Fault diagnosis method and system for industrial equipment based on integrated optimization of graph structure

Abstract

The present invention provides a fault diagnosis method for industrial equipment, a fault diagnosis system for industrial equipment, a storage medium, and a terminal device that optimize problems such as low prediction accuracy and poor robustness in fault diagnosis and significantly improve the stability of industrial Internet devices. [Solution] A method for fault diagnosis of industrial equipment based on integrated optimization of graph structure includes the steps of obtaining an original equipment dataset, constructing an original graph structure based on the original equipment dataset, extracting two base views based on the original graph structure, calculating graph node embeddings of the base views using a GCN, and recalculating the edge probabilities in the graph structure based on the graph node embeddings, performing view fusion based on the edge probabilities in the graph structure to obtain a view after initial optimization, and processing the fused view using a graph attention network to obtain an enhanced view.

Description

[Technical Field]

[0001] The present invention relates to the field of fault diagnosis technology, and more particularly to a fault diagnosis method and system for industrial equipment based on integrated optimization of graph structures. [Background technology]

[0002] With the rapid development of fault diagnosis technology for industrial Internet equipment, the system has become an important tool for equipment maintenance and management. By analyzing the equipment's historical data, operating status, and fault records, the fault diagnosis system can provide operators with timely warnings and repair advice, thereby improving the stability and efficiency of equipment operation.

[0003] Currently, traditional fault diagnosis methods, such as rule-based methods, have insufficient performance when dealing with complex equipment failure modes and variations, and statistical models often lack comprehensiveness in capturing the deep operating conditions and failure modes of equipment. Furthermore, traditional diagnostic systems face performance bottlenecks in terms of accuracy, real-time performance, and large-scale data processing.

[0004] Graph neural networks (GNNs), as an advanced technology for processing graph-structured data, show great potential in equipment fault diagnosis. However, when directly applying GNNs to fault diagnosis, significant problems exist, including sparse graph structures, data noise, and computational resource limitations. The sparsity of the graph structure makes it difficult for GNNs to capture the potential relationships between equipment and faults, while data noise can cause the model to learn inaccurate features. To improve the effectiveness of fault diagnosis systems, how to effectively utilize GNNs for graph data modeling and solve these problems has become a focus of current research and application.

[0005] Traditional GNN algorithms typically train by directly inputting the original graph structure into the network, but this poses two major challenges. First, extracting and fusing a valid graph structure from the original data is a challenge. After processing real-world data, graph data typically contains noise and incomplete node features, complicating the construction of an accurate graph structure and the extraction of useful features from it. Second, GNN algorithms are based on a message-passing mechanism, which requires the model to constantly aggregate information from the multi-hop neighborhood of nodes during training. This process requires processing large amounts of node data, slowing down the training and inference speed of GNN models. Furthermore, GNN models are generally considered black boxes, resulting in a lack of transparency in their internal feature transformation and decision-making processes. Therefore, to address these issues in the practical application of GNNs, improvements are needed in terms of efficiency, interpretability, and model complexity. Summary of the Invention [Problem to be solved by the invention]

[0006] To solve the above-mentioned problems, the present invention provides a fault diagnosis method and system for industrial equipment based on joint optimization of graph structure. [Means for solving the problem]

[0007] In a first aspect, the fault diagnosis method for industrial equipment based on integrated optimization of graph structure provided by the present invention adopts the following technical solution. A fault diagnosis method for industrial equipment based on integrated optimization of graph structures is obtaining an original equipment data set; constructing an original graph structure based on the original equipment dataset; Extracting two base views based on the original graph structure, calculating graph node embeddings of the base views using GCN, and recalculating the probabilities of edges in the graph structure based on the graph node embeddings; performing view fusion based on the probability of edges in the graph structure to obtain a first optimized view; processing the fused view by a GAT network to obtain an enhanced view; The method includes adopting a probabilistic block model as a generative model, iteratively optimizing the reinforced view using Bayesian estimation and expectation maximization algorithms, and reconstructing the graph structure.

[0008] Furthermore, the step of constructing an original graph structure based on the original equipment dataset includes creating the collected original equipment data into a graph dataset that is applied to network training, with the equipment being nodes, the communication forms between the equipment being edges, and various data and attributes of the equipment being node features.

