Fault event type identification method and device, server and storage medium

By performing multi-dimensional feature extraction and deep learning model training on fault event signals, the problem of insufficient single-dimensional and local information in the existing technology is solved, and effective identification of fault event types is achieved.

CN119025916AInactive Publication Date: 2024-11-26ZHONGTIAN ELECTRIC POWER OPTICAL CABLES CO LTD +1
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411107149.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, the fault event type identification method only considers the single-dimensional information and local information of the signal, resulting in the inability to effectively identify the fault event type.

Method used

By obtaining the original signal data of the fault event, data processing is performed to extract energy signals of multiple different periodic ratios, multi-dimensional feature data extraction, standard data set is established, and deep learning models are used for training to generate fault event type identification results. Consider multi-dimensional information such as time domain, frequency domain, scale domain, etc. and global and local information.

Benefits of technology

Effectively identifying the fault event type improves the recognition accuracy and considers multi-dimensional and global information, so that the fault type can be judged more accurately.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119025916A_ABST
    Figure CN119025916A_ABST
Patent Text Reader

Abstract

The invention provides a fault event type identification method and device, a server and a storage medium, and the method comprises the steps: carrying out the data processing of the obtained original signal data of any fault event, so as to obtain the processed signal data; identifying a periodic energy signal according to the processed signal data, and dividing the periodic energy signal into a plurality of proportions to obtain a plurality of energy signals with different periodic ratios; performing multi-dimensional feature data extraction on each energy signal with different period ratios to obtain multiple pieces of feature data; establishing a standard data set according to all the feature data corresponding to all the energy signals with different cycle ratios, and dividing the standard data set into a training set and a test set; inputting the test set into a deep learning model trained according to the training set to generate a type recognition result of the fault event; and outputting the type identification result of the fault event, so that the type of the fault event can be effectively identified.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, server and storage medium for identifying the type of a fault event. Background Art

[0002] Pattern recognition algorithm is a method in the industrial field that identifies fault event signals to determine the corresponding type. It can detect fault problems in a timely manner and take corresponding solutions, thereby increasing the effective working time of related equipment.

[0003] In the prior art, the traditional method for identifying the type of fault events mainly uses machine learning technology to identify the types of different fault events to be identified, so as to determine the type of each fault event to be identified.

[0004] However, in the prior art, the traditional method for identifying the type of fault event only considers the single-dimensional information and local information of the corresponding signal, which makes it impossible to effectively identify the type of fault event. Summary of the invention

[0005] The present application provides a method, device, server and storage medium for identifying the type of a fault event to solve the problem that the traditional method for identifying the type of a fault event only considers the single-dimensional information and local information of the corresponding signal, thereby failing to effectively identify the type of the fault event.

[0006] In a first aspect, the present application provides a method for identifying the type of a fault event, which is applied to a server, comprising:

[0007] Acquire original signal data of any fault event, and perform data processing on the original signal data to obtain processed signal data;

[0008] Identifying the periodic energy signal of the fault event according to the processed signal data, and dividing the periodic energy signal into a plurality of ratios to obtain a plurality of energy signals with different period ratios;

[0009] Performing multi-dimensional feature data extraction on each energy signal with different period ratios to obtain multiple feature data;

[0010] Establishing a standard data set according to all characteristic data corresponding to energy signals with different period ratios, and dividing the standard data set into a training set and a test set according to a preset ratio;

[0011] Inputting the test set into the trained deep learning model to generate a type recognition result of the fault event, wherein the trained deep learning model is trained according to the training set;

[0012] Output the type identification result of the fault event.

[0013] In one possible design, the training process of the deep learning model includes: constructing a segmented deep learning network structure, wherein the segmented deep learning network structure includes multiple network structure branches, and the number of the multiple network structure branches corresponds to the number of energy signals with different period ratios; inputting multiple feature data corresponding to the energy signals with different period ratios in the training set into each network structure branch for processing to output corresponding standard cross entropy loss functions; controlling each network structure branch to respectively identify the energy signals with different period ratios to determine the preset weights corresponding to each network structure branch, wherein the preset weights correspond to the energy signals with different period ratios; determining the total loss function according to the standard cross entropy loss functions and the corresponding preset weights; iterating the total loss function until the total loss function does not change to obtain a trained deep learning model.

