Fault diagnosis method, system and storage medium

Abstract

The present invention relates to the technical field of equipment fault diagnosis, and discloses a fault diagnosis method, system and storage medium. The method takes into account multiple characteristics of data variables and analyzes the differentiated sensitivities of multiple variables to the same characteristic, thereby realizing differentiated feature selection, reducing the impact of diagnostic performance caused by sensitivity differences of characteristics and redundant and invalid characteristics, and updating the fault parameter library in real time, thereby identifying and isolating unknown faults, and realizing intelligent early warning and fault diagnosis for equipment operation safety.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment fault diagnosis, and in particular to a fault diagnosis method, system and storage medium. Background Art

[0002] With the development of data mining and fault diagnosis, a wide variety of features can be extracted from data. However, not every feature can reflect the information required for fault diagnosis. Therefore, it is necessary to select the features with the best classification performance from the numerous feature variables. In the fault diagnosis of large equipment, since different sample variables may have different sensitivities to features, the differences in sensitivity of different samples to features can lead to adverse effects. Therefore, the fault diagnosis methods in the existing technology are significantly affected by the type of sample. There is an urgent need to provide a fault diagnosis method that can perform differentiated sensitivity analysis on different features to improve diagnostic capabilities. Summary of the Invention

[0003] The present invention provides a fault diagnosis method, system and storage medium to achieve differentiated sensitivity analysis of different features and improve diagnostic capabilities.

[0004] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0005] In a first aspect, the present invention provides a fault diagnosis method, comprising:

[0006] S1: Obtain a historical normal operation data training set and a historical fault operation data set as data to be analyzed, and extract characteristic variables from the data to be analyzed;

[0007] S2: performing differential sensitivity analysis on the feature variables to obtain selected feature variables, wherein the selected feature variables include a set of feature variables without faults and each fault type after selection; and calculating canonical correlation analysis (CCA) parameters and fault detection thresholds using the set of feature variables without faults after feature selection;

[0008] S3: Calculate the typical correlation analysis parameters corresponding to each fault type based on the selected characteristic variable set of each fault type, and build a fault parameter library containing parameters of each fault type;

[0009] S4: Select real-time collected data, calculate CCA residual statistics based on the feature variable set of the fault-free type after feature selection and the corresponding canonical correlation analysis parameters, compare the residual statistics with the fault detection threshold, and if a fault occurs, input its feature data into the fault parameter library to calculate the residual statistics corresponding to each fault type parameter;

[0010] S5: Based on the residual statistics corresponding to each fault type parameter, the type confidence rate is obtained. According to the type confidence rate, it is determined whether it is an unknown fault. If it is an unknown fault, the parameters are calculated and updated to the fault parameter library so that subsequent diagnosis can identify such faults. Otherwise, the predicted classification probability of each type of fault is calculated, and the fault type of the data is determined based on the classification probability.

[0011] In a second aspect, the present application provides a fault diagnosis system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the computer program.

[0012] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method steps described in the first aspect.

[0013] Beneficial effects:

[0014] The fault diagnosis method provided by the present invention performs differential sensitivity analysis on feature variables to obtain selected feature variables, calculates typical correlation analysis parameters corresponding to each fault type based on the feature variable set of each selected fault type, and constructs a fault parameter library containing parameters of each fault type; selects test feature data, calculates residuals based on the feature variable set of the selected fault-free type and the corresponding typical correlation analysis parameters, compares the residuals with the fault detection threshold, and if a fault occurs, inputs the test feature data into the fault parameter library to calculate the residuals corresponding to each fault type parameter; obtains a type confidence rate based on the residuals corresponding to each fault type parameter, and determines whether it is an unknown fault based on the type confidence rate. If it is an unknown fault, the parameters are calculated and updated to the fault parameter library, allowing subsequent diagnosis to identify such faults. Otherwise, the predicted classification probability of each fault type is calculated, and the fault type of the data is determined based on the classification probability. In this way, multiple features of the data variables are considered, and the differential sensitivity of multiple variables to the same feature is analyzed, thereby achieving differentiated feature selection, reducing the impact of diagnostic performance caused by feature sensitivity differences and redundant and invalid features, and updating the fault parameter library in real time, thereby identifying and isolating unknown faults, and realizing intelligent early warning and fault diagnosis for the safe operation of large-scale equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flowchart of a fault diagnosis method according to a preferred embodiment of the present invention;