[0009] Furthermore, the step of extracting two base views based on the original graph structure and calculating graph node embeddings of the base views using a GCN includes extracting two base views, namely an adjacency matrix and a transfer matrix, from the original graph structure, performing elementary processing on the selected two base views, and obtaining view embeddings using a graph convolutional network (GCN); Q 1 =σ(GCN(V1,X)) Q 2 =σ(GCN(V2,X)) where σ is a nonlinear activation function, which introduces a nonlinear transformation and improves the network's representation ability and ability to learn complex data.

[0010] Furthermore, the step of recalculating the probabilities of edges in the graph structure based on the graph node embeddings includes, for the target node: The node embedding isIt involves connecting with other node embeddings, then normalizing the node weights to obtain the existence probability of edges between node pairs, combining it with the overall probability to obtain a probability matrix, and combining the probability matrix with the original graph structure to obtain the processed base view.

[0011] moreover, The step of processing the fused view by the GAT network to obtain an enhanced view includes: This involves using a self-attention mechanism to enhance interactions and feature representations between nodes, dynamically evaluating, weighting, and aggregating features of neighboring nodes, thereby generating an enhanced view with more information.

[0012] Further, the step of enhancing the interactions and feature representations between nodes using a self-attention mechanism includes calculating an attention coefficient between a pair of nodes; e ij =d([Mh i ||Mh j ]),j∈N i where M is a shared parameter, h represents the node feature, and d is a real parameter for projecting the connected high-dimensional features onto a real field.

[0013] Furthermore, the step of adopting a probabilistic block model as a generative model includes using a probabilistic block model SBM as a generative model to simulate an optimal graph structure, and the process of generating a simulated optimal graph structure G is probabilistically formalized as follows: JPEG0007795843000002.jpg17128JPEG0007795843000003.jpg1449Here, SBM divides the whole network into multiple communities, and i represents the community to which node i belongs, Ω is a parameter of SBM, and Ω cicj is any node v in community c i and node v in community c j represents the probability that an edge exists between L represents tag information, G ijrepresents the probability that an edge exists between node i and node j in the simulated graph structure G, Z is the prediction obtained after performing softmax processing on the acquired node represented by H, and z i is node v i represents the predicted result of y i is v i represents the truth tag of v L represents all nodes in the training set, and node v i If there is a tagged value in c i The value of is the tagged value y i If not, the value is the predicted value z i is.

[0014] In a second aspect, a fault diagnosis system for industrial equipment based on integrated optimization of a graph structure includes: a data acquisition module configured to acquire an original equipment dataset and construct an original graph structure based on the original equipment dataset; an embedding module configured to extract two base views based on the original graph structure, calculate graph node embeddings of the base views using a GCN, and recalculate the probabilities of edges in the graph structure based on the graph node embeddings; a fusion module configured to perform view fusion based on the probabilities of edges in the graph structure to obtain a preliminary optimized view; an enhancement module configured to process the fused view by the GAT network to obtain an enhanced view; and an optimization module configured to adopt the probabilistic block model as a generative model, iteratively optimize the reinforced view using Bayesian estimation and an expectation maximization algorithm, reconstruct the graph structure, and obtain a final graph structure.

[0015] In a third aspect, the present invention provides a computer-readable storage medium having stored thereon a plurality of instructions, the instructions being suitable for being loaded by a processor of a terminal device and for executing the method for fault diagnosis of industrial equipment based on the integrated optimization of a graph structure.

[0016] In a fourth aspect, the present invention provides a terminal device including a processor used to realize each instruction and a computer-readable storage medium used to store a plurality of instructions, the instructions being loaded by the processor and suitable for executing the method for fault diagnosis of industrial equipment based on integrated optimization of the graph structure. [Effects of the Invention]

[0017] In summary, the present invention has the following beneficial effects: (1) This invention introduces artificial intelligence and deep learning algorithms, combines a recommendation system with a graph neural network, and utilizes a series of technologies, including graph structure learning algorithms, base view processing, graph attention mechanisms, and probabilistic structure inversion, to realize fault diagnosis of industrial Internet devices. (2) The present invention provides an advanced fault diagnosis method for industrial Internet equipment. The system combines base view fusion processing and attention mechanism, and combines it with Bayesian structural inverse estimation to optimize the problems of traditional fault diagnosis, such as low prediction accuracy and poor robustness, thereby significantly improving the stability of industrial Internet equipment. (3) Solve the problem of low fit between the dataset and the training model by extracting a base view based on the source data, performing coding and reconstruction fusion using GCN to obtain a rudimentary view, performing secondary optimization using GAT network to enhance the node information representation, and then introducing Bayesian thinking to generate an optimal graph frame based on SBM, and reconstructing the final graph structure using node mapping to improve the accuracy of downstream tasks. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a schematic diagram of a fault diagnosis method for industrial equipment based on integrated optimization of graph structures according to a first embodiment of the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0019] The present invention will be described in more detail below in conjunction with the drawings. Example 1

