Generalized zero sample fault diagnosis method for thermal power generation feed pump set based on fault similarity

Abstract

The invention discloses a thermal power generation feed pump set generalized zero sample fault diagnosis method based on fault similarity, and belongs to the field of thermal power generation feed pump set generalized zero sample fault diagnosis. The method comprises the following steps: constructing a feature extraction model of a fault-related stacked noise reduction auto-encoder, and extracting fault-related features from historical production data; according to shallow expert knowledge, a similarity matrix between the seen faults and a similarity matrix between the seen faults and unseen faults are constructed. Calculating a similarity matrix among the seen faults by using the data of the seen faults, and forming a similarity matrix mapping relation between the data and expert knowledge in combination with shallow expert experience; and on the basis of the off-line model, the seen faults and the non-seen faults of the water feeding pump set are diagnosed in real time. The fault type of the water feeding pump set can be diagnosed in real time, the defects of dependence on deep knowledge such as fault reasons, positions and degrees and domain offset are overcome, and normal operation of the water feeding pump set in thermal power generation is guaranteed.

Description

Technical Field

[0001] The invention belongs to the field of generalized zero-sample fault diagnosis of a thermal power generation water supply pump group and provides an effective zero-sample fault diagnosis method. Background Art

[0002] In the global energy structure, thermal power generation is an important part of power supply. As an efficient and economical way to produce electricity, coal-fired power plants are widely used in large-scale power supply. However, the efficient operation of thermal power plants not only depends on the coordinated work of major equipment such as boilers, steam turbines and generators, but also places the same requirements on the stability and reliability of auxiliary equipment. As one of the core auxiliary equipment of thermal power plants, feedwater pump sets play an indispensable role in the entire power generation system.

[0003] The main function of the feedwater pump group is to transport condensed water back to the boiler, ensure smooth water circulation in the boiler, and maintain the thermal balance and operating efficiency of the system. Its working performance directly affects the steam generation rate of the boiler, the power output of the steam turbine, and the economy and safety of the entire power plant. If the feedwater pump group fails, condensed water may not be supplied to the boiler in time, causing a series of serious problems such as dry burning of the boiler and pressure loss. This will not only cause damage to the equipment, but may also cause the power plant to shut down, thereby affecting power supply and economic benefits, and may even cause environmental pollution and safety accidents.

[0004] In actual production, fault data is usually stored in the historical production database, and the fault type can be judged based on the operator's experience and workshop production records, which is called a seen fault. However, as time goes by, the equipment gradually ages, many parameters change, and some foreseeable new faults may be encountered, or even multiple faults may occur at the same time. These faults without data records are called unseen faults. Since these faults are not detected in advance and not recorded in the historical database, they often cannot be diagnosed in time when they occur. Therefore, the rapid identification and diagnosis of seen faults and unseen faults constitutes a classic industrial zero-sample fault diagnosis problem. Solving this problem can significantly improve the production efficiency of enterprises, reduce unnecessary economic losses, and even prevent casualties caused by equipment failures. In the context of the continuous development of advanced sensing technology and data processing technology, the introduction of intelligent diagnosis systems will help improve the operating reliability of feedwater pump groups and the overall performance of coal-fired thermal power plants.