[0014] In a possible design, the multiple feature data corresponding to the energy signals of different period ratios in the training set are input into each network structure branch for processing to output corresponding standard cross entropy loss functions, including: inputting the multiple feature data corresponding to the energy signals of different period ratios in the training set into each network structure branch to perform the following processing: predicting the probability of a target fault event type corresponding to the energy signal of the period ratio according to the multiple feature data corresponding to the energy signal of the period ratio in any network structure branch; calculating the cross entropy loss function of the target fault event type corresponding to the energy signal of the period ratio according to the preset classification label corresponding to the target fault event type and the probability of the target fault event type; performing a corresponding function transformation on the cross entropy loss function of the target fault event type to obtain a corresponding standard cross entropy loss function; and controlling each network structure branch to output corresponding standard cross entropy loss functions according to the energy signals of different period ratios in the training set.

[0015] In a possible design, the cross entropy loss function of the target fault event type corresponding to the energy signal of the period ratio is calculated according to the preset classification label corresponding to the target fault event type and the probability of the target fault event type, and the calculation formula is:

[0016]

[0017] Where CrssEntropyLoss is the cross entropy loss function; c is the number of categories; y i is a one-hot encoding of the true label (if the sample belongs to category i, then y i =1, otherwise y i =0); is the probability of the target fault event type corresponding to the i-th label.

[0018] In a possible design, the cross entropy loss function of the target fault event type is subjected to a corresponding function transformation to obtain a corresponding standard cross entropy loss function, and the calculation formula is:

[0019] F = th (CrssEntropyLoss)

[0020] Wherein, F is the standard cross entropy loss function; th is the hyperbolic tangent corresponding to the function transformation; CrssEntropyLoss is the cross entropy loss function.

[0021] In a possible design, the multiple energy signals with different period ratios are energy signals with three different period ratios; accordingly, the total loss function is determined according to the standard cross entropy loss functions and the corresponding preset weights, and the calculation formula is:

[0022] F=λ1F1+λ2F2+λ3F3

[0023] Wherein, F is the total loss function; λ1 is the weight of the energy signal of the first different period ratio; λ2 is the weight of the energy signal of the second different period ratio; λ3 is the weight of the energy signal of the third different period ratio; F1 is the standard cross entropy loss function corresponding to the first branch; F2 is the standard cross entropy loss function corresponding to the second branch; F3 is the standard cross entropy loss function corresponding to the third branch.

[0024] In one possible design, the data processing of the original data to obtain processed signal data includes: performing abnormality analysis processing on the original data to obtain signal data after preliminary processing; and filtering processing on the signal data after preliminary processing to obtain processed signal data.

[0025] In one possible design, a standard data set is established based on all the characteristic data corresponding to the energy signals of all different period ratios, including: normalizing all the characteristic data corresponding to the energy signals of all different period ratios to obtain all the normalized characteristic data; and establishing a standard data set based on all the normalized characteristic data.

[0026] In a second aspect, the present application provides a device for identifying the type of a fault event, applied to a server, comprising:

[0027] A processing module, used for acquiring original signal data of any fault event, and performing data processing on the original signal data to obtain processed signal data;

[0028] A division module, used for identifying the periodic energy signal of the fault event according to the processed signal data, and dividing the periodic energy signal into a plurality of ratios to obtain energy signals with a plurality of different period ratios;

[0029] An extraction module, used for performing multi-dimensional feature data extraction on each energy signal with different period ratios to obtain multiple feature data;

[0030] An establishment module is used to establish a standard data set according to all characteristic data corresponding to energy signals with different period ratios, and divide the standard data set into a training set and a test set according to a preset ratio;

[0031] A first input module is used to input a test set into a trained deep learning model to generate a type recognition result of the fault event, wherein the trained deep learning model is trained according to the training set;

[0032] The output module is used to output the type identification result of the fault event.

[0033] In a third aspect, the present application provides a server, comprising: at least one processor and a memory;

[0034] The memory stores computer-executable instructions;

[0035] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method for identifying the type of fault event as described in the first aspect and various possible designs of the first aspect.

[0036] In a fourth aspect, the present application provides a computer storage medium storing computer execution instructions. When a processor executes the computer execution instructions, the method for identifying the type of fault events as described in the first aspect and various possible designs of the first aspect is implemented.

