Intelligent early warning method and device for fault diagnosis of factory building drainage system

Abstract

The invention provides an intelligent early warning method and device for fault diagnosis of a factory building drainage system, and relates to the technical field of fault diagnosis, and the method comprises the steps: carrying out the decomposition and development of a fault and a reason thereof, and determining a possible reason corresponding to the fault; acquiring a sensor historical data set containing various fault information of a plant drainage system as training data; performing multi-modal data feature extraction on the training data; optimizing the fault diagnosis model to obtain corresponding optimal parameters; inputting the training data into the fault diagnosis model and the fault early warning model for training, and obtaining the trained fault diagnosis model and fault early warning model; and inputting real-time data into the trained fault diagnosis model and the fault early warning model, and outputting a corresponding fault type and an early warning result. Through a multi-modal data feature extraction method, data is analyzed from different angles or views, which is helpful for capturing richer information and structures in the data, and improves the accuracy of fault diagnosis.

Description

Technical Field

[0001] The present application relates to the technical field of fault diagnosis, and in particular to an intelligent early warning method and device for diagnosing faults in a plant drainage system. Background Art

[0002] The plant drainage system is a key component of industrial facilities. It is responsible for effectively draining wastewater and rainwater generated during the production process to keep the production environment clean and the production process smooth. With the development of industrial automation and intelligence, higher requirements are placed on the reliability and maintenance cost-effectiveness of the drainage system. However, due to factors such as long-term operation, equipment aging, improper operation or changes in the external environment, the drainage system may have various faults that affect its normal operation. Therefore, the development of an effective fault diagnosis and early warning system is of great significance to ensure the normal operation of the factory, prevent environmental pollution, and maintain production safety.

[0003] At present, only local parts of the system can be diagnosed without considering the overall system. The method of distributed monitoring of the drainage system is adopted to realize fault diagnosis, but the order of local fault diagnosis is not considered. In addition, none of the above methods combines fault diagnosis with early warning, and can only realize the diagnosis of existing faults, which requires processing a large amount of information, and the diagnostic effect is not ideal and the accuracy is low. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the first purpose of this application is to propose an intelligent early warning method for fault diagnosis of plant drainage system.

[0006] The second object of the present application is to provide a device.

[0007] The third objective of the present application is to provide an electronic device.

[0008] A fourth objective of the present application is to provide a computer-readable storage medium.

[0009] A fifth object of the present application is to provide a computer program product.

[0010] To achieve the above-mentioned purpose, the first embodiment of the present application proposes an intelligent early warning method for diagnosing faults in a plant drainage system, comprising:

[0011] Use the tree structure to decompose and expand the faults and causes in the plant drainage system, and determine the possible causes of various faults;

[0012] Obtain the historical sensor data set containing various fault information of the plant drainage system as training data;

[0013] Performing multimodal data feature extraction on the training data, and initializing a fault diagnosis model and a fault warning model;

[0014] Optimizing the fault diagnosis model according to a monopoly search algorithm to obtain corresponding optimal parameters, wherein the monopoly search algorithm includes a probabilistic perturbation strategy;

[0015] Inputting the training data after feature extraction into the fault diagnosis model and the fault warning model for training, and obtaining the trained fault diagnosis model and the fault warning model;

[0016] The real-time data collected by the sensor is input into the trained fault diagnosis model and fault warning model respectively, and the corresponding fault type and warning result are output, and warning processing is performed according to the fault type and warning result.

[0017] Optionally, the tree structure is used to decompose and expand the faults existing in the plant drainage system and their causes to determine possible causes corresponding to various faults, including:

[0018] Monitor flow signals, temperature signals, positive and negative pressure signals and vibration signals.

[0019] When the signal is abnormal, according to the type of abnormal signal, the direct cause of the fault is found step by step with the drainage pipe, water pump motor and water pump as units, and a reasonable fault gate is selected until the cause of the fault cannot be decomposed, and the possible cause corresponding to the fault is obtained.

[0020] Optionally, the extracting multimodal data features from the training data includes:

[0021] Determine a fault information data set and a normal data set in the training data;

[0022] Centralizing the fault information data set and the normal data set to obtain centralized data samples;

[0023] Using the selected principal component directions and data sample set, calculate the scatter matrix;

[0024] Performing eigenvalue decomposition on the scatter matrix to obtain corresponding eigenvectors and principal components;

[0025] The original tensor is reconstructed according to the principal component and the eigenvector to obtain the training data after feature extraction.