[0020] Referring to FIG. 1, the fault diagnosis method for industrial equipment based on integrated optimization of graph structures according to the present embodiment includes: S1: Create a graph dataset from the collected original equipment dataset for network training. The graph dataset is composed of equipment (machines, sensors, control systems, etc.) as nodes, the communication patterns between equipment as edges, and various data and attributes of equipment as node features. In the initial stage, S2 extracts and preprocesses two base views based on the original graph structure, calculates graph node embeddings using GCN, and recalculates the probability of edge connections based on the node embeddings. S3 obtains the prediction results of the base view using the GCN network, integrates the information using a view fusion algorithm, and generates a view after initial optimization; S4 processes the fused graph using a graph attention network (GAT), enhances the interactions and feature representations between nodes using a self-attention mechanism, dynamically evaluates, weights, and aggregates the features of neighboring nodes, and thereby generates an enhanced view with more information; S5 involves constructing a multi-level observation set, using a probabilistic block model (SBM) as a generative model, iteratively optimizing the graph structure using Bayesian inference and the expectation maximization (EM) algorithm, and finally generating a high-quality graph structure, improving the performance and robustness of the graph neural network in downstream tasks.

[0021] The step S1 specifically includes S1.1.

[0022] In S1.1, the acquired equipment data is preprocessed, and a network structure graph G = (V, A, X, Y) is established based on the preprocessed data set, where V is a node set consisting of basic equipment information, A is the node adjacency matrix, X is a node feature set consisting of equipment attribute information, and Y is the equipment tag.

[0023] Specifically, step S2 includes steps S2.1 to S2.3.

[0024] In S2.1, two base views, the adjacency matrix and the transfer matrix, are extracted from the original graph structure, and then elementary processing is performed on the two selected base views to obtain view embeddings using a graph convolutional network (GCN). Q 1 =σ(GCN(V1,X)) Q 2 =σ(GCN(V2,X)) where σ is a nonlinear activation function, which introduces a nonlinear transformation and improves the network's representation ability and ability to learn complex data.

[0025] In S2.2, the node embedding is used to recalculate the probability of edges in the graph structure. For the target node, connect it with other node embeddings, then normalize the node weights to obtain the connection probability between node pairs. This process is expressed as follows: W ij =W·[q i ||q j ]+a JPEG0007795843000004.jpg1752Here, W represents the connection weight between nodes i and j, q represents the node weight, and a is an offset vector.

[0026] In S2.3, a probability matrix p1 is obtained based on the probability that an edge appears for each node pair. The matrix is ​​combined with the original structure to obtain the processed base view. This process is expressed as follows: JPEG0007795843000005.jpg841where, λ 1 is the fusion coefficient, the value of which is related to the type of dataset.

[0027] Specifically, step S3 includes steps S3.1 to S3.3.

[0028] In S3.1, we fuse views using a method to assign view weights based on prediction confidence. First, we use a graph convolutional network (GCN) to obtain the prediction results of two base views. JPEG0007795843000006.jpg13128JPEG0007795843000007.jpg12128Here, σ is a nonlinear activation function.

[0029] The view weight reflects the uncertainty based on the predicted structural distribution. Specifically, for the same maximum value, the larger the difference between the maximum value and other values, the smaller the uncertainty, and therefore the larger the difference, the higher the view weight. For example, between [0.7, 0.1, 0.2] and [0.4, 0.35, 0.3], the former is more different from the maximum value and should be given a higher weight. Similarly, for the same maximum value, the view with a larger difference between the maximum value and the second-largest value should have a higher weight. For example, between [0.8, 0.7, 0.3] and [0.8, 0.5, 0.1], the latter should be given a higher weight.