[0037] In a fifth aspect, an embodiment of the present disclosure provides a computer program product, including a computer program, which, when executed by a processor, implements the method for identifying the type of fault events as described in the first aspect and various possible designs of the first aspect.

[0038] The method, device, server and storage medium for identifying the type of fault event provided in the present application process the acquired original signal data of the fault event to obtain processed signal data; identify the periodic energy signal of the fault event according to the processed signal data, and divide the periodic energy signal into multiple energy signals with different period ratios, so as to extract feature data of each energy signal with different period ratios to obtain multiple feature data; establish a standard data set according to all the feature data, and divide the standard data set into a training set and a test set according to a preset ratio; input the test set into a deep learning model trained according to the training set to generate a type identification result of the fault event; output the type identification result of the fault event, so that multi-dimensional information such as time domain, frequency domain, scale domain, as well as global and local information are considered, so that the type of fault event can be effectively identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0040] Figure 1 A schematic diagram of an application scenario of the method for identifying the type of fault event provided in an embodiment of the present application;

[0041] Figure 2 Schematic diagram of the method for identifying the type of fault event provided in the embodiment of the present application Figure 1 ;

[0042] Figure 3 Schematic diagram of the method for identifying the type of fault event provided in the embodiment of the present application Figure 2 ;

[0043] Figure 4 A schematic diagram of the structure of a device for identifying the type of a fault event provided in an embodiment of the present application;

[0044] Figure 5 A schematic diagram of the hardware structure of the server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0046] Pattern recognition algorithm is a technology that uses computers and mathematical techniques to study the automatic processing and interpretation of patterns. It enables the machine to have automatic recognition capabilities through specific algorithms, and can classify specific samples into a certain pattern without manual observation and analysis. It is also a method in the industrial field to identify fault event signals to determine the corresponding type. It can promptly discover fault problems and take corresponding solutions, thereby increasing the effective working time of related equipment. In the prior art, the traditional method for identifying the type of fault events mainly uses machine learning technology to identify the types of different fault events to be identified to determine the type of each fault event to be identified. However, in the prior art, the traditional method for identifying the type of fault events only considers the single-dimensional information and local information of the corresponding signal, which makes it impossible to effectively identify the type of fault event.

[0047] In order to solve the above technical problems, the embodiments of the present application propose the following technical concept: the inventors take into account the shortcomings of the traditional fault event type identification method, obtain multiple feature data corresponding to the energy signals with different period ratios based on the signal data of any fault event, and use the trained deep learning model to process the multiple feature data corresponding to the energy signals with different period ratios to obtain the type identification result of the fault event, so that multi-dimensional information is taken into account, as well as global and local information, so that the type of fault event can be effectively identified.

[0048] Figure 1 It is a schematic diagram of an application scenario of the method for identifying the type of fault events provided in an embodiment of the present application.

[0049] like Figure 1 As shown, the scenario includes: a server 101 and a display terminal 102 .

[0050] The server 101 may be an independent server or a cluster composed of multiple servers.

[0051] The display terminal 102 may be a display screen or a terminal such as a personal computer.

[0052] The server 101 obtains the original signal data of any fault event, and performs periodic energy signal recognition on the processed signal data obtained by data processing the original signal data; divides the periodic energy signal into multiple ratios, and extracts multi-dimensional feature data of each energy signal with different period ratios to obtain multiple feature data; establishes a standard data set based on all feature data, and divides the standard data set into a training set and a test set; inputs the test set into a deep learning model trained according to the training set to generate a type recognition result of the fault event; and outputs the type recognition result of the fault event to the display terminal 102 via a wireless network. The following is a detailed description using a detailed embodiment.

[0053] Figure 2 Schematic diagram of the method for identifying the type of fault event provided in the embodiment of the present application Figure 1 , the execution subject of this embodiment can be Figure 1 The server in the embodiment shown is not particularly limited in this embodiment. Figure 2 As shown, the method includes:

[0054] S201: Acquire original signal data of any fault event, and perform data processing on the original signal data to obtain processed signal data.

[0055] Specifically, step S201 includes:

[0056] S2011: Perform abnormal analysis on the original data to obtain signal data after preliminary processing.