[0026] Optionally, the initialization of the fault diagnosis model and the fault warning model includes:

[0027] Initializing the fault diagnosis model, giving representations of nodes and edges by estimating soft masks, filtering out causal and shortcut features through an attention module, and generating causal suggestions and trivial suggestions based on the obtained features;

[0028] Use attention scores to construct soft masks and use graphs to get preliminary representations of causal and trivial adjoint graphs;

[0029] We decouple the causal and non-causal features from the graph, create the initial participation graph by computing the soft mask, use the graph neural network module to obtain the representation of the participation graph, and predict the category of the input graph through the readout function and classifier;

[0030] Backdoor adjustment parameterization using causal theory, combining causal and non-causal features;

[0031] The random addition method was used for the intervention formula to obtain the learning objectives.

[0032] Optionally, the initialization fault diagnosis model and fault early warning model include:

[0033] Initialize the fault warning model and sort the historical data set by time to obtain a sequence;

[0034] Construct an input embedding using convolutional layers and positional encodings;

[0035] Use temporal feature extractors to learn temporal and channel dependencies;

[0036] A linear layer is used to output the final prediction result.

[0037] Optionally, optimizing the fault diagnosis model according to the monopoly search algorithm to obtain corresponding optimal parameters includes:

[0038] Setting the objective function of the non-monopolistic search algorithm as the fault diagnosis accuracy, determining the population type as the number of hidden layer nodes and the learning rate of the fault diagnosis model; determining the number of iterations, the dimension size, and the upper and lower limits of the search space;

[0039] When the iteration counter is less than or equal to the preset threshold, the solution space is randomly explored and new potential optimal solutions are discovered;

[0040] When the iteration counter is greater than a preset threshold, a local search is performed in the neighborhood of the current solution to obtain the final solution.

[0041] To achieve the above-mentioned purpose, the second embodiment of the present application proposes an intelligent early warning device for diagnosing faults in a plant drainage system, comprising:

[0042] The cause determination module is used to decompose and expand the faults and their causes in the plant drainage system using a tree structure, and determine the possible causes corresponding to various faults;

[0043] The acquisition module is used to obtain the sensor historical data set containing various fault information of the plant drainage system as training data;

[0044] A feature extraction module, used to extract multimodal data features from the training data and initialize a fault diagnosis model and a fault warning model;

[0045] An optimization module, used for optimizing the fault diagnosis model according to a monopoly search algorithm to obtain corresponding optimal parameters, wherein the monopoly search algorithm includes a probabilistic perturbation strategy;

[0046] A training module, used for inputting the training data after feature extraction into the fault diagnosis model and the fault warning model for training, and obtaining the trained fault diagnosis model and the fault warning model;

[0047] The fault diagnosis module is used to input the real-time data collected by the sensor into the trained fault diagnosis model and fault warning model respectively, and output the corresponding fault type and warning result, and perform warning processing according to the fault type and warning result.

[0048] To achieve the above-mentioned purpose, the third aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0049] The memory stores computer-executable instructions;

[0050] The processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of the first aspects.

[0051] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present application proposes a computer-readable storage medium, in which computer-readable storage medium is stored computer execution instructions, and when the computer execution instructions are executed by a processor, they are used to implement the method as described in any one of the first aspects.

[0052] To achieve the above-mentioned purpose, the fifth aspect of the present application proposes a computer program product, which implements any method in the first aspect when executed by a processor.

[0053] The intelligent early warning method, device, electronic device and storage medium for diagnosing faults in a plant drainage system provided by the present application have the following beneficial effects:

[0054] 1. Effective use of the Multiview mechanism allows the PCA algorithm to analyze data from different angles or views, which helps capture richer information and structure in the data and builds a more comprehensive data representation than a single perspective, thereby improving the quality of feature extraction. In addition, multi-view feature extraction helps the model have better generalization capabilities when facing new data, can help reveal the complex structure in the data, and provide a more solid foundation for further analysis.

[0055] 2. The CTA-GNN model takes advantage of its ability to process graph-structured data and applies it to fault diagnosis in drainage systems. The model is able to naturally capture the complex relationships between components in a drainage system and go directly from raw sensor data to fault diagnosis results through end-to-end learning without the need for complex preprocessing steps. In addition, the CTA-GNN model is easily scalable and suitable for drainage systems of different sizes. It is able to analyze the propagation path and impact of faults in the system, and helps identify the root cause of the fault. It can also fuse information from multiple sensors and data sources to provide a comprehensive perspective on fault diagnosis. By improving the accuracy and timeliness of fault diagnosis, the CTA-GNN model helps reduce downtime and maintenance costs of drainage systems.