[0030] In S3.2, based on the above principle, the weight of the view represented by H in the fusion process is calculated as follows: H vi =e β (γlogn 1 +(1-γ)logn 1 -n 2 ) where n1 and n 2 represent the largest and second largest predictive distributions, and β and γ are hyperparameters.

[0031] In S3.3, the weights are normalized, JPEG0007795843000008.jpg16128Based on the calculated weights, generate the final view of the node. JPEG0007795843000009.jpg853

[0032] Specifically, step S4 includes steps S4.1 and S4.2.

[0033] In S4.1, we calculate the attention coefficient between node pairs, e ij =d([Mh i ||Mh j ]),j∈N i where M is a shared parameter, h represents the node feature, and d is a real parameter for projecting the connected high-dimensional features into a real field. The result requires normalization, and LeakyReLU is selected as the activation parameter. JPEG0007795843000010.jpg1885Here, k represents all neighboring nodes of a node.

[0034] In S4.2, the attention coefficient of each node pair is multiplied by the original node feature, and the weights are added to obtain new node features. JPEG0007795843000011.jpg1461The above process is performed for each node in the three views (two base views and one fused view), self-adaptively updating the node features in the graph structure so that the node features can aggregate information hidden in the relationships between nodes.

[0035] Specifically, step S5 includes steps S5.1 to S5.4.

[0036] In S5.1, the GCN network structure is used to construct the observation set, and the aggregation rules of the k-th layer of the GCN are as follows: JPEG0007795843000012.jpg1284JPEG0007795843000013.jpg627 represents the normalized adjacency matrix, D is a diagonal matrix, σ is a nonlinear activation function, and H (k-1) and H (k) represents the node features of the k-1th layer and the kth layer, and W(k) is the weight matrix for layer k, which facilitates the linear transformation of the nodes.

[0037] After each aggregation, the current node represents capturing the structural information of this order, and the extraction node represents constructing a kNN subgraph, placing the original image into it, to form the following multi-level observation set: JPEG0007795843000014.jpg856

[0038] In S5.2, the probabilistic block model (SBM) is used as a generative model, and the optimal graph structure G is simulated. The process of generating the simulated optimal graph structure G is probabilistically formalized as follows: JPEG0007795843000015.jpg17128JPEG0007795843000016.jpg1449Here, Ω is a parameter of SBM, and Ω cicj represents the probability that an edge exists between any node v in community c and node vj in community cj, variable Z represents the node-specific feature set in the original dataset, and YL corresponds to the node tag set.

[0039] In S5.3, the simulated optimal graph structure currently generated by SBM is mapped to the observation set. The rule is to compare the observed graph with the high-quality graph. If both graphs have one edge, the probability of this situation occurring is called the true goodness rate p. If one edge in the observed graph does not exist in the high-quality graph, the probability of this situation occurring is called the false goodness rate q. Conversely, the true error rate is 1-p, and the false error rate is 1-q.

[0040] According to this rule, the probability that a high-quality graph G is mapped to an observed graph O is expressed as follows: JPEG0007795843000017.jpg52151

[0041] In S5.4, each mapping between the SBM graph and the observation set produces a different possible value of the high-quality graph. Adding these values ​​together, we obtain the following parameterized posterior probability: JPEG0007795843000018.jpg13128 Compute the adjacency matrix Q of the high quality graph by maximizing the posterior probability, which is then used to construct the determined high quality graph as follows: JPEG0007795843000019.jpg1347This structure is input into the downstream task to obtain the equipment fault detection results. Example 2

[0042] This embodiment provides a fault diagnosis system for industrial equipment based on integrated optimization of graph structure, and the system includes: a data acquisition module configured to acquire an original equipment dataset and construct an original graph structure based on the original equipment dataset; an embedding module configured to extract two base views based on the original graph structure, calculate graph node embeddings of the base views using a GCN, and recalculate the probabilities of edges in the graph structure based on the graph node embeddings; a fusion module configured to perform view fusion based on the probabilities of edges in the graph structure to obtain a preliminary optimized view; an enhancement module configured to process the fused view by the GAT network to obtain an enhanced view; and an optimization module configured to adopt the probabilistic block model as a generative model and iteratively optimize the reinforced view using Bayesian inference and expectation maximization algorithms to obtain a final graph structure.