[0057] In this embodiment, the original data may also be data of different types.

[0058] In this embodiment, the abnormal analysis process is used to eliminate data such as incomplete and empty data.

[0059] S2012: Filter the signal data after preliminary processing to obtain processed signal data.

[0060] In this embodiment, filtering is used to eliminate interference signals.

[0061] S202: Identify the periodic energy signal of the fault event according to the processed signal data, and divide the periodic energy signal into multiple ratios to obtain multiple energy signals with different period ratios.

[0062] Specifically, step S202 is specifically: identifying the periodic energy signal of the fault event according to the processed signal data using an energy method.

[0063] In this embodiment, the multiple energy signals with different cycle ratios may be any number of 3, 5 or 7 energy signals with different cycle ratios.

[0064] Exemplarily, the energy signals with multiple different cycle ratios can be energy signals with 5 different cycle ratios; the energy signals with 5 different cycle ratios are: A10, A25, A50, A75 and A100, and their corresponding cycle ratios are: 10%, 25%, 50%, 75% and 100% respectively.

[0065] S203: Perform multi-dimensional feature data extraction on each energy signal with different period ratios to obtain multiple feature data.

[0066] In this embodiment, the multiple dimensions may cover dimensions such as time domain, frequency domain and scale.

[0067] In this embodiment, the multiple characteristic data can be short-time average energy, short-time zero-crossing rate, duration, short-time average amplitude, short-time level crossing rate, short-time double-level crossing rate, power spectrum energy and frequency band energy variance, etc.

[0068] S204: Establish a standard data set according to all feature data corresponding to energy signals with different period ratios, and divide the standard data set into a training set and a test set according to a preset ratio.

[0069] In this embodiment, the preset ratio may be any ratio of 7:3, 8:2 or 9:1, or other ratios.

[0070] Specifically, step S204 includes:

[0071] S2041: normalizing all characteristic data corresponding to energy signals with different period ratios to obtain all normalized characteristic data.

[0072] S2042: Establish a standard data set based on all normalized feature data.

[0073] In this embodiment, corresponding labels may also be added to the standard data set.

[0074] Exemplarily, the sub-tags contained in the tag event A are the above-mentioned A10, A25, A50, A75 and A100.

[0075] S205: Inputting the test set into the trained deep learning model to generate a type recognition result of the fault event, wherein the trained deep learning model is trained based on the training set.

[0076] S206: Output the type identification result of the fault event.

[0077] In summary, the method for identifying the type of fault event provided in the present embodiment processes the acquired original signal data of the fault event to obtain processed signal data; identifies the periodic energy signal of the fault event based on the processed signal data, and divides the periodic energy signal into multiple energy signals with different period ratios, so as to extract feature data for each energy signal with different period ratios to obtain multiple feature data; establishes a standard data set based on all the feature data, and divides the standard data set into a training set and a test set according to a preset ratio; inputs the test set into a deep learning model trained according to the training set to generate a type identification result of the fault event; outputs the type identification result of the fault event, so that multi-dimensional information such as time domain, frequency domain, scale domain, as well as global and local information are considered, so that the type of fault event can be effectively identified.

[0078] In addition, by normalizing all characteristic data corresponding to energy signals of all different period ratios, subsequent processing and calculation speeds can be accelerated, and the correlation between characteristic parameters corresponding to energy signals of different period ratios can be better compared.

[0079] Figure 3 Schematic diagram of the method for identifying the type of fault event provided in the embodiment of the present application Figure 2 In the embodiment of the present application, Figure 2 Based on the embodiment provided, the specific implementation method of the training process of the trained deep learning model in step S205 is described in detail. Figure 3 As shown, the method includes:

[0080] S301: Construct a segmented deep learning network structure, wherein the segmented deep learning network structure includes multiple network structure branches, and the number of the multiple network structure branches corresponds to the number of energy signals with different period ratios.

[0081] In this embodiment, the segmented deep learning network structure can be a 1DCNN deep learning network structure.

[0082] In this embodiment, the discussion on multiple energy signals with different period ratios has been described in detail in step S202, which will not be repeated here.

[0083] S302: Input multiple feature data corresponding to energy signals of different period ratios in the training set into each network structure branch for processing to output corresponding standard cross entropy loss functions.