[0056] 3. In view of the problems of low operating efficiency and large input data requirements in traditional fault information prediction and early warning methods, the present invention adopts the MTS-Mixers model through the time and channel factor decomposition strategy and the low-rank characteristics of time series data, which effectively simplifies the complexity of the model and improves efficiency; in addition, the MTS-Mixers model improves prediction performance by capturing the time and channel correlation and the redundancy of multivariate time series data and simplifying redundant data, which is particularly useful when processing complex drainage system data. Therefore, the use of the MTS-Mixers model is crucial for drainage system fault early warnings that require rapid response;

[0057] 4. The traditional drainage system fault diagnosis method can only diagnose and identify existing faults, and it is carried out under the condition that the equipment has already failed, so it cannot effectively avoid the actual economic losses and time costs caused by equipment failure. The present invention effectively combines fault diagnosis with fault warning, and accurately predicts the equipment operation status at future times on the basis of accurate identification of existing faults, and judges the equipment working status at future times based on the prediction results, effectively reducing economic losses and time costs.

[0058] 5. The monopoly search algorithm has problems such as slow convergence speed and easy to fall into local optimality during the optimization process. Therefore, the present invention adds a probabilistic perturbation strategy to the population update process of the monopoly search algorithm to help the algorithm jump out of the local optimality and improve the search efficiency.

[0059] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0061] Figure 1 A flow chart of an intelligent early warning method for diagnosing a plant drainage system fault provided in an embodiment of the present application;

[0062] Figure 2 It is an expanded diagram of the tree-structure drainage system failure and its causes;

[0063] Figure 3 A flow chart of an intelligent early warning method for diagnosing a plant drainage system fault provided in an embodiment of the present application;

[0064] Figure 4 A schematic diagram of the structure of an intelligent early warning device for diagnosing faults in a factory drainage system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0065] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0066] The plant drainage system is a key component of industrial facilities. It is responsible for effectively draining wastewater and rainwater generated during the production process to keep the production environment clean and the production process smooth. With the development of industrial automation and intelligence, higher requirements are placed on the reliability and maintenance cost-effectiveness of the drainage system. However, due to factors such as long-term operation, equipment aging, improper operation or changes in the external environment, the drainage system may have various faults that affect its normal operation. Therefore, the development of an effective fault diagnosis and early warning system is of great significance to ensure the normal operation of the factory, prevent environmental pollution, and maintain production safety.

[0067] With the improvement of industrial automation and intelligence, higher requirements are placed on the reliability and maintenance cost-effectiveness of drainage systems. Fault diagnosis technology can help timely discover and solve problems in drainage systems, reduce downtime, and avoid environmental pollution and economic losses. Fault warning systems can predict potential failures and take measures in advance, thereby improving the safety and reliability of the system and reducing unexpected downtime and maintenance costs. The plant drainage system involves a variety of equipment and complex physical processes, generating a large amount of multimodal data. How to extract useful information from this data and establish accurate fault diagnosis and early warning models is a technical challenge. Machine learning technology, especially deep learning and data fusion technology, makes it possible to process complex data sets and establish efficient fault diagnosis and early warning models. In order to improve the accuracy of fault diagnosis and early warning, it is necessary to improve and optimize the existing machine learning algorithms to better adapt to specific application scenarios.

[0068] At present, the solutions to this problem can be mainly divided into two categories: physical model and data-driven. For example, the collected equipment signals are compared with typical fault characteristic signals to achieve fault diagnosis, but the specific location and specific components of the fault source cannot be determined; the rough set theory is combined with Petri nets to construct a fault diagnosis model to diagnose the fault of the mine drainage system, but a large amount of information needs to be processed, and the diagnosis effect is not ideal; the fault monitoring system for the water collection well drainage device is developed using Lab VIEW, which can achieve good results, but can only diagnose the local part of the system without considering the integrity of the system. The method of distributed monitoring of the drainage system is adopted to achieve fault diagnosis, but the order of local fault diagnosis is not considered. In addition, none of the above methods combines fault diagnosis with early warning, and can only diagnose existing faults, but cannot make early warnings of faults for the equipment operation status at future moments in advance, thereby reducing the time cost.

[0069] In order to solve the above problems, the present invention proposes a method for constructing an intelligent early warning system for fault diagnosis of a factory drainage system. On the basis of overcoming the problems existing in the above drainage system fault diagnosis solution method, the fault early warning and fault diagnosis of the drainage system are cleverly combined, which can accurately realize the accurate identification of existing faults in the factory drainage system and timely early warning of the equipment operation status in the future.

