy. Finally, by constructing a hyperplane function based on fault classification data and calculating the optimal solution of the hyperplane function, the fault diagnosis result of the device to be diagnosed is obtained. The hyperplane function can determine the type of fault according to the input feature vector, which helps to quickly locate the cause of the fault, further improves the efficiency of fault diagnosis, and takes into account the relationship between the diagnosis accuracy and the operation speed.
[0040] In this embodiment, a data-driven two-layer fault diagnosis method is provided. Figure 2 is a flow chart of a data-driven dual-layer fault diagnosis method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:
[0041] Step S201, obtaining operation data of the device to be diagnosed, performing feature extraction on the operation data, and obtaining a feature vector.
[0042] Specifically, the above step S201 includes:
[0043] Step S2011, filtering the operating data to obtain filtered data.
[0044] Specifically, the lithium battery operation data will inevitably be affected by interference, sensor failure and other factors during the collection process, resulting in abnormal fault data, which affects the accuracy of fault diagnosis. Therefore, the present invention uses discrete cosine transform and inverse transform to process the data set, remove abnormal data, and increase data reliability. The following formula is used for discrete cosine filtering:
[0045]
[0046] in:
[0047] The inverse transform is:
[0048]
[0049] Wherein, k=0, 1, 2, 3…N-1, c(k) is the kth cosine transform coefficient, u is the generalized frequency variable, N is the number of original signals, c(k) is the compensation coefficient, and x(n) is the lithium battery operating data voltage, current, temperature, etc.
[0050] Step S2012, performing dimensionality reduction processing on the filtered data to obtain dimensionality reduced data.
[0051] In some optional embodiments, step 2012 includes:
[0052] Step a1, constructing a variable matrix based on the filtered data, and calculating the covariance matrix of the variable matrix.
[0053] First, perform dimensionality reduction on the filtered data by principal component analysis. Assuming there are n samples and each sample has p variables, construct an n×p-order matrix x of the observed samples:
[0054]
[0055] Among them, X i1 is the SOC of the data at the i-th moment, X i2 is the cell terminal voltage at the i-th moment, X i3 is the battery temperature of the data at the i-th moment. np is the pth variable of the nth sample. For this solution, the project selected SOC, voltage, and temperature as input variables n, and p is a positive integer. As an example, n is 18318 and p is 3.
[0056] The n×p-order matrix is standardized, where: Xnp is the p-th variable of the n-th sample after standardization; Xp is the p-th variable of the sample after standardization, X i1 -X ip They are the SOC, voltage and temperature data after standardization.
[0057]
[0058] Find the covariance matrix R. The covariance matrix of the standardized matrix X is R. At this time, the covariance matrix R is of order p×n:
[0059]
[0060] Among them, r ij is the correlation coefficient, is the correlation coefficient between two selected variables, and its calculation formula is as follows:
[0061]
[0062] in, is the average value of the i-th group of data, The mean of the j-th variable data set.
[0063] Step a2, extracting the eigenvalue and the first eigenvector of the covariance matrix.
[0064] When performing principal component analysis, the eigenvalues are sorted by size, and their eigenvectors are arranged in the same order to form a characteristic matrix. Since the variance of the principal components analyzed decreases step by step, the amount of information they contain also decreases step by step. Among them, the contribution rate C refers to the proportion of a certain eigenvalue to all eigenvalues, and its calculation formula is:
[0065]
[0066] Among them, y j is the jth eigenvalue; y k is the kth eigenvalue, for example, C1 is the proportion of SOC to all eigenvalues.
[0067] Step a3, performing principal component analysis based on the eigenvalue and the first eigenvector to obtain the principal component value and the corresponding contribution rate to obtain dimension-reduced data.
[0068] The variance contribution rate C of the principal component is used to reflect the amount of information contained, and the cumulative variance contribution rate can determine the number of principal components, that is, the value of m in F1, F2, ..., Fm. The principle for determining the number of principal components is: while i is as small as possible, make C(m) as large as possible to ensure that the amount of information lost by the sample is minimized. When the cumulative contribution rate is greater than 95%, it is considered to be sufficient to reflect the information of the original variable, and the corresponding m is the first m principal components extracted. Calculate the values of m principal components, that is:
[0069] F i =a 1i X1+a 2i X2+…+a pi X pi i=1,2,…,m
[0070] Among them, Fi refers to the i-th principal component of the original variables X1, X2, ..., XP, X1 corresponds to SOC, X2 corresponds to voltage, and X3 corresponds to temperature; a 1i , a 2i , …, a pi is the combination coefficient of Fi, a pi is the eigenvector corresponding to the i-th eigenvalue.