[0016] Figure 2 Schematic diagram of the detection effect of Fault 3 in a preferred embodiment of the present invention;

[0017] Figure 3 This is a diagram showing the fault isolation effect of an unknown fault in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following is a clear and complete description of the technical solutions of the present invention. It should be understood that the embodiments described are only a portion of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0019] Unless otherwise defined, the technical or scientific terms used in the present invention shall have the usual meanings understood by persons of ordinary skill in the field to which the present invention belongs. The words "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "a" do not indicate a quantity limitation, but rather indicate the existence of at least one. Words such as "connected" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0020] It should be understood that the fault diagnosis method of the present application is applicable to data-driven systems such as gearboxes, high-speed train braking systems, and bearings. This is for illustrative purposes only and is not intended to be limiting.

[0021] See Figure 1 , this application provides a fault diagnosis method, comprising:

[0022] S1: Obtain a historical normal operation data training set and a historical fault operation data set as the data to be analyzed, and extract characteristic variables from the data to be analyzed;

[0023] S2: Perform differential sensitivity analysis on the feature variables to obtain selected feature variables. The selected feature variables include the feature variable set of no faults and each fault type after selection. The feature variable set of no faults after feature selection is used to calculate the canonical correlation analysis parameters and the fault detection threshold.

[0024] S3: Calculate the typical correlation analysis parameters corresponding to each fault type based on the selected characteristic variable set of each fault type, and build a fault parameter library containing parameters of each fault type;

[0025] S4: Select real-time collected data, calculate CCA residual statistics based on the characteristic variable set and typical correlation analysis parameters of each fault type after feature selection, compare the residual statistics with the fault detection threshold, and if a fault occurs, input its characteristic data into the fault parameter library to calculate the residual statistics corresponding to each fault type parameter;

[0026] S5: Based on the residual statistics corresponding to each fault type parameter, the type confidence rate is obtained. Based on the type confidence rate, it is determined whether it is an unknown fault. If it is an unknown fault, the parameters are calculated and updated to the fault parameter library, allowing subsequent diagnosis to identify this type of fault. Otherwise, the predicted classification probability of each type of fault is calculated, and the fault type of the data is determined based on the classification probability. This achieves intelligent fault diagnosis.

[0027] In this embodiment, the predicted classification probability is a basis for determining the type of data fault. Intelligent fault diagnosis refers to identifying unknown faults and updating the parameter library.

[0028] The above fault diagnosis method fully considers the differences in sensitivity of different features to different variable data, and differentially selects the feature variables with the best classification effect as the data set for diagnosis to reduce the impact of feature sensitivity differences and redundant invalid feature information. It can accurately and quickly detect faults and diagnose fault types, and update fault types in real time.

[0029] Optionally, the S1 includes:

[0030] S11: Select the historical normal operation data training set, and assume that each sample data contains n variable data, which are V 0i , where i is the i-th variable data, i=1,…,n;

[0031] S12: Select historical fault data, expressed as V fi , where f is the fth data type, f = 0, 1, ..., m, where when f = 0 it indicates normal type data, and when f is not 0 it indicates the fth type of fault data, with a total of m types of faults;

[0032] S13: Based on the n variable data in the sample, the data features in normal and various fault states are calculated respectively, including: mean value, effective value, center of gravity frequency, relative power spectrum entropy, and deep neural network middle layer output, denoted as I fip , where p is the pth feature type, p = 1, ..., l, l is the total number of extracted feature types, and the calculation formula is as follows:

[0033]

[0034]

[0035]

[0036]

[0037] Where, I fi1 ~I fi4 Vfi The mean, effective value, center frequency and relative power spectrum entropy, N c is the number of sampling points in the calculation cycle, k is the kth sampling moment, j is the jth frequency sample, i is the i-th variable data, i = 1, ..., n; f is the f-th data type, f = 0, 1, ..., m, m is the total number of fault types, f = 0 is the normal data type, z, s k (j), f k (j) are variable data V fi The number of frequency samples, spectrum amplitude, and frequency value after Fourier transform.