[0070] Aiming at the problems that traditional drainage system fault diagnosis methods cannot determine the specific location and specific components of the fault source, can achieve accurate fault diagnosis but need to process a large amount of information, resulting in unsatisfactory diagnosis effect, can only diagnose part of the system without considering the integrity of the system, and although the fault is accurately identified, the order of local fault diagnosis is not considered. A method for constructing an intelligent early warning system for factory drainage system fault diagnosis is proposed. A comprehensive method for factory drainage system fault diagnosis and early warning is proposed, which uses a tree structure to decompose fault types, multi-sensor data sets, improved multimodal feature extraction algorithms, optimized neural network models and evolutionary algorithms. The aim is to improve the accuracy of fault detection and the timeliness of early warning through data-driven analysis.

[0071] The present application embodiment provides an intelligent early warning method for diagnosing plant drainage system faults. Figure 1 The following is a flow chart of an intelligent early warning method for diagnosing a factory drainage system fault provided by an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0072] Step 101, using a tree structure to decompose and expand the faults existing in the plant drainage system and their causes, and determine the possible causes corresponding to various faults;

[0073] Step 102, obtaining a sensor historical data set containing various fault information of the plant drainage system as training data;

[0074] Step 103, extracting multimodal data features from the training data, and initializing a fault diagnosis model and a fault warning model;

[0075] In this embodiment, the Multiview mechanism is added to the KPCA algorithm to improve the KPCA algorithm, and the improved Multiview KPCA algorithm is used to extract multimodal data features from all acquired data. A CTA-GNN fault diagnosis model and an MTS-Mixers fault warning model are established. The MTS-Mixers fault warning model introduces the BatchNormalization regularization method into the original MTS-Mixers model to improve the original regularization method of the MTS-Mixers model and obtain an improved MTS-Mixers model.

[0076] Step 104, optimizing the fault diagnosis model according to a monopoly search algorithm to obtain corresponding optimal parameters, wherein the monopoly search algorithm includes a probabilistic perturbation strategy;

[0077] In this embodiment, the probability perturbation strategy is added to the population update process of the monopoly search algorithm to help the algorithm jump out of the local optimum. The optimization of the non-monopoly search algorithm is realized, and the optimized INO algorithm is obtained. The INO algorithm is used to optimize the number of hidden layer nodes and learning rate hyperparameters of the CTA-GNN model to obtain the corresponding optimal parameters, and the optimized CTA-GNN fault diagnosis model is used to diagnose the fault conditions of the plant drainage system at the current and future moments.

[0078] Step 105, inputting the training data after feature extraction into the fault diagnosis model and the fault warning model for training, and obtaining the trained fault diagnosis model and the fault warning model;

[0079] In this embodiment, the enhanced data is used as the input of the CTA-GNN fault diagnosis model and the BN-Conv-Trans fault warning model, the two models are trained, and the trained fault diagnosis model and fault warning model are obtained; the real-time data is sent to the trained CTA-GNN fault diagnosis model and the MTS-Mixers fault warning model to perform fault type diagnosis and identification and predict the future status of each device, and judge whether the warning threshold is exceeded according to the prediction results to perform warning processing.

[0080] Step 106, input the real-time data collected by the sensor into the trained fault diagnosis model and fault warning model respectively, and output the corresponding fault type and warning result, and perform warning processing according to the fault type and warning result.

[0081] Optionally, the step 101 uses a tree structure to decompose and expand the faults existing in the plant drainage system and their causes, and determines possible causes corresponding to various faults, including:

[0082] Monitor flow signals, temperature signals, positive and negative pressure signals and vibration signals.

[0083] When the signal is abnormal, according to the type of abnormal signal, the direct cause of the fault is found step by step with the drainage pipe, water pump motor and water pump as units, and a reasonable fault gate is selected until the cause of the fault cannot be decomposed, and the possible cause corresponding to the fault is obtained.

[0084] Figure 2 The following is an expanded diagram of the tree-structure drainage system failure and its causes, such as Figure 2As shown, in this embodiment, the types of fault signals that can be collected by the plant drainage system are mainly abnormal drainage flow, abnormal motor temperature, abnormal drainage pressure, abnormal voltage and current, and abnormal vibration signals. By monitoring the flow, temperature, positive and negative pressure, and vibration signals, it can be determined whether the drainage system is operating normally, so these four signal abnormalities are selected as the first-level intermediate events. Then, according to the type of abnormal signal, the direct cause of the fault is searched step by step downward with the drainage pipe, water pump motor, and water pump as units, and a reasonable fault gate is selected until the cause of the fault cannot be decomposed, and the root cause of the system failure is obtained.

