Industrial equipment real-time fault diagnosis method and system based on generative model

Abstract

The invention relates to the technical field of industrial equipment fault diagnosis, and provides an industrial equipment real-time fault diagnosis method and system based on a generative model, and the method comprises the following steps: collecting the original operation data of industrial equipment in real time through a plurality of sensors, carrying out the preprocessing of the original operation data, obtaining the real-time operation data, and carrying out the calculation of the real-time operation data; the preprocessing comprises abnormal value detection and standardization; constructing a generative model based on the real-time operation data, generating a normal data sample of the industrial equipment in a normal state through the generative model, and integrating the real-time operation data and the normal data sample to obtain a multi-dimensional data sample; performing feature extraction on the multi-dimensional data sample by using a deep learning model to obtain a fault diagnosis vector; performing fault mode identification on the fault diagnosis vector through a classification model to obtain a fault diagnosis result, and performing state judgment on the fault diagnosis result according to a preset threshold to obtain an industrial equipment state; and the state and the fault diagnosis result of the industrial equipment are displayed in real time through a dynamic monitoring interface. According to the method, the real-time performance and the overall efficiency of fault diagnosis are improved, and the loss caused by equipment faults can be effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment fault diagnosis, and in particular to a real-time fault diagnosis method and system for industrial equipment based on a generative model. Background Art

[0002] In modern industrial production, the normal operation of industrial equipment is the key to ensuring production efficiency and product quality. However, various faults may occur during the operation of the equipment. If these faults are not discovered and handled in time, they will lead to production stagnation, equipment damage and economic losses.

[0003] Existing technologies often use historical fault data and expert experience for fault diagnosis. However, when data is insufficient or the equipment operating status is complex, it is impossible to obtain enough samples for training, which will lead to model overfitting and difficulty in achieving accurate fault identification. Summary of the invention

[0004] In view of this, the present invention proposes a real-time fault diagnosis method and system for industrial equipment based on a generative model, which solves the problem that the prior art cannot accurately identify faults of industrial equipment when data is insufficient or the equipment operating status is complex.

[0005] The technical solution of the present invention is implemented as follows: In a first aspect, the present invention provides a real-time fault diagnosis method for industrial equipment based on a generative model, comprising the following steps:

[0006] S1, collecting original operation data of industrial equipment in real time through multiple sensors, preprocessing the original operation data to obtain real-time operation data, wherein the preprocessing includes outlier detection and standardization;

[0007] S2, constructing a generative model based on the real-time operation data, generating a normal data sample of the industrial equipment in a normal state through the generative model, and integrating the real-time operation data and the normal data sample to obtain a multi-dimensional data sample;

[0008] S3, extracting features from the multidimensional data sample using a deep learning model to obtain a fault diagnosis vector;

[0009] S4, performing fault pattern recognition on the fault diagnosis vector through a classification model to obtain a fault diagnosis result, and performing state judgment on the fault diagnosis result according to a preset threshold to obtain an industrial equipment state;

[0010] S5, displaying the industrial equipment status and the fault diagnosis result in real time through a dynamic monitoring interface.

[0011] On the basis of the above technical solution, preferably, step S1 includes:

[0012] S11, setting a data collection frequency, and collecting original operating data of the industrial equipment in real time at the data collection frequency through multiple sensors, wherein the original operating data includes operating temperature, operating voltage, operating power, vibration frequency and operating current;

[0013] S12, performing outlier detection on the original operation data by statistical methods, identifying and eliminating outliers beyond a preset range, obtaining outlier-eliminated operation data, and unifying the data scale of the original operation data by Z-score standardization to obtain real-time operation data.

[0014] Based on the above technical solution, preferably, step S2 includes:

[0015] S21, constructing an initial generative model based on a generative adversarial network, and training the initial generative model through the real-time operation data to obtain a generative model;

[0016] S22, inputting random noise into the generative model to obtain original sample data similar to a normal state, evaluating the quality of the original sample data, retaining original sample data corresponding to an evaluation value greater than a preset evaluation value, and obtaining normal sample data;

[0017] S23, arranging the real-time operation data and the normal data samples according to the time series, and performing unified data dimension processing to obtain multi-dimensional data samples.