[0071] The variable contributions are accumulated to obtain the cumulative contribution of variable correlation as shown in Table 1:
[0072] Table 1
[0073]
[0074] F1=0.816X1+0.134X2-0.05X3
[0075] F2=0.241X1+0.873X2-0.12X3
[0076] For this project, when two principal components are extracted, the cumulative variance contribution rate of the principal components can reach more than 95%, so two principal components are selected as the input feature parameters of the support vector machine.
[0077] The data-driven double-layer fault diagnosis method provided in this embodiment solves the problem that the lithium battery operation data will inevitably be affected by interference, sensor failure and other factors during the collection process, generating abnormal fault data and affecting the accuracy of fault diagnosis by filtering and reducing the operating data. The present invention uses discrete cosine transform and inverse transform to process the data set, removes abnormal data, and increases data reliability. Through dimensionality reduction, high-dimensional data can be converted into low-dimensional data, making the data easier to process and understand. This helps to simplify the data analysis process and improve efficiency.
[0078] Step S2013, extract features from the dimension-reduced data to obtain feature vectors.
[0079] Specifically, the above step S2013 includes:
[0080] Step b1, periodically segment the dimension-reduced data based on the feature extraction method to construct a data feature set.
[0081] Specifically, the battery fault detection problem can be transformed into a binary classification problem of machine learning, such as Figure 3 As shown, Figure 3 FIG. 1 is a flow chart of a battery fault detection method according to an embodiment of the present invention. First, the data is segmented into cycles based on a feature extraction method, and a L f dimensional feature vector.
[0082] Step b2: set labels for the data feature set to obtain a machine learning data set.
[0083] Specifically, according to the actual working conditions of each sample, a fault or normal label is set, represented by 0 and 1 respectively, so as to construct a data set for machine learning.
[0084] Step b3, sampling the machine learning data set based on a preset number of times to obtain a data sampling set.
[0085] Specifically, let the capacity of the original sample set be Nr, and the elements in the sample set be L fdimensional attribute vector. By sampling the feature data set Nr times, a new training set containing Nr samples can be obtained.
[0086] Step b4, performing set processing on multiple data sampling sets to obtain feature vectors.
[0087] Specifically, the sampling is repeated K times, and a training sample set of K×Nr samples can be obtained.
[0088] The data-driven two-layer fault diagnosis method provided in this embodiment can divide the dimension reduction data into different time periods through period segmentation and feature extraction, and extract a data feature set reflecting the data characteristics from each time period. These data feature sets contain rich information, which helps to understand the characteristics and laws of the data more comprehensively. Setting labels for the data feature sets can make the machine learning data set have clear classification or identification, which is convenient for subsequent model training and learning. By corresponding the labels to the data feature sets, the performance and accuracy of the model can be better evaluated. By sampling the machine learning data set through a sampling method, multiple data sampling sets can be obtained. These data sampling sets can be used for different model training and analysis tasks, which improves the utilization and flexibility of the data. By performing set processing on multiple data sampling sets, more comprehensive and diversified feature vectors can be obtained. These feature vectors can capture more patterns and features of the data, which helps to improve the accuracy and robustness of fault diagnosis.
[0089] Step S202, obtaining fault classification data based on the feature vector and a preset fault analysis classifier; the fault analysis classifier is obtained by training based on negative sample data with fault identification and positive sample data with normal identification.
[0090] Specifically, the above step S202 includes:
[0091] Step S2021, randomly extract the data sampling set, and obtain the Gini index based on the random extraction result.