[0085] Optionally, the extracting multimodal data features from the training data includes:

[0086] First, determine the set of acquired fault information data set and normal data set D = [r1 d1 d2 d3 d4], and define the analysis goal as feature extraction. Among them: Define a q-dimensional index subset Where c1<… <c q , p is the rank of the tensor, and q is the selected PCA depth. This direction determines the subset of the data that will be analyzed.

[0087] Centralize the input samples:

[0088]

[0089] Where D(n) c is the centralized data sample; D(n) represents the nth data sample; Represents the average value of all data samples; n = 1,…,N; n represents the index of the data sample; N is the total number of data samples.

[0090] Using the selected PCA direction S and the data sample set, calculate the scatter matrix:

[0091]

[0092] This matrix reflects the variance of the data in different directions.

[0093] Perform eigenvalue decomposition on the scatter matrix and find the corresponding eigenvector U * As the basis tensor, calculate the principal component V * We can get:

[0094] V * (n) = D(n) c × S U * (3)

[0095] Using the principal component V *and the basis tensor U * Reconstruct the original tensor:

[0096]

[0097] Where j represents the number of selected principal components; Indicates V * The tensor of the jth principal component in ; Indicates U * The basis tensor associated with the j-th principal component in .

[0098] In order to further improve the performance of compression and feature extraction, multiple non-overlapping directions S = {S1, ..., S K}, repeat the above PCA process to form a global multi-view PCA. Get the fault information data set and normal data set after feature processing:

[0099] Optionally, the initialization of the fault diagnosis model and the fault warning model includes:

[0100] (1) Representations of nodes and edges are given by estimating soft masks. First, an attention module is needed to filter out causal and shortcut features. Then, based on the obtained features, causal proposals and trivial proposals are generated. Let f(·) denote the GNN-based encoder, G = {B, S} denote the input graph, and the node representation is as follows:

[0101] H=f(B,S) (6)

[0102] Among them, B is the adjacency matrix and S is the node feature matrix.

[0103] Calculate node-level attention scores through MLPs (Multi-layer Perceptrons) and edge-level attention scores and their non-causal corresponding scores and For node v i and edge (v i ,v j ) can be obtained:

[0104]

[0105]

[0106] where h i is node v i Feature representation, MLP node (·) is the MLP for the node. i ||h j is node v i and vj Feature concatenation, MLP node (·) is the MLP for edges and σ(·) is the Softmax function.

[0107] Using Attention Score To construct a soft mask Mnode, Medge, Finally, using Figure G: G c = {B⊙M edge , S⊙M node}and A preliminary representation of causal and trivial adjoint graphs is obtained.

[0108] (2) Obtain causal and non-causal features from the graph through disentanglement. Create an initial participation graph by calculating a soft mask. In order to obtain causal and trivial participation graphs, GNN modules can be used to obtain the representation of the participation graph respectively. Finally, the category of the input graph is predicted by the readout function and the classifier:

[0109]

[0110]

[0111] The goal of the causal adjoint graph is to estimate causal features whose representation can be classified as ground truth labels. Accordingly, the supervised loss for the graph classification problem is defined as:

[0112]

[0113] In contrast, the trivial attendance graph is designed to approximate non-causal trivial patterns. Therefore, the prediction of the trivial attendance graph can be motivated for all known fault classes, and the unified loss on the graph classification problem is defined as:

[0114]

[0115] Where KL is the KL-divergence, y unif is a uniform distribution. By optimizing the above two objectives, causal features and trivial features are distinguished.

[0116] (3) Use the backdoor adjustment parameterization of causal theory to combine causal and non-causal features. Backdoor adjustment can effectively weaken the confounding effect by stratifying the confounding factors and pairing each layer of the target causal attention graph with the trivial attention graph to form an intervention graph. Due to the irregularity of graph data, it prevents us from intervening at the data level and can only intervene implicitly at the representation level. We propose a loss guided by backdoor adjustment (causal intervention loss):

[0117]

[0118]

[0119] Among them, z G′ It is the classification result of the implicit intervention graph G' in the classifier Φ. It is a causal adjoint graph G c Representation. It is layered G t′ Representation. is a collection of estimates of the hierarchical trivial participation graph that yields trivial features present in the training data.

[0120] The random addition method is used to intervene in formula (9), and the learning objective of CTA-GNN, namely the total loss, is obtained:

[0121] L=L sup +ψ1L unif +ψ2L caus (15)

[0122] where ψ1 and ψ2 are constants that control the degree of disentanglement and causal interference, and they are hyperparameters that can be adjusted.