[0038] It's important to note that the output of a deep neural network's intermediate layer refers to the output of intermediate layers in deep machine learning, including convolutional neural networks and autoencoders. This intermediate layer abstracts the features of the input data in the machine learning process into another dimensional space, thereby outputting more abstract features and achieving better linear partitioning.

[0039] Optionally, the S2 includes:

[0040] S21: Use the Fisher score method to perform differential sensitivity analysis on the extracted characteristic variables and select them. The calculation formula is:

[0041]

[0042]

[0043]

[0044]

[0045]

[0046] Where, F i (p) is the Fisher score of the p-th characteristic variable in the i-th sample variable, and are the intra-class variance and inter-class variance of the p-th characteristic variable in the i-th sample variable, L i is the total number of characteristic variable samples of the i-th type sample variable, L fi is the number of characteristic variable samples of the fth type of data type for the i-th type sample variable, I fip (k) is the value of the p-th characteristic variable of the f-th data type in the i-th sample variable at the k-th sampling time, and are the f-th data type and the average value of all data types on the p-th feature of the i-th sample variable;

[0047] Then, the Fisher scores of each characteristic variable in the obtained n types of sample variables are sorted from large to small, and the top h characteristic variables of each sorting result are selected to form the selected characteristic variable set. The calculation formula is as follows:

[0048] J fi =[Q fi1 ,Q fi2 ,……,Q fih ] (10)

[0049] Where Q fi1 -Q fih is the characteristic variable I fip The first h feature variables after sorting from largest to smallest;

[0050] S22: Split the selected normal feature variable set into the first and second data sets, where the first data set is J 01 -J 0d , the second data set is J 0d+1 -J 0n ,in, Right now:

[0051] y a (k)=[J 01 (k),J 02 (k),……,J 0d (k)] (11)

[0052] y b (k)=[J 0d+1 (k),……,J 0n (k)] (12)

[0053] Where y a (k) is the data of the kth sampling moment of the first data set, y b (k) is the data of the kth sampling moment of the second data set, J 0i (k) is the value of the selected feature variable set of the normal data type of the i-th sample variable at the k-th moment, i is 1, 2, ... n, d represents the intermediate value, which is used to divide the selected normal feature variable set into two data sets;

[0054] Use Y a and Y b Represent the first and second datasets:

[0055] Y a =[y a (1),…,y a (N)] (13)

[0056] Y b =[y b(1),…,y b (N)] (14)

[0057] Where N is the total number of samples;

[0058] S23: Calculate typical correlation analysis parameters and fault detection thresholds, where the typical correlation analysis parameters include a correlation matrix and residual signal statistics:

[0059] According to Y a and Y b The two data sets are combined with the CCA method to obtain the canonical correlation analysis parameters Σ, L and J, as well as the residual signal statistics and

[0060] The fault detection threshold calculation formula is:

[0061]

[0062]

[0063] Where, J th,r1 and J th,r2 are two thresholds for fault detection, χ α 2 (m a ) and χ α 2 (m b ) represent the degrees of freedom m a and m b The chi-square distribution of α represents the confidence level, which can be set by the user according to the allowable false alarm rate.

[0064] Optionally, the S3 includes:

[0065] S31: Select various fault characteristic variable sets J after selection fi , calculate the typical related variables corresponding to each type of fault data;

[0066] S32: Calculate the fault parameter set based on typical related variables of multiple fault data:

[0067] R f =[Σ f ,L f ,J f ] (17)

[0068] Where R f is the fault parameter set of the fth type fault, Σ f ,L f ,J f are the relevant analysis parameters corresponding to various types of fault data, and R fEstablish a fault parameter library for various types of faults.