[0005] In traditional zero-shot fault diagnosis tasks, the model only needs to have the ability to diagnose unseen faults. However, in actual industrial applications, it is usually expected that the model can effectively diagnose both seen and unseen faults. To solve this problem, scholars have proposed a generalized zero-shot fault diagnosis method. In addition to single faults, in industrial production, compound faults are often faced. Compound faults refer to the simultaneous occurrence of multiple faults in the system, which makes the system behavior more complicated, which brings significant challenges to fault diagnosis. The prediction of compound faults is usually based on single faults that occurred in the past, and its predictability provides the possibility for generalized zero-shot fault diagnosis. At present, the research on generalized zero-shot compound fault diagnosis mainly focuses on the field of bearing faults, where compound faults can be regarded as a linear combination of multiple single faults. Therefore, the auxiliary information of compound faults can be obtained by linearly combining the auxiliary information of multiple single faults. However, existing methods all rely on attributes such as the location, degree and cause of the fault. However, obtaining these attributes requires a lot of professional deep expert knowledge. This deep knowledge involves a deep understanding of the entire industrial production process, including complex workflows, equipment characteristics, etc. Obtaining this deep knowledge is often time-consuming and laborious. In contrast, operators with general experience are more likely to provide easily accessible shallow knowledge based on data records, basic understanding of process mechanisms, and communication with experts. Based on this shallow knowledge, a fault similarity matrix is ​​constructed to calculate the membership of each sample relative to seen and unseen faults to diagnose unseen faults. To this end, the present invention proposes a generalized zero-sample composite fault diagnosis method for water supply pump groups based on fault similarity. Summary of the invention

[0006] The present invention aims to solve the technical problems existing in the prior art such as low fault diagnosis efficiency and poor accuracy of diagnostic results, and provides a generalized zero-sample fault diagnosis method for thermal power feedwater pump groups based on fault similarity. This method uses the production data of historical faults to construct a fault-related stacked denoising autoencoder model, which can remove redundant information in the production data, improve the resistance to noise, and improve the accuracy of fault recognition. At the same time, a fault similarity matrix construction method based on a combination of shallow expert experience and data-driven is proposed to effectively identify seen faults and unseen faults, and implement real-time diagnosis of unseen faults, which can effectively improve the accuracy of diagnosis.

[0007] In order to achieve the above purpose, the technical solution adopted by the present invention is: A generalized zero-sample fault diagnosis method for a thermal power feedwater pump group based on fault similarity includes offline establishment of a generalized zero-sample composite fault diagnosis model for a feedwater pump group and online diagnosis based on the generalized zero-sample composite fault diagnosis model for a feedwater pump group; based on historical data, a fault-related stacked denoising autoencoder is constructed to extract fault-related features for preliminary classification; a fault similarity matrix is ​​constructed using shallow expert experience, and the similarity matrix between seen faults is calculated in combination with historical production data to establish a mapping relationship between data-driven and expert experience. A gating mechanism is used to preliminarily distinguish seen faults from unseen faults. Unseen faults are further distinguished using mapping relationships and shallow expert experience. Through the above steps, seen faults and unseen faults in a feedwater pump group are identified in real time, and real-time and accurate diagnosis of generalized zero-sample faults of a feedwater pump group is achieved.

[0008] The specific method is: Step 1) Extract fault-related features from historical production data by building a feature extraction model of fault-related stacked denoising autoencoders; The fault-related stacked denoising autoencoder is used to extract valuable information and eliminate redundant information, enhance the model's resistance to noise interference, accurately identify and separate key feature information, and complete the distinction of fault types; the model training is divided into two stages.