[0123] Optionally, the initialization fault diagnosis model and fault early warning model include:

[0124] The signal time series collected by the drainage flow sensor Example, where n is the length of the time series and c is the number of channels. The predicted value of the sequence in the next m time steps is expressed as follows:

[0125]

[0126] Where P f represents the predicted value of the next m time steps; f represents the time step.

[0127] Using the Input Embedding module, we get the input embedding through a 1D convolutional layer and positional encoding:

[0128]

[0129] Where h represents the data related to the historical time series; PE(·) represents the position code.

[0130] Use a temporal feature extractor MLP to learn temporal and channel dependencies:

[0131]

[0132]

[0133] In the formula, Indicates time dependence; Indicates channel dependency.

[0134] Use a linear layer to output the final prediction result:

[0135]

[0136] Optionally, optimizing the fault diagnosis model according to the monopoly search algorithm to obtain corresponding optimal parameters includes:

[0137] The NO search algorithm has two main phases, the exploration phase and the exploitation phase. In these phases, the NO algorithm modifies the candidate solutions in each dimension (decision variable). The NO algorithm flowchart is shown in Figure 1. Figure 1 shown.

[0138] Set the objective function of the NO algorithm to the CTA-GNN fault diagnosis accuracy, determine the number of hidden layer nodes and the learning rate of the CTA-GNN model; determine the number of iterations, dimension size, and upper and lower limits of the search space;

[0139] In the exploration phase, the algorithm randomly explores the solution space to find new potential optimal solutions. The exploration phase is performed when the iteration counter (t) is less than or equal to half of the total number of iterations (T), as shown in the following formula:

[0140] Anew(j)=rand*A(RP) (21)

[0141] Where Anew(j) is the jth dimension in the newly generated solution, rand is a random value between 0 and 1. j is the dimension index defined between 1 and the maximum dimension of the search space (D), A is the current solution, and A(RP) is the selected random dimension of the current solution. RP is a random number between 1 and D.

[0142] In the development phase, the algorithm performs local search in the neighborhood of the current solution to refine the quality of the solution. The development phase is executed when the iteration count is higher than half of T. This phase aims to utilize the neighborhood around the current solution. The current solution is randomly perturbed to search for a better position around it. When the algorithm gradually converges, it will be difficult to jump out once it is trapped in the local optimum. Therefore, the present invention introduces a probabilistic perturbation strategy into the algorithm population update process to help the algorithm jump out of the local optimum. The specific implementation process is as follows:

[0143] Anew(j)=A(j)-(A(SRP)*rand)*eps-(A(j)-NO) (22)

[0144]

[0145]

[0146] Where: P r is the disturbance probability; D a is the dimension of the algorithm optimization problem and D a =2; T a is the maximum number of iterations. In the early stages of algorithm iteration, the perturbation probability is small, so that the algorithm can quickly find the global optimal solution; in the later stages of algorithm iteration, the perturbation probability increases, enhancing population diversity. Generally speaking, T a is a large constant. When the algorithm iterates to the later stage t a →T a , disturbance probability:

[0147]

[0148] After adding the probability perturbation, the formula for updating the position of the algorithm particles is optimized as follows:

[0149] Anew(j)=A(j)-(A(SRP)*rand1)*eps-(A(j)-NO); rand1 <P r (25)

[0150] where A(j) is the jth position in the current solution, rand1 is a random value between [0, 1], A(SRP) is a dimension randomly selected from the current solution dimension, eps is a very small value, SRP is a solution random dimension, a random number between 1 and the size of the dimension; NO is a tuning parameter used to adjust the search process of the proposed method.

[0151] Figure 3 The following is a flow chart of an intelligent early warning method for diagnosing a factory drainage system fault provided by an embodiment of the present application. Figure 3As shown, the power plant drainage system is mainly composed of valves, pump stations, motors, pipes and other equipment. Due to its special working conditions and the need for close collaboration between multiple important equipment, the failure rate of important components of the entire system is high and it is difficult to directly identify the fault. To address the above problems, this patent proposes an intelligent early warning system for fault diagnosis of the power plant drainage system. Firstly, the faults and their causes in the drainage system are decomposed and expanded using a tree structure, and the root possible causes corresponding to each fault are matched one by one; secondly, the sensor historical data set containing various fault information of the plant drainage system is obtained as the training data of the diagnosis and early warning model; various signal data collected by the sensor in real time are obtained as the target data of fault diagnosis and early warning; and all the acquired data are enhanced using the Multiview PCA model; then, the CTA-GNN fault diagnosis model and the MTS-Mixers fault early warning model are trained respectively using the enhanced historical data to obtain the trained models; finally, the enhanced real-time data is sent to the CTA-GNN fault diagnosis model for fault type diagnosis and identification; it is sent to the MTS-Mixers fault early warning model to predict the operating status of each equipment in the future and the prediction results are used as the input of the trained CTA-GNN fault diagnosis model to realize the fault diagnosis and early warning of the plant drainage system equipment in the future.