[0084] Specifically, step S302 is as follows:

[0085] S3021: Input multiple feature data corresponding to energy signals of different period ratios in the training set into each network structure branch to perform the following processing:

[0086] S30211: Predict the probability of a target fault event type corresponding to an energy signal of a cycle ratio in any network structure branch based on a plurality of feature data corresponding to the energy signal of the cycle ratio.

[0087] In this embodiment, the structural composition of any network structure branch specifically includes: convolution layer, weight layer, normalization layer, weight layer, normalization layer, ReLU, fully connected layer, DropOUT layer, etc.

[0088] Among them, the weight layer and the normalization layer uniformly normalize the feature data to facilitate model calculation; ReLU is used to make the matrix sparse to prevent gradient disappearance and increase the nonlinear expression ability of the fully connected layer model; the DropOUT layer is used to reduce the probability of overfitting.

[0089] S30212: Calculate the cross entropy loss function of the target fault event type corresponding to the energy signal of the cycle ratio according to the preset classification label corresponding to the target fault event type and the probability of the target fault event type.

[0090] In this embodiment, according to the preset classification label corresponding to the target fault event type and the probability of the target fault event type, the cross entropy loss function of the energy signal of the cycle ratio corresponding to the target fault event type is calculated, and the calculation formula is:

[0091]

[0092] Where CrssEntropyLoss is the cross entropy loss function; c is the number of categories; y i is a one-hot encoding of the true label (if the sample belongs to category i, then y i =1, otherwise y i =0); is the probability of the target fault event type corresponding to the i-th label.

[0093] S30213: Perform a corresponding function transformation on the cross entropy loss function of the target fault event type to obtain a corresponding standard cross entropy loss function.

[0094] In this embodiment, the cross entropy loss function of the target fault event type is transformed into a corresponding function to obtain a corresponding standard cross entropy loss function, and the calculation formula is:

[0095] F = th (CrssEntropyLoss)

[0096] Where F is the standard cross entropy loss function; th is the hyperbolic tangent corresponding to the function transformation; CrssEntropyLoss is the cross entropy loss function.

[0097] S30214: According to the energy signals of different period ratios in the training set, the standard cross entropy loss functions corresponding to the outputs of the network structure branches are controlled.

[0098] Exemplarily, the energy signals with different period ratios are energy signals with three different period ratios, and their corresponding standard cross entropy loss functions are F1, F2 and F3.

[0099] S303: Control each network structure branch to respectively identify energy signals with different period ratios to determine each preset weight corresponding to each network structure branch, wherein each preset weight corresponds to an energy signal with different period ratio.

[0100] S304: Determine a total loss function according to each standard cross entropy loss function and each corresponding preset weight.

[0101] In this embodiment, the energy signals with multiple different period ratios are energy signals with three different period ratios; accordingly, the total loss function is determined according to each standard cross entropy loss function and each corresponding preset weight, and the calculation formula is:

[0102] F=λ1F1+λ2F2+λ3F3

[0103] Wherein, F is the total loss function; λ1 is the weight of the energy signal of the first different period ratio; λ2 is the weight of the energy signal of the second different period ratio; λ3 is the weight of the energy signal of the third different period ratio; F1 is the standard cross entropy loss function corresponding to the first branch; F2 is the standard cross entropy loss function corresponding to the second branch; F3 is the standard cross entropy loss function corresponding to the third branch.

[0104] S305: Iterate the total loss function until the total loss function does not change, so as to obtain a trained deep learning model.

[0105] In this embodiment, when the total loss function does not change, the deep learning model converges, that is, it is a trained deep learning model.

[0106] In summary, the method for identifying the type of fault events provided in this embodiment constructs a segmented deep learning network structure, which includes a plurality of network structure branches corresponding to the number of energy signals with different period ratios; inputs a plurality of feature data corresponding to the energy signals with different period ratios in the training set into each network structure branch for processing to output corresponding standard cross entropy loss functions; controls each network structure branch to determine the corresponding preset weights respectively; determines the total loss function according to each standard cross entropy loss function and the corresponding preset weights; iterates the total loss function until the total loss function does not change, so as to obtain a trained deep learning model, which takes into account multi-dimensional information such as time domain, frequency domain, scale domain, as well as global and local information, so that the type of fault event can be effectively identified.