[0092] Specifically, when constructing the nodes of the decision tree, the variable L is selected by random variable selection. f Randomly extract from the feature attributes to form an attribute subset and then split the node. The number of features selected during node splitting is not less than log2L f The CART algorithm is used to select Gini impurity as the evaluation criterion. The Gini index is calculated by the following formula according to the features selected in step S201, and the feature with the smallest Gini index of the current branch is selected as the split point. In the attribute set, the Gini index of the mth branch node can be defined as:
[0093]
[0094] Among them, Λ is the sample set; λj is the proportion of samples in the jth category; y is the total number of categories.
[0095] Step S2022, splitting is performed based on the Gini index to generate a decision tree.
[0096] Specifically, by splitting based on the Gini index, a decision tree can be generated to classify or predict new data. Each node of the decision tree represents a feature and a corresponding value, and the path from the root node to the leaf node can be regarded as a set of rules for classifying data into corresponding categories.
[0097] Step S2023, using majority voting method to integrate the decision trees to obtain a random forest.
[0098] Specifically, the relevant indicators are tested based on the test set segmented in step S201. Since the decision results are different, in order to obtain the final decision result of the random forest, the prediction results are selected by majority voting. The principle of majority voting is that when a prediction label obtains more than half of the decision tree votes, the final prediction is the prediction label; otherwise, the prediction label will be rejected, as shown in the following formula:
[0099]
[0100] Where: j is the number of categories; θj is the label category; y is the number of categories.
[0101] The data-driven two-layer fault diagnosis method provided in this embodiment generates a decision tree by splitting based on the Gini index. The random forest consists of multiple decision trees, each of which is trained based on a randomly extracted data set. Since each decision tree is trained using only part of the data, the risk of overfitting can be reduced. At the same time, the decision tree is integrated by the majority voting method to further improve the classification accuracy.
[0102] Step S2024, input the target running data into the binary classifier of the random forest model for classification prediction, and obtain the classifier decision results of each data.
[0103] Specifically, the feature vector to be diagnosed is input into the trained random forest binary classifier for classification prediction. If the classifier decision result is 0, it can be judged as a normal operating state; if the classifier decision result is 1, it is judged as a fault state, thereby realizing battery fault diagnosis.
[0104] Step S2025 , based on the classifier decision result, fault location is performed on the target operation data to obtain fault data, where the fault data includes the fault data type and the fault type.
[0105] Specifically, the fault data and fault data information detected this time are output. Battery fault data types include voltage too high, temperature too high, current 0, current too large, voltage too low, etc. When the battery voltage exceeds the normal range, it may be an indicator of battery failure. Too high voltage may cause battery damage or cause safety problems. Too high battery temperature may cause battery performance to deteriorate, and in severe cases may cause a fire. If the battery current is 0, the battery may have failed or been disconnected. If the battery current is too large, it may damage the battery or even cause a safety accident. If the battery voltage is too low, it may cause battery performance to deteriorate, and in severe cases may cause safety problems. Fault information includes fault, normal, and the location where the fault occurred.
[0106] The data-driven two-layer fault diagnosis method provided in this embodiment can simplify the fault diagnosis process by inputting the target operation data into the binary classifier of the random forest for classification prediction, obtaining the classifier decision results of each data, and using the binary classifier of the random forest for fault location. Compared with the traditional fault detection method, the fault location method based on the classifier decision result has higher accuracy and efficiency. Fault location based on the classifier decision result can more accurately identify fault data.
[0107] Step S203, constructing a hyperplane function based on the fault classification data, calculating the optimal solution of the hyperplane function, and obtaining the fault diagnosis result of the device to be diagnosed.
[0108] Specifically, the above step S203 includes:
[0109] Step S2031, inputting the feature vector into a preset vector machine model to obtain a hyperplane function corresponding to the fault classification data.
[0110] Specifically, the abnormal values of the battery fault detection results are diagnosed and classified. For this project, y i is the category function, x i is a vector constructed using F1-i and F2-i.
[0111]
[0112] Among them, x i is the training sample input, y i is the training sample output, corresponding to battery cell short circuit fault, aging fault and short circuit fault.
[0113] Step S2032, calculating the optimal solution of the hyperplane to obtain the fault diagnosis result corresponding to the fault classification data.