[0152] The Multiview PAC data feature extraction method designed and adopted by the present invention effectively utilizes the Multiview mechanism to allow the PCA algorithm to analyze data from different angles or views, which helps to capture richer information and structure in the data, and can construct a more comprehensive data representation than a single perspective, thereby improving the quality of feature extraction. In addition, multi-perspective feature extraction helps the model have better generalization ability when facing new data, can help reveal the complex structure in the data, and provide a more solid foundation for further analysis.

[0153] 2. The CTA-GNN model takes advantage of its ability to process graph-structured data and applies it to fault diagnosis in drainage systems. The model is able to naturally capture the complex relationships between components in a drainage system and go directly from raw sensor data to fault diagnosis results through end-to-end learning without the need for complex preprocessing steps. In addition, the CTA-GNN model is easily scalable and suitable for drainage systems of different sizes. It is able to analyze the propagation path and impact of faults in the system, and helps identify the root cause of the fault. It can also fuse information from multiple sensors and data sources to provide a comprehensive perspective on fault diagnosis. By improving the accuracy and timeliness of fault diagnosis, the CTA-GNN model helps reduce downtime and maintenance costs of drainage systems.

[0154] 3. In view of the problems of low operating efficiency and large input data requirements in traditional fault information prediction and early warning methods, the present invention adopts the MTS-Mixers model through the time and channel factor decomposition strategy and the low-rank characteristics of time series data, which effectively simplifies the complexity of the model and improves efficiency; in addition, the MTS-Mixers model improves prediction performance by capturing the time and channel correlation and the redundancy of multivariate time series data and simplifying redundant data, which is particularly useful when processing complex drainage system data. Therefore, the use of the MTS-Mixers model is crucial for drainage system fault early warnings that require rapid response;

[0155] 4. The traditional drainage system fault diagnosis method can only diagnose and identify existing faults, and it is carried out under the condition that the equipment has already failed, so it cannot effectively avoid the actual economic losses and time costs caused by equipment failure. The present invention effectively combines fault diagnosis with fault warning, and accurately predicts the equipment operation status at future times on the basis of accurate identification of existing faults, and judges the equipment working status at future times based on the prediction results, effectively reducing economic losses and time costs.

[0156] 5. The monopoly search algorithm has problems such as slow convergence speed and easy to fall into local optimality during the optimization process. Therefore, the present invention adds a probabilistic perturbation strategy to the population update process of the monopoly search algorithm to help the algorithm jump out of the local optimality and improve the search efficiency.

[0157] In order to implement the above-mentioned embodiment, the present application also proposes an intelligent early warning device for diagnosing faults in a plant drainage system. Figure 4 This is a schematic diagram of the structure of an intelligent early warning device for diagnosing factory drainage system faults provided in an embodiment of the present application. Figure 4 As shown, the device comprises:

[0158] Cause determination module 410, used to decompose and expand the faults existing in the plant drainage system and their causes using a tree structure, and determine the possible causes corresponding to various faults;

[0159] The acquisition module 420 is used to obtain the sensor historical data set containing various fault information of the plant drainage system as training data;

[0160] A feature extraction module 430 is used to extract multimodal data features from the training data and initialize a fault diagnosis model and a fault warning model;

[0161] An optimization module 440, configured to optimize the fault diagnosis model according to a monopoly search algorithm to obtain corresponding optimal parameters, wherein the monopoly search algorithm includes a probabilistic perturbation strategy;

[0162] The training module 450 is used to input the training data after feature extraction into the fault diagnosis model and the fault warning model for training, and obtain the trained fault diagnosis model and the fault warning model;

[0163] The fault diagnosis module 460 is used to input the real-time data collected by the sensor into the trained fault diagnosis model and the fault warning model respectively, and output the corresponding fault type and warning result, and perform warning processing according to the fault type and warning result.

