CN-120974433-B - Industrial equipment fault detection method, device and equipment based on vertical domain large model

CN120974433BCN 120974433 BCN120974433 BCN 120974433BCN-120974433-B

Abstract

The application discloses an industrial equipment fault detection method, device and equipment based on a vertical domain large model, and relates to the field of artificial intelligence; the method comprises the steps of processing processed data into a fault type result through a fault detection model, wherein the fault detection model comprises a structural perception enhancement module and a multi-mode fusion module which are connected with each other, the fault detection model is obtained through training based on fault sample data and target pseudo-samples, part of the fault sample data is marked with fault label results, the target pseudo-samples are obtained according to the fault label results and the fault sample data by taking acquired domain prior data as constraint parameters, the structural perception enhancement module is used for carrying out feature enhancement processing on the structural data to obtain structural feature vectors, and the multi-mode fusion module is used for carrying out semantic fusion processing on the structural feature vectors, time sequence data, image data and text data. The application improves the fault recognition precision.

Inventors

XIAO BO
ZHANG MAOSEN
LIANG ZHUOSHAN
HUANG QIUMING
Deng Daner

Assignees

广东知业科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251016

Claims (8)

1. The industrial equipment fault detection method based on the vertical domain large model is characterized by comprising the following steps of: acquiring multi-mode data to be detected in the operation process of target industrial equipment; preprocessing the multi-mode data to be detected to obtain processed data, wherein the processed data comprises structural data, time sequence data, image data and text data; The fault detection model comprises a structural perception enhancement module and a multi-mode fusion module which are connected with each other, wherein the fault detection model is obtained by training based on fault sample data and target pseudo-samples, part of the fault sample data is marked with fault label results, the target pseudo-samples are obtained by taking acquired field priori data as constraint parameters and according to the fault label results and the fault sample data, the structural perception enhancement module is used for carrying out feature enhancement processing on the structural data to obtain structural feature vectors, and the multi-mode fusion module is used for carrying out semantic alignment and fusion processing on the structural feature vectors, the time sequence data, the image data and the text data; the fault detection model further comprises a fault classification module, wherein the fault classification module is connected with the multi-mode fusion module; Processing the processed data in a fault detection model to obtain a fault class result, wherein the processing comprises the following steps: the structural data is subjected to feature extraction and component relation strengthening treatment through the structural perception enhancement module to obtain a structural feature vector; performing feature extraction, modal deletion repair and semantic alignment processing on the image data, the time sequence data and the text data through the multi-modal fusion module to obtain a joint representation vector with consistent semantics; Performing fault classification processing on the joint representation vector through the fault classification module to obtain a fault class result; the structural data is subjected to feature extraction and component relation strengthening treatment through the structural perception enhancement module to obtain structural feature vectors, and the method comprises the following steps: Carrying out topology modeling on the structural data to construct a topology graph, wherein the topology graph comprises nodes and edges, the nodes are used for representing component information in equipment, and the edges are used for representing structural information among components; Coding the structural information through a graph neural network to obtain a structural embedded vector; for hierarchical structure data in the structure data, acquiring hierarchical information of a subsystem where a component is located and position information in the topological graph; Generating a hierarchical position code according to the hierarchical information and the position information; adding the hierarchical position code into the component information and carrying out fusion processing on the hierarchical position code and the component information to obtain a structural fusion vector; and carrying out attention distribution processing on the structure fusion vector and the structure embedding vector through an attention layer to obtain the structure feature vector.
2. The industrial equipment fault detection method based on the vertical domain large model according to claim 1, wherein the multi-mode fusion module comprises an encoder, a missing repair unit and a semantic alignment unit which are connected in sequence; performing feature extraction, modal deletion repair and semantic alignment processing on the image data, the time sequence data and the text data through the multi-modal fusion module to obtain a joint representation vector with consistent semantics, wherein the joint representation vector comprises: processing the image data, the time sequence data and the text data through corresponding encoders respectively, and extracting an image feature vector, a time sequence feature vector and a text feature vector; determining a missing vector from the structural feature vector, the image feature vector, the time sequence feature vector and the text feature vector by the missing repairing unit, and performing modal repairing treatment on the missing vector to obtain a target vector, wherein the target vector comprises a target image vector, a target time sequence vector, a target text vector and a target structural vector; and aligning the target image vector, the target time sequence vector, the target text vector and the target structure vector in a common semantic space by the semantic alignment unit, and extracting a joint representation vector with consistent semantics.
3. The industrial equipment fault detection method based on the vertical domain large model according to claim 1, wherein the fault detection model is constructed by the steps of: obtaining a target pseudo sample and original equipment data, and preprocessing the original equipment data into fault sample data, wherein the fault sample data comprises first data marked with a fault label result, second data not marked with the fault label result and equipment operation data; Acquiring equipment attribute data from a preset industry knowledge base, and vectorizing the equipment attribute data into a knowledge vector; Dividing the knowledge vector, the fault sample data and the target pseudo sample into a training set and a verification set according to a preset dividing rule; Inputting the training set into an initial model to obtain a fault output result; Constructing a loss function according to the fault output result and the fault label result, minimizing the loss function, and performing iterative optimization on parameters of each module in the initial model by adopting a small sample learning optimization strategy to obtain a model to be verified; and inputting the verification set into the model to be verified for verification to obtain a fault detection model.