[0114] In some optional embodiments, a hyperplane divides the short-circuit fault data, aging fault data and short-circuit fault data of the device to be diagnosed, and the expression of the hyperplane is:
[0115] w T *x i +b=0
[0116] Among them, w is the weight vector, x i is the training sample input; b is the threshold.
[0117] In the data-driven two-layer fault diagnosis method provided in this embodiment, the weight vector and threshold in the hyperplane expression are adjustable. By adjusting these parameters, the performance of the model can be optimized and adapted to different data sets and tasks.
[0118] The problem of finding the optimal plane (battery fault classification problem) can be summarized as an optimization problem. The maximum classification boundary can be obtained by finding the minimum value of the formula, thereby achieving the best battery fault classification effect.
[0119]
[0120] In the above formula, C is the penalty parameter; ξ i is a non-negative relaxation factor.
[0121] In order to solve this constrained optimization problem, this project introduces the Lagrange function:
[0122] L(w,b,a)=1 / 2∥w∥-a(y((w·x)+b)-1)
[0123] In the formula, a i >0 is the Lagrange multiplier. The solution of the constrained optimization problem is determined by the saddle point of the Lagrange function, and the optimization problem satisfies the partial derivatives of w and b to be 0 at the saddle point. The QP problem is converted into the corresponding dual problem:
[0124] The QP problem is for predictive control with constraints. Generally speaking, it is impossible to obtain an analytical solution to the constrained optimization problem.
[0125]
[0126] Solve the optimal solution
[0127] The most accurate fault classification decision function for satellite batteries is obtained through a series of solutions.
[0128]
[0129] In the above formula is the optimal Lagrange multiplier; b* a threshold for classifying battery failure, and x j With x i A vector constructed from SOC and voltage representing different fault data.
[0130] In order to improve the generalization ability, the radial basis kernel function is used:
[0131]
[0132] where δ 2 is the kernel function width.
[0133] The firefly algorithm is used to optimize the kernel function parameters to determine the penalty parameter C and kernel width δ 2 The optimal solution. Set the parameters of the firefly algorithm, including the firefly population size N, the spatial dimension of the optimization problem D, the medium's absorption coefficient of light γ, the firefly's initial attraction β0, and the maximum number of iterations t max Among them, x i represents the location of the i-th firefly individual, x j represents the location of the jth firefly. The distance between the i-th firefly and the j-th firefly is r ij It can be expressed as:
[0134] r ij =||x i -x j ||
[0135] Determine the fitness function and set the diagnostic accuracy as the fitness value. The calculation formula is:
[0136]
[0137] Where: ψ is the diagnostic accuracy; x n is the number of test set samples; m is the type of battery failure, x m is the number of samples with correct diagnosis of the mth type of fault. The positions of individual fireflies are randomly initialized, and each individual firefly represents a set of solutions of (C, δ). The initial brightness of the individual firefly is calculated according to the fitness function; the attractiveness β of the firefly is updated using the following formula:
[0138]
[0139] If the jth firefly is brighter, the formula for the ith firefly to move toward the jth firefly is:
[0140] x i (t+1)=x i (t)+β(x j (t)-xi (t))+α i (t)ε
[0141] Among them, x i (t+1) is the position of the i-th firefly in the space at the t+1-th iteration and the t-th iteration; x j (t) is the position of the jth firefly in the space at the tth iteration; α i (t) is the step factor of the i-th firefly at the t-th iteration; ε is the coefficient that obeys the normal distribution, and its value range is [-0.5, 0.5]. Update the position of the firefly and then recalculate the brightness of the firefly; iterate until the target accuracy or the maximum number of iterations is reached, output the optimal solution of (C, δ), and assign it to the fault classification decision function, so that the test set samples can be diagnosed.
[0142] Specifically, by classifying the fault data based on the hyperplane, the fault diagnosis results of battery cell short circuit fault, aging fault and short circuit fault can be obtained. Hyperplane segmentation is achieved by constructing a hyperplane function and calculating the optimal solution. By dividing the fault data into positive samples (normal state) and negative samples (fault state), the hyperplane function can be used to classify the normal state and the fault state. The fault diagnosis results of battery cell short circuit fault, aging fault and short circuit fault can be interpreted and evaluated based on the results of hyperplane segmentation. If the diagnosis result shows that the battery cell has a short circuit fault, the circuit and connection of the battery cell can be further checked to determine whether there are problems such as circuit failure or poor contact. If the diagnosis result shows that the battery cell has an aging fault, the degree of battery aging and service life can be evaluated, and appropriate measures can be taken for maintenance or replacement. If the diagnosis result shows that the battery cell has other types of faults, further fault analysis and repair can be performed.