[0009] Phase 1: Layered pre-training; training multiple basic denoising autoencoders one by one; The construction process of the first denoising autoencoder is divided into the following 5 steps: Step 1: Build the input layer; define the number of input layer nodes to be consistent with the original data dimension, and receive training samples of historical production data; Step 2: Add a corruption layer; perform random corruption on the input data, including randomly setting zeros or adding Gaussian noise to generate noisy input data; Step 3: Construct the encoding layer; after the destruction layer, an encoding layer is constructed to map the noisy input into a low-dimensional feature representation; the encoding layer consists of multiple neural network layers, including a linear layer, an activation function layer, and a possible regularization layer; Step 4: Construct the decoding layer; input the low-dimensional feature representation obtained in the previous step into the decoding layer, and restore the low-dimensional features to an output close to the original input data. The number of output nodes is consistent with the input layer; the structure of the encoding layer is opposite to that of the decoding layer, and its purpose is to use low-dimensional features to reconstruct the original clean input data; Step 5: Train the parameters of the autoencoder; take the original uncorrupted data as the reconstruction target, define the mean square error as the loss function, select the Adam optimizer and set the learning rate and batch size hyperparameters, and use the back propagation algorithm to optimize the parameters of the encoder and decoder; The construction process of the second denoising autoencoder is similar to that of the first denoising autoencoder and is divided into the following 5 steps: Step 1: Construct the input layer; fix the parameters of the first denoising autoencoder, and use the low-dimensional features output by its encoding layer as the input data of the second denoising autoencoder; Step 2: Add a corruption layer; inject noise into the input features, the noise type includes random zero or Gaussian noise, to generate noisy features; Step 3: Construct the encoding layer. The construction process is the same as the encoding layer of the first autoencoder. The noisy features are further compressed to obtain lower-dimensional features containing deeper information. Step 4: Construct the decoding layer: The construction process is consistent with the decoding layer of the first denoising autoencoder, reconstructing the lower-dimensional features into the undestructed features of the input layer; Step 5: Train the autoencoder parameters; fix all the parameters of the first denoising autoencoder and only optimize the parameters of the second denoising autoencoder. The loss function, optimizer and hyperparameter settings are the same as those of the first denoising autoencoder. The construction process of the third denoising autoencoder is similar to that of the above denoising autoencoder. The output of the encoding layer of the second denoising autoencoder is used as the input data of the third denoising autoencoder. During training, the parameters of the second denoising autoencoder are fixed, and only the relevant parameters of the current denoising autoencoder are optimized. The second stage is global fine-tuning. After completing the layered pre-training, the stacked denoising autoencoders are globally optimized end-to-end. The input layer, noise layer, and encoding layers of multiple denoising autoencoders obtained by layered pre-training are stacked in sequence to form a feature extraction network. ; The initialization parameters inherit the parameters of the layered pre-training stage and will have N tr Historical production data of samples X tr Input the stacked network, the network encodes the input data layer by layer to generate a high-level feature representation H; this high-level feature representation is used in two parts: The first part: reconstruct the original data; construct a decoding layer, which consists of multiple neural network layers, including a linear layer and an activation function layer, and input the high-level feature representation into the decoding layer; the decoding layer decodes the high-level features and outputs reconstructed data with the same dimension as the original input data ; Calculate the mean square error between the reconstructed data and the original data as the reconstruction loss. The reconstruction loss expression is as follows: ;in and They are X tr and The nth sample of ; The second part: fault classification; establish a SoftMax regression model as a classifier to distinguish between seen and unseen faults and determine the category of a specific seen or unseen fault, map the high-level feature representation to the probability distribution of the fault category and obtain the fault category judged by the classifier as , using cross entropy as the classification loss function, the expression of the cross entropy loss function is as follows: ; where p represents the category, 1{ =p} is the indicator function, that is, if When , its value is 1, otherwise it is 0; Combining the above two parts of loss, the overall loss function of the model is defined as follows: ; Among them, α1 and α2 are weight parameters, and the back propagation algorithm is used to optimize the parameters of the entire network.

[0010] Step 2) Based on shallow expert knowledge, construct a similarity matrix between seen faults and seen faults, and a similarity matrix between seen faults and unseen faults; use seen fault data to calculate the similarity matrix between seen faults, and combine with shallow expert experience to form a similarity matrix mapping relationship between data and expert knowledge; Step 1: Based on the shallow expert experience and historical production data obtained, a similarity matrix between the seen faults and the seen faults of the water supply pump group is constructed: ; and the similarity matrix between seen and unseen faults ; Step 2: Using the historical data of seen faults, by calculating the feature center of each seen fault category, further obtain the similarity matrix between the seen fault categories; first, based on the seen fault samples in the historical production data of the water supply pump group, calculate the feature center of each seen fault category, and the feature center of the pth seen fault is defined as: ;in ∈H tr is the nth sample of the pth type of seen faults, is the number of samples of the p-th type of faults that have been seen, and then the similarity between the feature centers of any two seen fault categories is calculated, which is defined as: ; Finally, the similarity matrix between the seen fault categories calculated based on the seen fault data is obtained: ;