[0107] Figure 4 This is a schematic diagram of the structure of a device for identifying the type of fault event provided in an embodiment of the present application. Figure 4 As shown, the fault event type identification device includes: a processing module 401 , a classification module 402 , an extraction module 403 , a creation module 404 , a first input module 405 and an output module 406 .

[0108] The processing module 401 is used to obtain the original signal data of any fault event and perform data processing on the original signal data to obtain processed signal data;

[0109] A division module 402, configured to identify the periodic energy signal of the fault event according to the processed signal data, and divide the periodic energy signal into a plurality of ratios to obtain energy signals with a plurality of different period ratios;

[0110] An extraction module 403 is used to extract multi-dimensional feature data from each energy signal with different period ratios to obtain multiple feature data;

[0111] Establishing module 404, used to establish a standard data set according to all characteristic data corresponding to energy signals with different period ratios, and divide the standard data set into a training set and a test set according to a preset ratio;

[0112] A first input module 405 is used to input the test set into the trained deep learning model to generate a type recognition result of the fault event, wherein the trained deep learning model is trained according to the training set;

[0113] The output module 406 is used to output the type identification result of the fault event.

[0114] In a possible implementation manner, the device further includes:

[0115] A construction module 407 is used to construct a segmented deep learning network structure, wherein the segmented deep learning network structure includes a plurality of network structure branches, and the number of the plurality of network structure branches corresponds to the number of the plurality of energy signals with different period ratios;

[0116] The second input module 408 is used to input a plurality of feature data corresponding to energy signals of different period ratios in the training set into each network structure branch for processing, so as to output corresponding standard cross entropy loss functions;

[0117] A control module 409, configured to control each network structure branch to respectively identify energy signals of different period ratios, so as to determine each preset weight corresponding to each network structure branch, wherein each preset weight corresponds to the energy signal of different period ratios;

[0118] A determination module 410 is used to determine a total loss function according to each of the standard cross entropy loss functions and the corresponding preset weights;

[0119] The loop module 411 is used to iterate the total loss function until the total loss function does not change, so as to obtain a trained deep learning model.

[0120] In a possible implementation, the second input module 408 specifically includes:

[0121] The input unit 4081 is used to input a plurality of feature data corresponding to energy signals of different period ratios in the training set into each network structure branch to perform the following processing:

[0122] A prediction unit 4082, configured to predict the probability of a target fault event type corresponding to an energy signal of a period ratio in any network structure branch according to a plurality of feature data corresponding to the energy signal of the period ratio;

[0123] A calculation unit 4083 is used to calculate the cross entropy loss function of the energy signal of the period ratio corresponding to the target fault event type according to the preset classification label corresponding to the target fault event type and the probability of the target fault event type;

[0124] A transformation unit 4084 is used to perform a corresponding function transformation on the cross entropy loss function of the target fault event type to obtain a corresponding standard cross entropy loss function;

[0125] The control unit 4085 is used to control the standard cross entropy loss functions corresponding to the outputs of the network structure branches according to the energy signals of different period ratios in the training set.

[0126] In a possible implementation, the cross entropy loss function of the target fault event type corresponding to the energy signal of the period ratio is calculated according to the preset classification label corresponding to the target fault event type and the probability of the target fault event type, and the calculation formula is:

[0127]

[0128] Where CrssEntropyLoss is the cross entropy loss function; c is the number of categories; y i is a one-hot encoding of the true label (if the sample belongs to category i, then y i =1, otherwise y i =0); is the probability of the target fault event type corresponding to the i-th label.

[0129] In a possible implementation, the cross entropy loss function of the target fault event type is subjected to a corresponding function transformation to obtain a corresponding standard cross entropy loss function, and the calculation formula is:

[0130] F = th (CrssEntropyLoss)

[0131] Wherein, F is the standard cross entropy loss function; th is the hyperbolic tangent corresponding to the function transformation; CrssEntropyLoss is the cross entropy loss function.