[0143] The data-driven two-layer fault diagnosis method provided in this embodiment can quickly detect and identify abnormal data by classifying fault data based on a hyperplane. When abnormal data appears, the hyperplane model can quickly make a judgment, and the fault diagnosis results based on the hyperplane have good interpretability. By understanding and analyzing the hyperplane, the nature and mode of the fault can be better understood, and the credibility and transparency of the fault diagnosis results can be improved.
[0144] In the present embodiment, a double-layer fault diagnosis device based on data drive is also provided, and the device is used to implement the above-mentioned embodiment and preferred implementation mode, and the description has been made no further. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware is also possible and conceived.
[0145] This embodiment provides a data-driven two-layer fault diagnosis device, such as Figure 4 As shown, including:
[0146] The feature extraction module 401 is used to obtain the operating data of the device to be diagnosed, perform feature extraction on the operating data, and obtain a feature vector.
[0147] The fault analysis module 402 is used to obtain fault classification data based on the feature vector and a preset fault analysis classifier; the fault analysis classifier is obtained by training based on negative sample data with fault identification and positive sample data with normal identification.
[0148] The fault diagnosis module 403 is used to construct a hyperplane function based on the fault classification data, calculate the optimal solution of the hyperplane function, and obtain the fault diagnosis result of the device to be diagnosed.
[0149] The data-driven double-layer fault diagnosis device provided in this embodiment first extracts features from the operating data to obtain feature vectors, and converts the original data into feature vectors, which can remove noise and redundant information, thereby simplifying the data set and making data processing and analysis more efficient. In machine learning and data analysis, using feature vectors as input can help improve the performance of the model. By learning and optimizing feature vectors, the model can better capture the patterns and relationships of the data and improve the accuracy of prediction and classification. Secondly, the fault classification data is obtained by using feature vectors and preset fault analysis classifiers. The decision tree model is a powerful machine learning algorithm as a fault analysis classifier, which can be classified according to feature vectors. Compared with traditional fault detection methods, the fault classification method based on feature vectors and decision tree models has higher accuracy. Finally, by constructing a hyperplane function based on fault classification data and calculating the optimal solution of the hyperplane function, the fault diagnosis result of the device to be diagnosed is obtained. The hyperplane function can determine the type of fault according to the input feature vector, which helps to quickly locate the cause of the fault, further improves the efficiency of fault diagnosis, and takes into account the relationship between the diagnostic accuracy and the operation speed.
[0150] In some optional implementations, the feature extraction module 401 includes:
[0151] The filtering unit is used to filter the operating data to obtain filtered data.
[0152] The dimension reduction unit is used to perform dimension reduction processing on the filtered data to obtain dimension reduced data.
[0153] The extraction unit is used to extract features from the dimension-reduced data to obtain feature vectors.
[0154] The data-driven double-layer fault diagnosis device provided in this embodiment solves the problem that the lithium battery operation data will inevitably be affected by interference, sensor failure and other factors during the collection process, generating abnormal fault data and affecting the accuracy of fault diagnosis by filtering and reducing the operating data. The present invention uses discrete cosine transform and inverse transform to process the data set, removes abnormal data, and increases data reliability. Through dimensionality reduction, high-dimensional data can be converted into low-dimensional data, making the data easier to process and understand. This helps to simplify the data analysis process and improve efficiency.
[0155] In some optional implementations, the dimension reduction unit includes:
[0156] The covariance matrix calculation subunit is used to construct a variable matrix based on the filtered data and calculate the covariance matrix of the variable matrix.
[0157] The first eigenvector extraction subunit is used to extract the eigenvalues and the first eigenvector of the covariance matrix.
[0158] The principal component analysis subunit is used to perform principal component analysis based on the eigenvalue and the first eigenvector to obtain the principal component value and the corresponding contribution rate, and obtain the dimension reduction data.