[0011] Step 3: Combined with the similarity matrix obtained from shallow expert experience, the similarity matrix mapping relationship between data and shallow expert knowledge is obtained: ; The similarity diagnosis basis combining data-driven and expert knowledge is realized through mapping relationships, which is used to distinguish and diagnose the seen faults and unseen faults of the water supply pump group in the online stage.

[0012] Step 3) Based on the offline model, diagnose the observed and unseen faults of the water supply pump group in real time.

[0013] 3.1) Input the test sample data monitored online into the pre-trained feature extractor to obtain the feature representation corresponding to the test sample; input the obtained test sample features into the pre-trained classifier to obtain the posterior probability distribution of the test sample belonging to each type of seen fault; 3.2) Compare the maximum probability value of the posterior probability distribution obtained in 3.1) with the set gate threshold to distinguish whether the test sample belongs to a seen fault or an unseen fault; when the maximum posterior probability value is greater than the gate threshold, the test sample is determined to be a seen fault; when the maximum posterior probability value is less than or equal to the gate threshold, the test sample is determined to be an unseen fault; 3.3) For the test sample determined as a seen fault in step 3.2), its fault category is determined to be the seen fault category with the largest posterior probability: ; where g lp is the posterior probability that the lth test sample belongs to the pth type of seen fault; 3.4) For the test samples that are determined to be unseen faults in step 3.2), calculate the similarity between the sample features and the feature centers of each seen fault category obtained in the offline training phase: ;in is the feature of the lth test sample, is the feature center of the p-th type of seen fault; the similarity ω between the current test sample and each seen fault is obtained l =[ ,……, ], using the similarity matrix mapping relationship between offline constructed data and shallow expert knowledge, the mapped The similarity w of the qth type of unseen fault obtained by shallow expert knowledge in the offline stage with respect to each seen fault q ∈W Q,P , q=1,2,……,Q for comparison to determine the unseen fault category to which the current sample belongs: ;

[0014] 3.5) Based on the diagnosis results of step 3.3) and step 3.4), accurate classification and diagnosis of the observed faults and unobserved compound faults of the water supply pump group in the online stage are achieved.

[0015] The beneficial effects of the present invention are: (1) In traditional autoencoder training, the extracted features may not all be beneficial to a specific task, and the extracted features may be redundant. Therefore, the present invention constructs a fault-related stacked denoising autoencoder to extract fault-related features, which significantly enhances the ability to identify fault features in production data, and effectively overcomes the problem of low diagnostic accuracy of traditional models due to data noise interference; (2) Traditional zero-sample fault diagnosis methods often rely on attributes such as the location, degree, and cause of the fault, and have limited ability to identify unseen faults. Therefore, the present invention uses shallow expert knowledge and historical data to construct a fault similarity matrix, and establishes a similarity mapping relationship with expert experience in a data-driven manner, so that the diagnostic model can efficiently and accurately realize real-time diagnosis of unseen faults. The present invention can not only classify single faults, but also accurately classify unseen compound faults, which better meets actual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is the working flow diagram of the water supply pump group; Figure 2 This is the structure diagram of the generalized zero-sample fault diagnosis model for the water supply pump group; Figure 3 Schematic diagram of seen and unseen faults diagnosed based on posterior probability samples; Figure 4 The accuracy graph of each indicator under different numbers of training samples. DETAILED DESCRIPTION

[0017] The following specific embodiment has been put into practical use in a thermal power plant feed water pump workshop and has achieved remarkable results. Figure 1 As shown in the figure, the structure of the generalized zero-sample fault diagnosis model for the water supply pump group is as follows: Figure 2 shown.