[0164] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0165] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0166] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0167] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0168] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0169] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0170] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0171] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0172] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0173] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0174] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0175] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0176] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0177] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. An intelligent early warning method for fault diagnosis of a plant drainage system, characterized in that: The following steps are involved: Use the tree structure to decompose and expand the faults and causes in the plant drainage system, and determine the possible causes of various faults; Obtain the historical sensor data set containing various fault information of the plant drainage system as training data; Performing multimodal data feature extraction on the training data, and initializing a fault diagnosis model and a fault warning model; Optimizing the fault diagnosis model according to a monopoly search algorithm to obtain corresponding optimal parameters, wherein the monopoly search algorithm includes a probabilistic perturbation strategy; Inputting the training data after feature extraction into the fault diagnosis model and the fault warning model for training, and obtaining the trained fault diagnosis model and the fault warning model; The real-time data collected by the sensor is input into the trained fault diagnosis model and fault warning model respectively, and the corresponding fault type and warning result are output, and warning processing is performed according to the fault type and warning result.

2. The method according to claim 1, characterized in that The tree structure is used to decompose and expand the faults and their causes in the plant drainage system, and determine the possible causes of various faults, including: Monitor flow signals, temperature signals, positive and negative pressure signals and vibration signals. When the signal is abnormal, the direct cause of the fault is searched step by step downward with the drainage pipe, water pump motor and water pump as units according to the type of abnormal signal, and a reasonable fault gate is selected until the cause of the fault cannot be decomposed, and the possible cause corresponding to the fault is obtained.

3. The method according to claim 2, characterized in that The extracting multimodal data features from the training data includes: Determine a fault information data set and a normal data set in the training data; Centralizing the fault information data set and the normal data set to obtain centralized data samples; Using the selected principal component directions and data sample set, calculate the scatter matrix; Performing eigenvalue decomposition on the scatter matrix to obtain corresponding eigenvectors and principal components; The original tensor is reconstructed according to the principal component and the eigenvector to obtain the training data after feature extraction.

4. The method according to claim 3, characterized in that The initialization fault diagnosis model and fault early warning model include: Initializing the fault diagnosis model, giving representations of nodes and edges by estimating soft masks, filtering out causal and shortcut features through an attention module, and generating causal suggestions and trivial suggestions based on the obtained features; Use attention scores to construct soft masks and use graphs to get preliminary representations of causal and trivial adjoint graphs; We obtain causal and non-causal features from the graph by decoupling, create an initial participation graph by calculating soft masks, use graph neural network modules to obtain the representation of the participation graph, and predict the category of the input graph through readout functions and classifiers; Use backdoor tuning parameterization of causal theory to combine causal and non-causal features; The random addition method was used for the intervention formula to obtain the learning objectives.

5. The method according to claim 4, characterized in that The initialization fault diagnosis model and fault early warning model include: Initialize the fault warning model and sort the historical data set by time to obtain a sequence; Construct an input embedding using convolutional layers and positional encodings; Use temporal feature extractors to learn temporal and channel dependencies; A linear layer is used to output the final prediction result.

6. The method according to claim 5, characterized in that The step of optimizing the fault diagnosis model according to the monopoly search algorithm to obtain corresponding optimal parameters includes: Setting the objective function of the non-monopolistic search algorithm as the fault diagnosis accuracy, determining the population type as the number of hidden layer nodes and the learning rate of the fault diagnosis model; determining the number of iterations, the dimension size, and the upper and lower limits of the search space; When the iteration counter is less than or equal to the preset threshold, the solution space is randomly explored and new potential optimal solutions are discovered; When the iteration counter is greater than a preset threshold, a local search is performed in the neighborhood of the current solution to obtain the final solution.

7. An intelligent early warning device for fault diagnosis of a plant drainage system, characterized in that: include: The cause determination module is used to decompose and expand the faults and their causes in the plant drainage system using a tree structure, and determine the possible causes corresponding to various faults; The acquisition module is used to obtain the sensor historical data set containing various fault information of the plant drainage system as training data; A feature extraction module, used to extract multimodal data features from the training data and initialize a fault diagnosis model and a fault warning model; An optimization module, used for optimizing the fault diagnosis model according to a monopoly search algorithm to obtain corresponding optimal parameters, wherein the monopoly search algorithm includes a probabilistic perturbation strategy; A training module, used for inputting the training data after feature extraction into the fault diagnosis model and the fault warning model for training, and obtaining the trained fault diagnosis model and the fault warning model; The fault diagnosis module is used to input the real-time data collected by the sensor into the trained fault diagnosis model and fault warning model respectively, and output the corresponding fault type and warning result, and perform warning processing according to the fault type and warning result.

8. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images