[0159] In some optional embodiments, the extraction unit comprises:
[0160] Construct subunits for periodically segmenting the dimension-reduced data based on feature extraction methods and constructing data feature sets.
[0161] The label subunit is used to set labels for the data feature set to obtain a machine learning data set.
[0162] The sampling subunit is used to sample the machine learning data set based on a preset number of times to obtain a data sampling set.
[0163] The processing unit is used to perform collective processing on multiple data sampling sets to obtain feature vectors.
[0164] The data-driven double-layer fault diagnosis device provided in this embodiment can divide the dimension reduction data into different time periods through period segmentation and feature extraction, and extract a data feature set reflecting the data characteristics from each time period. These data feature sets contain rich information, which helps to understand the characteristics and laws of the data more comprehensively. Setting labels for the data feature sets can make the machine learning data set have clear classification or identification, which is convenient for subsequent model training and learning. By corresponding the labels to the data feature sets, the performance and accuracy of the model can be better evaluated. By sampling the machine learning data set through a sampling method, multiple data sampling sets can be obtained. These data sampling sets can be used for different model training and analysis tasks, which improves the utilization and flexibility of the data. By performing set processing on multiple data sampling sets, more comprehensive and diversified feature vectors can be obtained. These feature vectors can capture more patterns and features of the data, which helps to improve the accuracy and robustness of fault diagnosis.
[0165] In some optional implementations, the fault analysis module 402 includes:
[0166] The random sampling unit is used to randomly extract data from a sample set and obtain a Gini index based on the random sampling result.
[0167] The split unit is used to split based on the Gini index and generate a decision tree.
[0168] The integration unit is used to integrate the decision trees using the majority voting method to obtain a random forest.
[0169] The data-driven two-layer fault diagnosis device provided in this embodiment generates a decision tree by splitting based on the Gini index. The random forest consists of multiple decision trees, each of which is trained based on a randomly extracted data set. Since each decision tree is trained using only part of the data, the risk of overfitting can be reduced. At the same time, the decision tree is integrated by the majority voting method to further improve the classification accuracy.
[0170] In some optional implementations, the fault analysis module 402 includes:
[0171] The prediction unit is used to input the target running data into the binary classifier of the random forest for classification prediction and obtain the classifier decision results of each data.
[0172] The positioning unit is used to locate the fault of the target operation data based on the decision result of the classifier to obtain fault data, which includes the fault data type and the fault type.
[0173] The data-driven two-layer fault diagnosis device provided in this embodiment can simplify the fault diagnosis process by inputting the target operation data into the binary classifier of the random forest for classification prediction, obtaining the classifier decision results of each data, and using the binary classifier of the random forest for fault location. Compared with the traditional fault detection method, the fault location method based on the classifier decision result has higher accuracy and efficiency. Fault location based on the classifier decision result can more accurately identify fault data.
[0174] In some optional implementations, the fault diagnosis module 403 includes:
[0175] An input unit, used to input the feature vector into a preset vector machine model to obtain a hyperplane function corresponding to the fault classification data;
[0176] The optimal solution calculation unit is used to calculate the optimal solution of the hyperplane to obtain the fault diagnosis result corresponding to the fault classification data.
[0177] The data-driven two-layer fault diagnosis device provided in this embodiment can quickly detect and identify abnormal data by classifying fault data based on a hyperplane. When abnormal data appears, the hyperplane model can quickly make a judgment, and the fault diagnosis results based on the hyperplane have good interpretability. By understanding and analyzing the hyperplane, the nature and mode of the fault can be better understood, and the credibility and transparency of the fault diagnosis results can be improved.
[0178] In some optional embodiments, a hyperplane divides the short-circuit fault data, aging fault data, and short-circuit fault data of the device to be diagnosed, and the expression of the hyperplane is:
[0179] w T *x i +b=0
[0180] Among them, x i is the training sample input, w is the weight vector, and b is the threshold.
[0181] In the data-driven double-layer fault diagnosis device provided in this embodiment, the weight vector and threshold in the hyperplane expression can be adjusted. By adjusting these parameters, the performance of the model can be optimized and adapted to different data sets and tasks.
[0182] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.
[0183] The hardware encryption device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.