[0018] In the offline modeling stage, a fault-related stacked denoising autoencoder feature extraction model is constructed to remove redundant information from the training data, extract fault-related features from historical production data, and use shallow expert knowledge to construct a fault similarity matrix. The similarity matrix between seen faults calculated with available seen fault data is then analyzed to determine the relationship between the similarity matrix provided by experts. In the online diagnosis stage, a gating mechanism is used to divide the test samples into seen faults and unseen faults, and the fault similarity matrix is ​​used to diagnose which fault category the test samples belong to, completing real-time diagnosis of seen and unseen faults. The samples are diagnosed as seen and unseen faults such as Figure 3 shown.

[0019] The offline data classification and recognition algorithm of the present invention is used to model and diagnose the collected data online. In order to prove the effectiveness of the proposed method, the following indicators are defined: G S =N Ts / N S is the accuracy rate of the seen samples being diagnosed as seen faults, G U =N Tu / N u is the accuracy rate of unseen samples being diagnosed as unseen faults, where N s and N u are the number of seen test samples and the number of unseen test samples respectively; N Ts is the number of seen samples diagnosed as seen faults, N Tu is the number of unseen samples diagnosed as unseen faults.

[0020] The accuracy of seen samples and unseen samples that are correctly diagnosed as the corresponding fault type is defined as follows:

[0021] Acc S and Acc U Respectively represent the accuracy of seen samples and unseen samples that are correctly diagnosed as the corresponding fault type, N Cs and N Cu They are respectively represented as the number of correctly diagnosed seen samples and unseen samples. Two comprehensive indicators are used to evaluate the performance of the model:

[0022] Among them, G H is the comprehensive accuracy of gated diagnosis for seen and unseen faults, Acc H It indicates the accuracy of further diagnosis of specific fault types based on the gated diagnosis of seen faults and unseen faults. Figure 4 For a fixed number of 1800 test samples, the accuracy of each indicator under different numbers of training samples shows the impact of different numbers of training samples on model performance. The total number of different training samples is 800, 1600, 2400 and 3200 respectively. It can be seen that the performance indicators of the diagnosis model usually improve with the increase of the number of training samples. It shows that more training samples provide richer information about the seen faults, which helps to improve the performance of the proposed model.

[0023] In order to compare the fault diagnosis level of the model, the model of this embodiment is compared with some common zero-sample fault diagnosis methods, mainly including direct attribute prediction DAP, attribute label embedding ALE and SSHTN. DAP and ALE are traditional zero-sample learning methods, which do not include seen categories in the test phase, while SSHTN includes both seen and unseen categories in the test phase. This embodiment compares the fault diagnosis performance of these three models through the multiple evaluation indicators proposed above.

[0024]

[0025] As can be seen from the above table, the model proposed in the present invention shows performance that exceeds other existing methods in various evaluation indicators. Among them, the DAP and ALE methods are mainly used to solve the zero-sample learning problem, which may lead to domain shift problems. SSHTN solves the generalized zero-sample learning problem, which not only alleviates the domain shift problem, but also improves the accuracy. However, the above methods all rely on deep knowledge-based attribute information. The FSGZSCFD model proposed in the present invention not only effectively alleviates the domain shift problem, but also uses shallow knowledge to obtain the fault similarity matrix for generalized zero-sample fault diagnosis, and can still maintain high prediction accuracy and reliability when the information may be incomplete or difficult to obtain.

Claims

1. A generalized zero-sample fault diagnosis method for a thermal power feedwater pump group based on fault similarity, characterized in that: The steps are: Step 1) Extract fault-related features from historical production data by building a feature extraction model of fault-related stacked denoising autoencoders; Step 2) Based on the shallow expert knowledge, a similarity matrix between seen faults and seen faults, and a similarity matrix between seen faults and unseen faults are constructed; Use the seen fault data to calculate the similarity matrix between the seen faults, and combine it with shallow expert experience to form a similarity matrix mapping relationship between data and expert knowledge; Step 3) Based on the offline model, diagnose the observed and unseen faults of the water supply pump group in real time.