[0069] Optionally, the S4 includes:

[0070] S41: Select the real-time acquisition data set and calculate the characteristic variables after differential sensitivity selection;

[0071] S42: Calculate the residual signal statistics based on the obtained characteristic variables and when Greater than the fault detection threshold J th,r1 or Greater than the fault detection threshold J th,r2 When , it is determined to be a fault, otherwise it is determined to be normal, that is:

[0072]

[0073] S43: When the data is judged to be faulty, the parameters of each type of fault in the fault parameter library are used in turn to calculate the residual signal statistics corresponding to each fault type parameter and Among them, the CCA method is used to calculate the residual signal statistics corresponding to each fault type parameter and

[0074] Optionally, the S5 includes:

[0075] S51: Calculate the type confidence rate of each current fault type. The calculation formula is:

[0076]

[0077] Where u represents the uth statistical sample, u=1,…,L T , L T is the total length of the sample of data residual statistics, G f is the type confidence rate of the fth fault type, g f1 (u) and g f2 (u) is the classification factor at the u-th sample moment, and the calculation formula is as follows:

[0078]

[0079]

[0080] Where, and is the residual signal statistic of the fth fault type at the uth sample moment;

[0081] S52: Based on the current confidence rate of each fault type, determine whether it is an unknown fault and calculate the predicted classification probability of the data set. The judgment method is as follows:

[0082]

[0083] Where G o is the type confidence rate of the o-th fault type, o=1,…,m, m is the total number of fault types, o≠f, β is the confidence threshold, and the value range is (0, 1). The specific value is set according to the user's definition of unknown faults;

[0084] S53: Calculate the predicted classification probability when the type of the current data is not an unknown fault. The calculation formula is as follows:

[0085]

[0086] Where H f Determine the probability of each type of prediction classification for the current data sample,

[0087] S54: When the current data type is unknown fault, the typical related parameters of the unknown fault are calculated as: Σ′, L′ and J′, forming a fault parameter set of the new fault type:

[0088] R′=[Σ′,L′,J′] (24)

[0089] Where R′ is the fault parameter set of the new fault type. Adding R′ to the fault parameter library increases the total number of fault types.

[0090] As a preferred implementation of this embodiment, this embodiment takes a certain gearbox acceleration signal as an example to further illustrate and verify the method of the present invention.

[0091] This example considers four gearbox faults F1 to F4.

[0092] This embodiment uses a data set collected from an acceleration sensor on a gearbox to conduct experimental verification to illustrate the feasibility and effectiveness of the present invention.

[0093] First, we selected acceleration sensor measurements from four gearbox locations: Sensor 1, Sensor 2, Sensor 3, and Sensor 4. The sampling frequency was 6400 Hz, and 12,000 normal, fault-free data points were acquired from each sensor. This data was used to establish a fault detection method based on canonical correlation analysis for fault diagnosis. The fault parameter library used in the fault isolation algorithm was constructed using 12,000 data points per type of historical fault samples. Finally, we validated the proposed method using test data from different fault types.

[0094] In actual operation, the detection effect of fault 3 is as follows Figure 2 As shown, the detection index J T1 、J T2 They are the two statistical indicators of the fault diagnosis method of the present invention, and the fault injection starts from the 187th sample. Figure 2 It can be seen that the method of the present invention can accurately detect the occurrence of faults in vibration signals. Figure 2 The medium statistic is the CCA residual statistic, and the threshold is the fault detection threshold.

[0095] The nearest neighbor (KNN) method was used to compare with the proposed method. Due to the confusability of the gearbox vibration signal, the KNN method's correct isolation rate (CIR) for fault F2 was very low, at only 2.68%, misclassifying the easily confused F2 and F3 fault state signals. The proposed fault diagnosis method, however, achieved a good fault isolation rate. Table 1 shows a comparison of the two methods' performance indicators, fault detection rate (FDR) and CIR.

[0096] Table 1 Gearbox fault detection and isolation results

[0097]

[0098] The CIR of the method of the present invention in the above table is expressed by the predicted classification probability. It can be seen from the above table that the method of the present invention can effectively improve the correct fault isolation rate and has better isolation performance.