[0018] On the basis of the above technical solution, preferably, the calculation formula of the generative model includes:

[0019]

[0020] Where V(D(x;θ d ),G(z;θ g )) is the objective function of the generative adversarial network, G(z;θ g ) is the generating function, D(x;θ d ) is the discriminant function, θ d is the learnable parameter of the discriminant function, θ g is the learnable parameter of the generating function, x is the real-time running data, z is the random noise, P data (x) is the real data distribution of x, P z (z) is the random noise distribution, is the expected value of the discriminator’s predicted probability of x, is the discriminator pair generating function G(z;θ g ), λ is the regularization parameter, ||G(z;θ g)|| 2 is the generating function G(z;θ g ) is the L2 norm of the output sample, is the expected value of the L2 norm of the sample generating function.

[0021] On the basis of the above technical solution, preferably, step S3 includes:

[0022] S31, constructing an initial deep learning model, wherein the initial deep learning model adopts a convolutional neural network model, a recurrent neural network model or a long short-term memory network model, obtains annotated historical fault diagnosis data, performs model training on the initial deep learning model through the historical fault diagnosis data, optimizes model parameters, and obtains a trained deep learning model;

[0023] S32, inputting the multidimensional data sample into the deep learning model, performing forward propagation to obtain feature representation, and generating a fault diagnosis vector through the feature representation.

[0024] Based on the above technical solution, preferably, step S4 includes:

[0025] S41, constructing an initial classification model through a classification algorithm, wherein the classification algorithm includes a support vector machine algorithm, a decision tree algorithm, a random forest algorithm, and a logistic regression algorithm, obtaining a labeled historical fault diagnosis vector, and training the initial classification model through the labeled historical fault diagnosis vector to optimize model parameters;

[0026] S42, inputting the fault diagnosis vector into the classification model, and performing forward propagation to obtain a fault diagnosis result;

[0027] S43, setting a preset threshold, comparing the fault diagnosis result with the preset threshold, and obtaining the industrial equipment status.

[0028] On the basis of the above technical solution, preferably, the calculation formula of the classification model includes:

[0029]

[0030] in, is the fault diagnosis result, f(·) is the sigmoid activation function, X is the fault diagnosis vector, θ is the model parameter of the classification model, W is the model weight vector, and b is the bias;

[0031] The calculation formula for the state judgment includes:

[0032]

[0033] Among them, W status For industrial equipment status, W1status , They are the first industrial equipment status, the second industrial equipment status, and the third industrial equipment status. and They are the first preset threshold and the second preset threshold respectively.

[0034] In a second aspect, the present invention further provides a real-time fault diagnosis system for industrial equipment based on a generative model, the system comprising:

[0035] A data processing module, used to collect raw operating data of industrial equipment in real time through multiple sensors, and pre-process the raw operating data to obtain real-time operating data, wherein the pre-processing includes outlier detection and standardization;

[0036] A sample integration module, used to build a generative model based on the real-time operation data, generate a normal data sample of the industrial equipment in a normal state through the generative model, and integrate the real-time operation data and the normal data sample to obtain a multi-dimensional data sample;

[0037] A feature extraction module, used to extract features from the multidimensional data sample using a deep learning model to obtain a fault diagnosis vector;

[0038] A fault identification and state judgment module is used to perform fault pattern recognition on the fault diagnosis vector through a classification model to obtain a fault diagnosis result, and perform state judgment on the fault diagnosis result according to a preset threshold to obtain the state of the industrial equipment;

[0039] The monitoring and display module is used to display the industrial equipment status and the fault diagnosis result in real time through a dynamic monitoring interface.

[0040] In a third aspect, the present invention further provides an electronic device, comprising: at least one processor, at least one memory, a communication interface and a bus;

[0041] The processor, memory, and communication interface communicate with each other through the bus, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement steps of a real-time fault diagnosis method for industrial equipment based on a generative model.

[0042] In a fourth aspect, the present invention further provides a computer-readable storage medium storing computer instructions, wherein the computer instructions enable a computer to implement steps of a real-time fault diagnosis method for industrial equipment based on a generative model.

[0043] The real-time fault diagnosis method and system for industrial equipment based on generative models of the present invention have the following beneficial effects compared with the prior art:

[0044] (1) Real-time data collection by multiple sensors and generation of normal state samples using generative models are combined with deep learning models for feature extraction and fault pattern recognition. A dynamic monitoring interface is used to display equipment status and diagnostic results in real time, and a systematic fault diagnosis process is formed. In the case of insufficient data or complex equipment operation status, the real-time and accuracy of fault diagnosis is guaranteed, the overall efficiency and reliability are improved, and the losses caused by equipment failures can be effectively reduced;

[0045] (2) By constructing and optimizing the generative model and combining it with the training mechanism of the generative adversarial network, high-quality normal sample data is generated, which enhances the adaptability and accuracy of fault diagnosis in cases where data is insufficient or the equipment is in a complex operating state, thereby improving the efficiency of real-time fault diagnosis of industrial equipment;