4. The method for detecting industrial equipment failure based on the vertical domain large model according to claim 3, wherein obtaining the target pseudo-sample comprises: acquiring domain prior data and converting the domain prior data into constraint parameters, wherein the constraint parameters comprise conditional constraint parameters, latent variable distribution constraint parameters and multi-mode consistency rule parameters; Inputting the fault label result and fault sample data into an encoder to be processed by taking the constraint parameter as a generation condition to obtain pseudo sample data, wherein the pseudo sample data is configured with a weight value and a sample parameter; And calculating confidence coefficient according to the weight value and the sample parameter, filtering the pseudo sample data with the confidence coefficient smaller than a preset threshold value from all the pseudo sample data, and obtaining the target pseudo sample through comparison learning constraint processing, wherein the comparison learning constraint is used for shortening the semantic distance between the pseudo sample data and the fault label result.
5. The industrial equipment fault detection method based on the vertical domain large model according to claim 3, wherein the initial model comprises an initial structure perception enhancement module, an initial multi-mode fusion module and an initial fault classification module which are connected in sequence, the training set is input into the initial model to obtain a fault output result, and the method comprises the following steps: carrying out structure enhancement processing on the structure sample data of the training set through the initial structure perception enhancement module to obtain a structure sample vector; Performing feature extraction, modal deletion repair and semantic alignment processing on other sample data in the training set and the structural sample vector through the initial multi-modal fusion module to obtain a combined sample vector; and classifying the combined sample vector through the initial fault classification module to obtain the fault output result.
6. The industrial equipment fault detection method based on the vertical domain large model according to claim 1, wherein preprocessing the multi-mode data to be detected to obtain processed data comprises: performing time stamp alignment processing on the multi-mode data to be detected to obtain aligned data; Performing null value filling processing on the aligned data to obtain filled data, wherein the filled data comprises time sequence filling data and other filling data; Performing segmentation slicing processing on the time sequence filling data to obtain a plurality of continuous fragments with fixed lengths; And carrying out feature standardization processing according to the continuous fragments and the other filling data to obtain the processed data.
7. An industrial equipment fault detection device based on a vertical domain large model, which is characterized by comprising: The acquisition module is used for acquiring multi-mode data to be detected in the operation process of the target industrial equipment; The preprocessing module is used for preprocessing the multi-mode data to be detected to obtain processed data, wherein the processed data comprises structural data, time sequence data, image data and text data; The fault detection module is used for processing the processed data through a fault detection model to obtain a fault class result, the fault detection model comprises a structural perception enhancement module and a multi-mode fusion module which are connected with each other, the fault detection model is obtained based on fault sample data and target pseudo-sample training, the fault sample data is partially marked with a fault label result, the target pseudo-sample is obtained by taking acquired field priori data as constraint parameters and according to the fault label result and the fault sample data, the structural perception enhancement module is used for carrying out feature enhancement processing on the structural data to obtain a structural feature vector, the multi-mode fusion module is used for carrying out semantic alignment and fusion processing on the structural feature vector, the time sequence data, the image data and the text data, and the fault detection model further comprises a fault classification module, and the fault classification module is connected with the multi-mode fusion module; The fault detection module is specifically configured to: the structural data is subjected to feature extraction and component relation strengthening treatment through the structural perception enhancement module to obtain a structural feature vector; performing feature extraction, modal deletion repair and semantic alignment processing on the image data, the time sequence data and the text data through the multi-modal fusion module to obtain a joint representation vector with consistent semantics; Performing fault classification processing on the joint representation vector through the fault classification module to obtain a fault class result; The fault detection module is further configured to: Carrying out topology modeling on the structural data to construct a topology graph, wherein the topology graph comprises nodes and edges, the nodes are used for representing component information in equipment, and the edges are used for representing structural information among components; Coding the structural information through a graph neural network to obtain a structural embedded vector; for hierarchical structure data in the structure data, acquiring hierarchical information of a subsystem where a component is located and position information in the topological graph; Generating a hierarchical position code according to the hierarchical information and the position information; adding the hierarchical position code into the component information and carrying out fusion processing on the hierarchical position code and the component information to obtain a structural fusion vector; and carrying out attention distribution processing on the structure fusion vector and the structure embedding vector through an attention layer to obtain the structure feature vector.
8. A computer device comprising a memory, a processor to store a computer program executable on the processor, characterized in that the processor executes the computer program to implement the steps of the method for industrial device fault detection based on a vertical domain large model as claimed in any one of claims 1-6.