[0184] The embodiment of the present invention also provides a computer device having the above Figure 4 The hardware encryption device shown.
[0185] See also Figure 5 , Figure 5 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 5 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.
[0186] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.
[0187] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.
[0188] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0189] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.
[0190] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.
[0191] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.
[0192] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. A data-driven fault diagnosis method, characterized in that: Methods include: Acquiring operating data of the device to be diagnosed, performing feature extraction on the operating data, and obtaining a feature vector; Obtaining fault classification data based on the feature vector and a preset fault analysis classifier; The fault analysis classifier is obtained by training based on negative sample data with fault identification and positive sample data with normal identification; A hyperplane function is constructed based on the fault classification data, and a fault diagnosis result of the device to be diagnosed is obtained based on the hyperplane function.
2. The method according to claim 1, characterized in that Extract features from running data, including: Performing filtering on the operating data to obtain filtered data; Performing dimensionality reduction processing on the filtered data to obtain dimensionality reduced data; Feature extraction is performed on the dimension-reduced data to obtain the feature vector.
3. The method according to claim 2, characterized in that The performing dimensionality reduction processing on the filtered data to obtain dimensionality reduction data includes: Constructing a variable matrix based on the filtered data, and calculating a covariance matrix of the variable matrix; Extracting the eigenvalues and the first eigenvector of the covariance matrix; A principal component analysis is performed based on the eigenvalue and the first eigenvector to obtain the principal component value and the corresponding contribution rate, thereby obtaining the dimension reduction data.
4. The method according to claim 2, characterized in that: The extracting features from the dimension-reduced data comprises: Performing periodic segmentation on the dimension reduction data based on a feature extraction method to construct a data feature set; Setting labels for the data feature set to obtain a machine learning data set; Sampling the machine learning data set based on a preset number of times to obtain a data sampling set; The plurality of data sampling sets are collectively processed to obtain the feature vector.
5. The method according to claim 4, characterized in that The fault classification data obtained based on the feature vector and the preset fault analysis classifier includes: Randomly sampling the data sample set, and obtaining a Gini index based on the random sampling result; Perform splitting based on the Gini index to generate a decision tree; The decision trees are integrated using a majority voting method to obtain a random forest model; Input the operating data into the binary classifier of the random forest model for classification prediction, and obtain the classifier decision results of each data; The operating data is fault located based on the decision result of the classifier to obtain fault data, wherein the fault data includes a fault data type and a fault type.
6. The method according to claim 1, characterized in that The constructing a hyperplane function based on the fault classification data, and obtaining a fault diagnosis result of the device to be diagnosed based on the hyperplane function, comprises: Inputting the feature vector into a preset vector machine model to obtain a hyperplane function corresponding to the fault classification data; An optimal solution of the hyperplane is calculated to obtain a fault diagnosis result corresponding to the fault classification data.
7. The method according to claim 1, characterized in that The hyperplane function is used to separate the short-circuit fault data, aging fault data and short-circuit fault data of the device to be diagnosed. The expression of the hyperplane function is: w T *x i +b=0 Among them, x i is the training sample input, w is the weight vector, and b is the threshold.
8. A data-driven two-layer fault diagnosis device, characterized in that: The device includes: A feature extraction module, used to obtain operating data of the device to be diagnosed, perform feature extraction on the operating data, and obtain a feature vector; A fault analysis module, used to obtain fault classification data based on the feature vector and a preset fault analysis classifier; the fault analysis classifier is obtained by training based on negative sample data with fault identification and positive sample data with normal identification; The fault diagnosis module is used to construct a hyperplane function based on the fault classification data, and obtain a fault diagnosis result of the device to be diagnosed based on the hyperplane function.
9. A computer device, characterized in that: include: The memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the data-driven two-layer fault diagnosis method according to any one of claims 1 to 7 by executing the computer instructions.
10. A computer-readable storage medium, characterized in that: Computer instructions are stored on the computer-readable storage medium, and the computer instructions are used to enable a computer to execute the data-driven dual-layer fault diagnosis method according to any one of claims 1 to 7.
Citation Information
Cited By
Track circuit fault diagnosis method, device, equipment, medium and product
CN119881607A