[0132] In a possible implementation, the multiple energy signals with different period ratios are energy signals with three different period ratios; accordingly, the total loss function is determined according to the standard cross entropy loss functions and the corresponding preset weights, and the calculation formula is:

[0133] F=λ1F1+λ2F2+λ3F3

[0134] Wherein, F is the total loss function; λ1 is the weight of the energy signal of the first different period ratio; λ2 is the weight of the energy signal of the second different period ratio; λ3 is the weight of the energy signal of the third different period ratio; F1 is the standard cross entropy loss function corresponding to the first branch; F2 is the standard cross entropy loss function corresponding to the second branch; F3 is the standard cross entropy loss function corresponding to the third branch.

[0135] In a possible implementation, the processing module 401 specifically includes:

[0136] The first processing unit 4011 is used to perform abnormal analysis on the original data to obtain signal data after preliminary processing;

[0137] The second processing unit 4012 is used to perform filtering processing on the signal data after the preliminary processing to obtain processed signal data.

[0138] In a possible implementation, the establishing module 404 specifically includes:

[0139] The processing unit 4041 is used to perform normalization processing on all characteristic data corresponding to the energy signals of all different period ratios to obtain all normalized characteristic data;

[0140] The establishing unit 4042 is used to establish a standard data set according to all the normalized feature data.

[0141] The device provided in this embodiment can be used to execute the technical solution of the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0142] Figure 5 The hardware structure diagram of the server provided in the embodiment of the present application is shown in FIG. Figure 5 As shown, the server of this embodiment includes: a processor 501 and a memory 502; the memory stores computer execution instructions; at least one processor executes the computer execution instructions stored in the memory, so that at least one processor executes the above fault event type identification method.

[0143] Optionally, the memory 502 may be independent or integrated with the processor 501 .

[0144] When the memory 502 is independently provided, the server further includes a bus 503 for connecting the memory 502 and the processor 501 .

[0145] An embodiment of the present application further provides a computer storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, the method for identifying the type of a fault event as described above is implemented.

[0146] An embodiment of the present application further provides a computer program product, including a computer program, and when the computer program is executed by a processor, the method for identifying the type of a fault event as described above is implemented.

[0147] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0148] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to implement the solution of this embodiment.

[0149] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The above-mentioned module-composed unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0150] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the method described in each embodiment of the present application.

[0151] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be implemented by a combination of hardware and software modules in the processor.

[0152] The memory may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, etc.

[0153] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application is not limited to only one bus or one type of bus.

[0154] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0155] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic device or a main control device.

[0156] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for identifying the type of a fault event, characterized in that: Applicable to servers, including: Acquire original signal data of any fault event, and perform data processing on the original signal data to obtain processed signal data; Identifying the periodic energy signal of the fault event according to the processed signal data, and dividing the periodic energy signal into a plurality of ratios to obtain a plurality of energy signals with different period ratios; Performing multi-dimensional feature data extraction on each energy signal with different period ratios to obtain multiple feature data; Establishing a standard data set according to all characteristic data corresponding to energy signals with different period ratios, and dividing the standard data set into a training set and a test set according to a preset ratio; Inputting the test set into the trained deep learning model to generate a type recognition result of the fault event, wherein the trained deep learning model is trained according to the training set; Output the type identification result of the fault event.

2. The method according to claim 1, characterized in that The training process of the deep learning model includes: Constructing a segmented deep learning network structure, wherein the segmented deep learning network structure includes a plurality of network structure branches, and the number of the plurality of network structure branches corresponds to the number of the plurality of energy signals with different period ratios; Inputting a plurality of feature data corresponding to energy signals of different period ratios in the training set into each network structure branch for processing to output corresponding standard cross entropy loss functions; Controlling each network structure branch to respectively identify energy signals with different period ratios to determine each preset weight corresponding to each network structure branch, wherein each preset weight corresponds to the energy signal with different period ratios; Determine a total loss function according to each of the standard cross entropy loss functions and the corresponding preset weights; The total loss function is iterated and looped until the total loss function does not change, so as to obtain a trained deep learning model.