Description

Industrial equipment fault detection method, device and equipment based on vertical domain large model Technical Field The application relates to the technical field of artificial intelligence, in particular to an industrial equipment fault detection method, device and equipment based on a vertical domain large model. Background With the rapid development of industrial technology, industrial equipment is used as a core carrier of a manufacturing system in the manufacturing industry, and the running state of the industrial equipment directly determines the production efficiency, the product quality and the operation safety. In the advanced pushing process of industrialization and intelligent manufacturing, key equipment such as a numerical control machine tool, a heavy motor, wind power equipment, a rail transit traction system and the like develop towards high integration, high automation and high complexity, and once the equipment fails, the equipment can be stopped to cause huge economic loss, and safety accidents such as equipment damage, casualties and the like can be caused. Therefore, in order to ensure continuous and stable operation of industrial production, how to accurately, efficiently and early diagnose the faults of industrial equipment is important. At present, in an industrial production scene, a general large model is adopted to carry out fault diagnosis on industrial data in the related technology, however, the model is difficult to adapt to a vertical domain scene such as a complex equipment structure, changeable industrial working conditions, complex data modes and the like, and particularly, under the condition that the number of equipment fault data samples is small, the general large model is adopted to diagnose faults, so that the false alarm rate is high, and the fault recognition precision is poor. Disclosure of Invention The application aims to provide an industrial equipment fault detection method, device and equipment based on a vertical domain large model. In order to achieve the above object, the present application provides the following solutions: in a first aspect, the present application provides an industrial equipment fault detection method based on a vertical domain large model, including: acquiring multi-mode data to be detected in the operation process of target industrial equipment; preprocessing the multi-mode data to be detected to obtain processed data, wherein the processed data comprises structural data, time sequence data, image data and text data; The fault detection model comprises a structure perception enhancement module and a multi-mode fusion module which are connected with each other, the fault detection model is obtained by training based on fault sample data and target pseudo-samples, fault label results are marked in part of the fault sample data, the target pseudo-samples are obtained by taking acquired field priori data as constraint parameters and according to the fault label results and the fault sample data, the structure perception enhancement module is used for carrying out feature enhancement processing on the structure data to obtain structure feature vectors, and the multi-mode fusion module is used for carrying out semantic alignment and fusion processing on the structure feature vectors, the time sequence data, the image data and the text data. In a second aspect, the present application provides an industrial equipment fault detection device based on a vertical domain large model, the device comprising: The acquisition module is used for acquiring multi-mode data to be detected in the operation process of the target industrial equipment; The preprocessing module is used for preprocessing the multi-mode data to be detected to obtain processed data, wherein the processed data comprises structural data, time sequence data, image data and text data; The fault detection module is used for processing the processed data through a fault detection model to obtain a fault type result, the fault detection model comprises a structure perception enhancement module and a multi-mode fusion module which are connected with each other, the fault detection model is obtained by training based on fault sample data and target pseudo-samples, fault label results are marked in part of the fault sample data, the target pseudo-samples are obtained by taking acquired field priori data as constraint parameters and according to the fault label results and the fault sample data, the structure perception enhancement module is used for carrying out feature enhancement processing on the structure data to obtain a structure feature vector, and the multi-mode fusion module is used for carrying out semantic alignment and fusion processing on the structure feature vector, the time sequence data, the image data and the text data. In a third aspect, the present application provides a computer device comprising a memory, a processor to store a computer program on the memory and