[0046] (3) By constructing a classification model and introducing the sigmoid activation function and a multi-level state judgment mechanism, accurate classification of fault diagnosis results and real-time state feedback are achieved, thereby improving the accuracy of fault pattern recognition and the flexibility of equipment state judgment, and enhancing the fault diagnosis capability and response speed of industrial equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 A flow chart of a method for real-time fault diagnosis of industrial equipment based on a generative model of the present invention;

[0049] Figure 2 This is a structural diagram of a real-time fault diagnosis system for industrial equipment based on a generative model according to the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] See also Figure 1 The present invention provides a real-time fault diagnosis method for industrial equipment based on a generative model, comprising the following steps:

[0052] S1, collecting original operation data of industrial equipment in real time through multiple sensors, preprocessing the original operation data to obtain real-time operation data, wherein the preprocessing includes outlier detection and standardization;

[0053] S2, building a generative model based on the real-time operation data, generating a normal data sample of the industrial equipment in a normal state through the generative model, and integrating the real-time operation data and the normal data sample to obtain a multi-dimensional data sample;

[0054] S3, extracting features from the multidimensional data sample using a deep learning model to obtain a fault diagnosis vector;

[0055] S4, performing fault pattern recognition on the fault diagnosis vector through a classification model to obtain a fault diagnosis result, and performing state judgment on the fault diagnosis result according to a preset threshold to obtain an industrial equipment state;

[0056] S5, displaying the industrial equipment status and the fault diagnosis result in real time through a dynamic monitoring interface.

[0057] Specifically, this embodiment collects data in real time through multiple sensors and uses a generative model to generate normal state samples, combines a deep learning model to perform feature extraction and fault pattern recognition, uses a dynamic monitoring interface to display the equipment status and diagnosis results in real time, and forms a systematic fault diagnosis process. In the case of insufficient data or complex equipment operating conditions, the real-time and accuracy of fault diagnosis is achieved, the overall efficiency and reliability are improved, and the losses caused by equipment failures can be effectively reduced.

[0058] Step S1 includes:

[0059] S11, setting a data collection frequency, and collecting original operating data of the industrial equipment in real time at the data collection frequency through multiple sensors, wherein the original operating data includes operating temperature, operating voltage, operating power, vibration frequency and operating current;

[0060] S12, performing outlier detection on the original operation data by statistical methods, identifying and eliminating outliers beyond a preset range, obtaining outlier-eliminated operation data, and unifying the data scale of the original operation data by Z-score standardization to obtain real-time operation data.

[0061] In a specific embodiment, a suitable data collection frequency is set according to the operating characteristics of the industrial equipment, for example, data is collected once per second, to ensure that dynamic changes during the operation of the equipment can be captured.

[0062] The original operating data of industrial equipment is collected in real time at a set frequency through multiple sensors (such as temperature sensors, voltage sensors, power sensors, vibration sensors and current sensors). These data include key parameters such as operating temperature, operating voltage, operating power, vibration frequency and operating current.

[0063] Statistical methods are used to detect outliers in the collected raw operating data, identify and eliminate outliers that exceed the preset range, and ensure the accuracy and reliability of the data.

[0064] The original operation data after removing outliers is standardized (such as Z-score standardization) to unify the data scale and obtain real-time operation data.

[0065] Specifically, this embodiment ensures that the collected original operation data is of high quality by setting a reasonable data collection frequency and outlier detection, thereby reducing the risk of misdiagnosis due to data noise.

[0066] Real-time data collection and processing enables timely monitoring of equipment operating status, facilitating rapid response to potential faults and reducing equipment downtime.

[0067] Data standardization processing enables data collected by different sensors to be compared and analyzed at the same scale. This embodiment can adapt to the operating characteristics of different types of industrial equipment and flexibly adjust the data collection frequency and abnormal value detection standards.

[0068] Step S2 includes:

[0069] S21, constructing an initial generative model based on a generative adversarial network, and training the initial generative model through the real-time operation data to obtain a generative model;

[0070] S22, inputting random noise into the generative model to obtain original sample data similar to a normal state, evaluating the quality of the original sample data, retaining original sample data corresponding to an evaluation value greater than a preset evaluation value, and obtaining normal sample data;

[0071] S23, arranging the real-time operation data and the normal data samples according to the time series, and performing unified data dimension processing to obtain multi-dimensional data samples.