3. The method according to claim 2, characterized in that The inputting of multiple feature data corresponding to energy signals of different period ratios in the training set into each network structure branch for processing to output corresponding standard cross entropy loss functions includes: The multiple feature data corresponding to the energy signals of different period ratios in the training set are input into each network structure branch to perform the following processing: Predicting the probability of a target fault event type corresponding to an energy signal of a period ratio in any network structure branch according to a plurality of characteristic data corresponding to the energy signal of the period ratio; Calculate the cross entropy loss function of the target fault event type corresponding to the energy signal of the period ratio according to the classification label corresponding to the preset target fault event type and the probability of the target fault event type; Performing a corresponding function transformation on the cross entropy loss function of the target fault event type to obtain a corresponding standard cross entropy loss function; According to the energy signals of different period ratios in the training set, the standard cross entropy loss functions corresponding to the outputs of the network structure branches are controlled.

4. The method according to claim 3, characterized in that The cross entropy loss function of the target fault event type corresponding to the energy signal of the period ratio is calculated according to the preset classification label corresponding to the target fault event type and the probability of the target fault event type. The calculation formula is: Where CrssEntropyLoss is the cross entropy loss function; c is the number of categories; y i is a one-hot encoding of the true label (if the sample belongs to category i, then y i =1, otherwise y i =0); is the probability of the target fault event type corresponding to the i-th label.

5. The method according to claim 4, characterized in that The cross entropy loss function of the target fault event type is subjected to a corresponding function transformation to obtain a corresponding standard cross entropy loss function, and the calculation formula is: F = th (CrssEntropyLoss) Wherein, F is the standard cross entropy loss function; th is the hyperbolic tangent corresponding to the function transformation; CrssEntropyLoss is the cross entropy loss function.

6. The method according to claim 2, characterized in that wherein the multiple energy signals with different period ratios are energy signals with three different period ratios; Accordingly, the total loss function is determined according to the standard cross entropy loss functions and the corresponding preset weights, and the calculation formula is: F=λ1F1+λ2F2+λ3F3 Where F is the total loss function; λ1 is the weight of the energy signal of the first different period ratio; λ2 is the weight of the energy signal of the second different period ratio; λ3 is the weight of the energy signal of the third different period ratio; F1 is the standard cross entropy loss function corresponding to the first branch; F2 is the standard cross entropy loss function corresponding to the second branch; F3 is the standard cross entropy loss function corresponding to the third branch.

7. The method according to claim 1, characterized in that The processing of the original data to obtain processed signal data includes: Performing abnormal analysis on the raw data to obtain signal data after preliminary processing; The signal data after the preliminary processing is filtered to obtain processed signal data.

8. The method according to any one of claims 1 to 7, characterized in that: The standard data set is established according to all characteristic data corresponding to energy signals with different period ratios, including: Normalizing all characteristic data corresponding to the energy signals of all different period ratios to obtain all normalized characteristic data; A standard data set is established based on all the normalized feature data.

9. A device for identifying the type of a fault event, characterized in that: Applicable to servers, including: A processing module, used for acquiring original signal data of any fault event, and performing data processing on the original signal data to obtain processed signal data; A division module, used for identifying the periodic energy signal of the fault event according to the processed signal data, and dividing the periodic energy signal into a plurality of ratios to obtain energy signals with a plurality of different period ratios; An extraction module, used for performing multi-dimensional feature data extraction on each energy signal with different period ratios to obtain multiple feature data; An establishment module is used to establish a standard data set according to all characteristic data corresponding to energy signals with different period ratios, and divide the standard data set into a training set and a test set according to a preset ratio; A first input module is used to input a test set into a trained deep learning model to generate a type recognition result of the fault event, wherein the trained deep learning model is trained according to the training set; The output module is used to output the type identification result of the fault event.

10. A server, characterized in that: include: at least one processor and memory; The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method for identifying the type of a fault event according to any one of claims 1 to 8.

11. A computer storage medium, characterized in that: The computer storage medium stores computer executable instructions, and when the processor executes the computer executable instructions, the method for identifying the type of a fault event as described in any one of claims 1 to 8 is implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for identifying the type of a fault event according to any one of claims 1 to 8 is implemented.

Citation Information

Patent Citations

  • Rotating machinery fault diagnosis and early warning method

    CN109506921A

  • Vein identification method based on time domain short-time and long-time feature fusion

    CN113111744A

  • Image question and answer data acquisition method based on visual language large model

    CN117972053A

  • Battery fault prediction model training method and device, and prediction method and device

    CN118395370A

  • Method and apparatus for detecting electroencephalogram characteristic waveform, and storage medium

    WO2024113226A1