2. The generalized zero-sample fault diagnosis method for a thermal power generation water supply pump group based on fault similarity according to claim 1 is characterized by: In the step 1), the specific scheme is: using the fault-related stacked denoising autoencoder to extract valuable information and eliminate redundant information, enhance the model's resistance to noise interference, identify and separate key feature information, and complete the distinction of fault types; the training of the model is divided into two stages.

3. The generalized zero-sample fault diagnosis method for a thermal power feedwater pump group based on fault similarity according to claim 2 is characterized in that: The two stages of training the model are specifically: Phase 1: Layered pre-training; training multiple basic denoising autoencoders one by one; The construction process of the first denoising autoencoder is divided into the following 5 steps: Step 1: Build the input layer; define the number of input layer nodes to be consistent with the original data dimension, and receive training samples of historical production data; Step 2: Add a corruption layer; perform random corruption on the input data, including randomly setting zeros or adding Gaussian noise to generate noisy input data; Step 3: Construct the encoding layer; An encoding layer is constructed after the destruction layer to map the noisy input into a low-dimensional feature representation; The encoding layer consists of multiple neural network layers, including linear layers, activation function layers, and regularization layers; Step 4: Construct the decoding layer; input the low-dimensional feature representation obtained in the previous step into the decoding layer, and restore the low-dimensional features to an output close to the original input data. The number of output nodes is consistent with the input layer; the structure of the encoding layer is opposite to that of the decoding layer, and the low-dimensional features are used to reconstruct the original clean input data; Step 5: Train the parameters of the autoencoder; take the original uncorrupted data as the reconstruction target, define the mean square error as the loss function, select the Adam optimizer and set the learning rate and batch size hyperparameters, and use the back propagation algorithm to optimize the parameters of the encoder and decoder; The construction process of the second denoising autoencoder is similar to that of the first denoising autoencoder and is divided into the following 5 steps: Step 1: Construct the input layer; fix the parameters of the first denoising autoencoder, and use the low-dimensional features output by its encoding layer as the input data of the second denoising autoencoder; Step 2: Add a corruption layer; inject noise into the input features, the noise type includes random zero or Gaussian noise, to generate noisy features; Step 3: Construct the encoding layer. The construction process is the same as the encoding layer of the first autoencoder. The noisy features are further compressed to obtain lower-dimensional features containing deeper information. Step 4: Construct the decoding layer: The construction process is consistent with the decoding layer of the first denoising autoencoder, reconstructing the lower-dimensional features into the undestructed features of the input layer; Step 5: Train the autoencoder parameters; fix all the parameters of the first denoising autoencoder and only optimize the parameters of the second denoising autoencoder. The loss function, optimizer and hyperparameter settings are the same as those of the first denoising autoencoder. The construction process of the third denoising autoencoder is similar to that of the above denoising autoencoder. The output of the encoding layer of the second denoising autoencoder is used as the input data of the third denoising autoencoder. During training, the parameters of the second denoising autoencoder are fixed, and only the relevant parameters of the current denoising autoencoder are optimized. The second stage is global fine-tuning. After completing the layer-wise pre-training, the stacked denoising autoencoders are globally optimized end-to-end. The input layer, noise layer and encoding layers of multiple denoising autoencoders obtained by layered pre-training are stacked in sequence to form a feature extraction network ; The initialization parameters inherit the parameters of the layered pre-training stage and will have N tr Historical production data of samples X tr Input the stacked network, the network encodes the input data layer by layer to generate a high-level feature representation H; This high-level feature representation is used in two parts: The first part: reconstruct the original data; build a decoding layer, which consists of multiple neural network layers, including linear layers and activation function layers, and input high-level feature representations into the decoding layer; The decoding layer decodes the high-level features and outputs reconstructed data with the same dimension as the original input data. ; Calculate the mean square error between the reconstructed data and the original data as the reconstruction loss. The reconstruction loss expression is as follows: ;in and They are X tr and The nth sample of ; The second part: fault classification; establish a SoftMax regression model as a classifier to distinguish between seen and unseen faults and determine the category of a specific seen or unseen fault, map the high-level feature representation to the probability distribution of the fault category and obtain the fault category judged by the classifier as , using cross entropy as the classification loss function, the expression of the cross entropy loss function is as follows: ; where p represents the category, 1{ =p} is the indicator function, that is, if When , its value is 1, otherwise it is 0; Combining the above two losses, the overall loss function of the model is defined as follows: , where α1 and α2 are weight parameters, and the back propagation algorithm is used to optimize the parameters of the entire network.