[0072] In a specific embodiment, an initial generative model is constructed based on a generative adversarial network, which consists of a generator and a discriminator, wherein the generator is responsible for generating sample data similar to a normal state, and the discriminator is used to determine the authenticity of the generated samples.

[0073] The initial generative model is trained using real-time running data. By continuously optimizing the parameters of the generator and discriminator, the generator can generate higher quality normal sample data, and the discriminator can more accurately distinguish between real data and generated data.

[0074] Input random noise into the trained generative model to generate original sample data similar to the normal state. Perform quality assessment on the generated sample data, retain samples with evaluation values ​​greater than the preset evaluation value, and obtain normal sample data.

[0075] The real-time operation data and normal sample data are arranged according to the time series, and the data dimensions are processed uniformly to finally obtain a multi-dimensional data sample.

[0076] Specifically, this embodiment can generate diversified normal sample data through the training of generative adversarial networks, enhance the model's adaptability to different operating states, and improve the accuracy of fault diagnosis. In the case of insufficient data, the generative model can effectively generate normal samples, make up for the lack of historical data, reduce the risk of model overfitting, and improve the reliability of fault identification.

[0077] By evaluating the quality of the generated samples, we ensure that the retained sample data has high authenticity and representativeness. By uniformly processing the arrangement and dimensions of real-time operation data and normal sample data, we ensure the consistency and comparability of the data and improve the efficiency of feature extraction.

[0078] The calculation formula of the generative model includes:

[0079]

[0080] Where V(D(x;θ d ),G(z;θ g )) is the objective function of the generative adversarial network, G(z;θ g ) is the generating function, D(x;θ d ) is the discriminant function, θ d is the learnable parameter of the discriminant function, θ g is the learnable parameter of the generating function, x is the real-time running data, z is the random noise, P data (x) is the real data distribution of x, P z (z) is the random noise distribution, is the expected value of the discriminator’s predicted probability of x, is the discriminator pair generating function G(z;θ g ), λ is the regularization parameter, ||G(z;θ g )|| 2 is the generating function G(z;θ g) is the L2 norm of the output sample, is the expected value of the L2 norm of the sample generating function.

[0081] Specifically, the generative adversarial network of this embodiment consists of two main parts: a generator and a discriminator. The goal of the generator is to generate sample data that is as realistic as possible, while the goal of the discriminator is to distinguish between real samples and generated samples. Through this adversarial training, the generator continuously optimizes its generation ability, and the discriminator continuously improves its judgment ability.

[0082] This embodiment defines the objective functions of the generator and the discriminator so that when generating samples, the generator can maximize the prediction probability of the discriminator for the generated samples, while minimizing the prediction probability of the discriminator for the real samples, thereby forming a good competitive relationship between the generator and the discriminator and promoting the joint optimization of the model.

[0083] Introducing regularization parameters and L2 norm constraints in the objective function can effectively prevent the overfitting of the model and ensure that the samples generated by the generator strike a balance between diversity and authenticity, making the generated samples more consistent with the operating status of actual industrial equipment.

[0084] By optimizing the objective function, the generator can generate higher quality sample data that are closer to the real normal state data. After introducing regularization and constraints, the model is more stable during the training process, reducing the phenomenon of model non-convergence caused by insufficient training or insufficient data.

[0085] Through the adversarial training mechanism, the competitive relationship between the generator and the discriminator promotes the joint optimization of both parties, so that the generator can continuously improve its generation ability, and the discriminator can more accurately identify the authenticity of the samples.

[0086] The improved objective function can adapt to different types of industrial equipment and operating conditions, enhancing the versatility and applicability of the model.

[0087] Step S3 includes:

[0088] S31, constructing an initial deep learning model, wherein the initial deep learning model adopts a convolutional neural network model, a recurrent neural network model or a long short-term memory network model, obtains annotated historical fault diagnosis data, performs model training on the initial deep learning model through the historical fault diagnosis data, optimizes model parameters, and obtains a trained deep learning model;

[0089] S32, inputting the multidimensional data sample into the deep learning model, performing forward propagation to obtain feature representation, and generating a fault diagnosis vector through the feature representation.

[0090] In a specific embodiment, a suitable deep learning architecture is selected, such as a convolutional neural network, a recurrent neural network, or a long short-term memory network, and a model is designed according to the fault characteristics and data types of the industrial equipment.

[0091] Collect and organize annotated historical fault diagnosis data, which should include feature vectors and corresponding fault labels under different fault modes for use in model training.

[0092] The initial deep learning model is trained using labeled historical fault diagnosis data. The model parameters are optimized and the network structure is adjusted through the back propagation algorithm to improve the model's ability to identify fault features.