[0043] A computer-readable storage medium having a plurality of instructions stored thereon, the instructions being suitable for being loaded by a processor of a terminal device to execute the method for fault diagnosis of industrial equipment based on the joint optimization of the graph structure.

[0044] A terminal device including a processor used to realize each instruction and a computer-readable storage medium used to store a plurality of instructions, the instructions being loaded by the processor and suitable for executing the fault diagnosis method for industrial equipment based on the integrated optimization of the graph structure.

[0045] The above are all preferred embodiments of the present invention, and do not limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape and principle of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fault diagnosis method for industrial equipment based on integrated optimization of a graph structure, comprising: obtaining an original equipment data set; constructing an original graph structure based on the original equipment dataset; Extracting two base views based on the original graph structure, calculating graph node embeddings of the base views using GCN, and recalculating the probabilities of edges in the graph structure based on the graph node embeddings; performing view fusion based on the probability of edges in the graph structure to obtain a first optimized view; processing the fused view by a GAT network to obtain an enhanced view; adopting a probabilistic block model as a generative model, and iteratively optimizing the reinforced views using Bayesian estimation and expectation maximization algorithm to obtain a final graph structure; The step of constructing an original graph structure based on the original equipment data set includes: creating the collected original equipment data into a graph data set for network training, with the equipment as nodes, the connection forms between the equipment as edges, and various data and attributes of the equipment as node features; The step of extracting two base views based on the original graph structure and calculating graph node embeddings of the base views using a graph convolution network (GCN) includes extracting two base views, i.e., an adjacency matrix and a transfer matrix, from the original graph structure, performing elementary processing on the selected two base views, and obtaining view embeddings using a graph convolution network (GCN); Q １ =σ(GCN(V １ ,X)) Q ２ =σ(GCN(V ２ ,X)) where σ is a nonlinear activation function, which introduces a nonlinear transformation to improve the network's representation ability and ability to learn complex data. The step of recalculating the probability of edges in the graph structure based on the graph node embedding includes: for a target node, connecting its node embedding with other node embeddings, then normalizing the node weights to calculate the probability of an edge existing between the node pairs, combining with the overall probability to obtain a probability matrix, and combining the probability matrix with the original graph structure to obtain a processed base view; The step of processing the fused view through the GAT network to obtain an enhanced view includes using a self-attention mechanism to enhance interactions and feature representations between nodes, and dynamically evaluating, weighting, and aggregating features of neighboring nodes, thereby generating an enhanced view with more information content; The step of enhancing the interactions and feature representations between nodes using a self-attention mechanism includes calculating an attention coefficient between a pair of nodes; s ｉｊ J(N) ｉ ||((). ｊ N),E∈9 ｉ where M is a shared parameter, h represents the node feature, and d is a real parameter for projecting the connected high-dimensional features onto a real field. The step of adopting a probabilistic block model as a generative model includes using a probabilistic block model SBM as a generative model and simulating an optimal graph structure, and the process of generating a simulated optimal graph structure G is probabilistically formalized as follows: where Ω is a parameter of the SBM, and Ω ｃｉｃｊ represents the probability that an edge exists between any node v in community c and node vj in community cj, variable Z represents a node-specific feature set in the original dataset, and YL corresponds to a node tag set.

2. A fault diagnosis system for industrial equipment based on integrated optimization of graph structures that executes the method of claim 1, a data acquisition module configured to acquire an original equipment dataset and construct an original graph structure based on the original equipment dataset; an embedding module configured to extract two base views based on the original graph structure, calculate graph node embeddings of the base views using GCN, and recalculate the probabilities of edges in the graph structure based on the graph node embeddings; a fusion module configured to perform view fusion based on the probabilities of edges in the graph structure to obtain a preliminary optimized view; an enhancement module configured to process the fused view by the GAT network to obtain an enhanced view; and an optimization module configured to adopt a probabilistic block model as a generative model, and to iteratively optimize the reinforced view using Bayesian estimation and an expectation maximization algorithm to obtain a final graph structure.

3. 10. A computer-readable storage medium having stored thereon a plurality of instructions, said instructions being suitable for being loaded by a processor of a terminal device to perform the method of claim 1.

4. 10. A terminal device comprising: a processor adapted to implement instructions; and a computer-readable storage medium adapted to store a plurality of instructions, the instructions being loaded by the processor and adapted to perform the method of claim 1.

Images