4. The generalized zero-sample fault diagnosis method for a thermal power generation water supply pump group based on fault similarity according to claim 1 is characterized in that: In the step 2), the specific method of constructing the fault similarity matrix module is: Step 1: Based on the shallow expert experience and historical production data obtained, a similarity matrix between the seen faults and the seen faults of the water supply pump group is constructed: and the similarity matrix between seen and unseen faults ； Step 2: Using the historical data of seen faults, by calculating the feature center of each seen fault category, further obtain the similarity matrix between the seen fault categories; first, based on the seen fault samples in the historical production data of the water supply pump group, calculate the feature center of each seen fault category, and the feature center of the pth seen fault is defined as: ;in ∈H tr is the nth sample of the pth type of seen faults, is the number of samples of the p-th type of faults that have been seen, and then the similarity between the feature centers of any two seen fault categories is calculated, which is defined as: ; Finally, the similarity matrix between the seen fault categories calculated based on the seen fault data is obtained: ； Step 3: Combined with the similarity matrix obtained from shallow expert experience, the similarity matrix mapping relationship between data and shallow expert knowledge is obtained: ; The similarity diagnosis basis combining data-driven and expert knowledge is realized through mapping relationships, which is used to distinguish and diagnose the seen faults and unseen faults of the water supply pump group in the online stage.

5. The generalized zero-sample fault diagnosis method for a thermal power generation water supply pump group based on fault similarity according to claim 1 is characterized in that: In the step 3), the specific method is: 3.1) Input the test sample data monitored online into the pre-trained feature extractor to obtain the feature representation corresponding to the test sample; input the obtained test sample features into the pre-trained classifier to obtain the posterior probability distribution of the test sample belonging to each type of seen fault; 3.2) Compare the maximum probability value of the posterior probability distribution obtained in 3.1) with the set gate threshold to distinguish whether the test sample belongs to a seen fault or an unseen fault; when the maximum posterior probability value is greater than the gate threshold, the test sample is determined to be a seen fault; when the maximum posterior probability value is less than or equal to the gate threshold, the test sample is determined to be an unseen fault; 3.3) For the test sample determined as a seen fault in step 3.2), its fault category is determined to be the seen fault category with the largest posterior probability: ; where g lp is the posterior probability that the lth test sample belongs to the pth type of seen fault; 3.4) For the test samples that are determined to be unseen faults in step 3.2), calculate the similarity between the sample features and the feature centers of each seen fault category obtained in the offline training phase: ;in is the feature of the lth test sample, is the characteristic center of the p-th type of observed fault; Get the similarity ω between the current test sample and each seen fault l =[ ,……, ], using the similarity matrix mapping relationship between offline constructed data and shallow expert knowledge, the mapped The similarity w of the qth type of unseen fault obtained by shallow expert knowledge in the offline stage with respect to each seen fault q ∈W Q,P , q=1,2,……,Q for comparison to determine the unseen fault category to which the current sample belongs: ； 3.5) Based on the diagnosis results of step 3.3) and step 3.4), accurate classification and diagnosis of the observed faults and unobserved compound faults of the water supply pump group in the online stage are achieved.

Images