[0099] In order to verify the effectiveness of the method of the present invention in identifying and isolating unknown faults, a set of unknown fault data was selected and tested and diagnosed using the method of the present invention. The results of the detection and isolation before and after the fault parameter library was updated are shown in Table 2. Figure 3 This is the unknown fault isolation effect diagram after updating the fault parameter library.

[0100] Table 2 Unknown fault detection and isolation results before and after the fault parameter library is updated

[0101]

[0102] From Table 2 and Figure 3 It can be seen that the method of the present invention has a good effect on identifying and isolating unknown faults. After updating the fault parameter library according to the unknown fault data, such faults can be accurately detected and isolated.

[0103] The present application also provides a fault diagnosis system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method. The fault diagnosis system can implement various embodiments of the above-described fault diagnosis method and achieve the same beneficial effects, which are not described in detail here.

[0104] The present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method steps described above. This computer-readable storage medium can implement various embodiments of the method described above and achieve the same beneficial effects, which are not described in detail here.

[0105] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A fault diagnosis method, characterized in that: include: S1: Obtain a historical normal operation data training set and a historical fault operation data set as data to be analyzed, and extract characteristic variables from the data to be analyzed; S2: Performing differential sensitivity analysis on the feature variables to obtain selected feature variables, wherein the selected feature variables include a set of feature variables without faults and each fault type after selection; and calculating canonical correlation analysis (CCA) parameters and fault detection thresholds using the set of feature variables without faults after feature selection; S3: Calculate the typical correlation analysis parameters corresponding to each fault type based on the selected characteristic variable set of each fault type, and build a fault parameter library containing parameters of each fault type; S4: Select real-time collected data, calculate CCA residual statistics based on the feature variable set of the fault-free type after feature selection and the corresponding canonical correlation analysis parameters, compare the residual statistics with the fault detection threshold, and if a fault occurs, input its feature data into the fault parameter library to calculate the residual statistics corresponding to each fault type parameter; S5: Based on the residual statistics corresponding to each fault type parameter, a type confidence rate is obtained. Based on the type confidence rate, whether it is an unknown fault is determined. If it is an unknown fault, the parameters are calculated and updated to the fault parameter library so that subsequent diagnosis can identify such faults. Otherwise, the predicted classification probability of each type of fault is calculated, and the fault type of the data is determined based on the classification probability. The S5 includes: S51: Calculate the type confidence rate of each current fault type. The calculation formula is: Where u represents the uth statistical sample, u=1,…,L T , L T is the total length of the sample of data residual statistics, G f is the type confidence rate of the fth fault type, g f1 (u) and g f2 (u) is the classification factor at the u-th sample moment, and the calculation formula is as follows: Where, and is the residual signal statistic of the f-th fault type at the u-th sample moment.

2. The fault diagnosis method according to claim 1, characterized in that: Said S1 comprises: S11: Select the historical normal operation data training set, and assume that each sample data contains n variable data, which are V 0i , where i is the i-th variable data, i=1,…,n; S12: Select historical fault data, expressed as V fi , where f is the fth data type, f = 0, 1, ..., m, where when f = 0 it indicates normal type data, and when f is not 0 it indicates the fth type of fault data, with a total of m types of faults; S13: Based on the n variable data in the sample, the data features in normal and various fault states are calculated respectively, including: mean value, effective value, center of gravity frequency, relative power spectrum entropy, and deep neural network middle layer output, denoted as I fip , where p is the pth feature type, p=1,…,l, and l is the total number of extracted feature types.

3. The fault diagnosis method according to claim 1, characterized in that: The S2 includes: S21: Use the Fisher score method to perform differential sensitivity analysis on the extracted characteristic variables and select them. The calculation formula is: Where, F i (p) is the Fisher score of the p-th characteristic variable in the i-th sample variable, and are the intra-class variance and inter-class variance of the p-th characteristic variable in the i-th sample variable, L i is the total number of characteristic variable samples of the i-th type sample variable, L fi is the number of characteristic variable samples of the fth type of data type for the i-th type sample variable, I fip (k) is the value of the p-th characteristic variable of the f-th data type in the i-th sample variable at the k-th sampling time, and are the f-th data type and the average value of all data types on the p-th feature of the i-th sample variable; Then, the Fisher scores of each characteristic variable in the obtained n types of sample variables are sorted from large to small, and the top h characteristic variables of each sorting result are selected to form the selected characteristic variable set. The calculation formula is as follows: J fi =[Q fi1 ,Q fi2 ,……,Q fih ] (9) Where Q fi1 -Q fih is the characteristic variable I fip The first h feature variables after sorting from largest to smallest; S22: Split the selected normal feature variable set into the first and second data sets, where the first data set is J 01 -J 0d , the second data set is J 0d+1 -J 0n ,in, Right now: y a (k)=[J 01 (k),J 02 (k),……,J 0d (k)] (10) y b (k)=[J 0d+1 (k),……,J 0n (k)] (11) Where y a (k) is the data of the kth sampling moment of the first data set, y b (k) is the data of the kth sampling moment of the second data set, J 0i (k) is the value of the selected feature variable set of the normal data type of the i-th sample variable at the k-th moment, i is 1, 2...n, d represents the intermediate value, which is used to divide the selected normal feature variable set into two data sets; Use Y a and Y b Represent the first and second datasets: AND a =[and a (1),…,and a (N)] (12) AND b =[and b (1),…,and b (N)] (13) Where N is the total number of samples; S23: Calculate typical correlation analysis parameters and fault detection thresholds, where the typical correlation analysis parameters include a correlation matrix and residual signal statistics: According to Y a and Y b The two data sets are combined with the CCA method to obtain the canonical correlation analysis parameters Σ, L and J, as well as the residual signal statistics and The fault detection threshold calculation formula is: Where, J th,r1 and J th,r2 are two thresholds for fault detection, χ α 2 (m a ) and χ α 2 (m b ) represent the degrees of freedom m a and m b The chi-square distribution of , α represents the confidence level.

4. The fault diagnosis method according to claim 1, characterized in that: The S3 includes: S31: Select various fault characteristic variable sets J after selection fi , calculate the typical related variables corresponding to each type of fault data; S32: Calculate the fault parameter set based on typical related variables of multiple fault data: R f =[Σ f ,L f ,J f ] (16) Where R f is the fault parameter set of the fth type fault, Σ f ,L f ,J f are the relevant analysis parameters corresponding to various types of fault data, and R f Establish a fault parameter library for various types of faults.

5. The fault diagnosis method according to claim 1, characterized in that: The S4 includes: S41: Select the real-time acquisition data set and calculate the characteristic variables after differential sensitivity selection; S42: Calculate the residual signal statistics based on the obtained characteristic variables and when Greater than the fault detection threshold J th,r1 or Greater than the fault detection threshold J th,r2 , it is determined to be a fault, otherwise it is determined to be normal, that is: S43: When the data is judged to be faulty, the parameters of each type of fault in the fault parameter library are used in turn to calculate the residual signal statistics corresponding to each fault type parameter and 6. The fault diagnosis method according to claim 1, characterized in that: The S5 further includes: S52: Based on the current confidence rate of each fault type, determine whether it is an unknown fault and calculate the predicted classification probability of the data set. The judgment method is as follows: Where G o is the type confidence rate of the oth fault type, o = 1, ..., m, m is the total number of fault types, o ≠ f, β is the confidence threshold, and its value range is (0, 1); S53: Calculate the predicted classification probability when the type of the current data is not an unknown fault. The calculation formula is as follows: Where H f Determine the probability of each type of prediction classification for the current data sample, S54: When the current data type is unknown fault, the typical related parameters of the unknown fault are calculated as: Σ′, L′ and J′, forming a fault parameter set of the new fault type: R′=[Σ′,L′,J′] (20) Where R′ is the fault parameter set of the new fault type. Adding R′ to the fault parameter library increases the total number of fault types.

7. A fault diagnosis system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method steps according to any one of claims 1 to 6 are implemented.

Images