[0093] The multidimensional data samples obtained in step S2 are input into the trained deep learning model for forward propagation to obtain feature representations, and a fault diagnosis vector is generated through these feature representations.

[0094] Specifically, this embodiment can effectively extract key features from multidimensional data samples by constructing a suitable deep learning model. By selecting different types of deep learning models (such as convolutional neural networks, recurrent neural networks, or long short-term memory networks), this embodiment can adapt to different fault characteristics and data types. By using labeled historical fault diagnosis data for model training, it can fully mine the information in historical data, and the feature representation obtained through forward propagation provides high-quality input for fault pattern recognition.

[0095] Step S4 includes:

[0096] S41, constructing an initial classification model through a classification algorithm, wherein the classification algorithm includes a support vector machine algorithm, a decision tree algorithm, a random forest algorithm, and a logistic regression algorithm, obtaining a labeled historical fault diagnosis vector, and training the initial classification model through the labeled historical fault diagnosis vector to optimize model parameters;

[0097] S42, inputting the fault diagnosis vector into the classification model, and performing forward propagation to obtain a fault diagnosis result;

[0098] S43, setting a preset threshold, comparing the fault diagnosis result with the preset threshold, and obtaining the industrial equipment status.

[0099] In a specific embodiment, a suitable classification algorithm is selected, such as support vector machine, decision tree, random forest or logistic regression, to construct an initial classification model. These algorithms can effectively process multi-dimensional data samples and perform fault pattern recognition.

[0100] Collect and organize annotated historical fault diagnosis vectors, which should contain feature representations of different fault modes and their corresponding fault labels for use in model training.

[0101] The labeled historical fault diagnosis vectors are used to train the initial classification model. By optimizing the model parameters, the model's ability to identify fault modes is improved, ensuring that the classifier can accurately distinguish different fault types.

[0102] The fault diagnosis vector generated in step S3 is input into the trained classification model for forward propagation to obtain the fault diagnosis result.

[0103] Set a preset threshold, compare the fault diagnosis result with the preset threshold, and determine the status of the equipment (such as normal, faulty, or waiting for maintenance, etc.).

[0104] Specifically, this embodiment can identify different fault modes by constructing a suitable classification model. By selecting a variety of classification algorithms (such as SVM, decision tree, etc.), this embodiment can be flexibly adjusted according to specific application scenarios and data characteristics. By using labeled historical fault diagnosis vectors for model training, it can fully mine the information in historical data. By setting preset thresholds for status judgment, it can provide timely feedback on the operating status of the equipment, facilitate operators to respond quickly, and reduce the risks brought by equipment failures.

[0105] The calculation formula of the classification model includes:

[0106]

[0107] in, is the fault diagnosis result, f(·) is the sigmoid activation function, X is the fault diagnosis vector, θ is the model parameter of the classification model, W is the model weight vector, and b is the bias;

[0108] The calculation formula for the state judgment includes:

[0109]

[0110] Among them, W status For industrial equipment status, W1 status , They are the first industrial equipment status, the second industrial equipment status, and the third industrial equipment status. and They are the first preset threshold and the second preset threshold respectively.

[0111] Specifically, this embodiment introduces a sigmoid activation function to map the fault diagnosis vector to a probability value range (usually 0 to 1), so that the result output by the model can be interpreted as the probability of a fault occurring, and the classification model can more intuitively reflect the fault status of the equipment.

[0112] In state judgment, by setting multiple preset thresholds (such as the first preset threshold and the second preset threshold), the fault diagnosis results can be divided into different states (such as normal, faulty, waiting for maintenance, etc.), so that the fault diagnosis results are more detailed and accurate. By optimizing the model parameters (such as the weight vector W and the bias b), the model's ability to recognize fault states can be improved.

[0113] This embodiment introduces the sigmoid activation function to interpret the fault diagnosis result as the probability of a fault occurring. By setting multiple preset thresholds, multi-level judgment of the equipment status is achieved, which facilitates taking corresponding measures according to different fault severities. By optimizing the model parameters, the model's ability to recognize fault states is improved, misjudgments and missed judgments are reduced, and the operating status of industrial equipment is fed back in time, helping operators to respond quickly and reducing the risks brought by equipment failures.

[0114] In a specific embodiment, the implementation steps of step S5 include:

[0115] Design a user-friendly dynamic monitoring interface that displays the real-time status of industrial equipment, fault diagnosis results, and related operating parameters (such as operating temperature, operating voltage, operating power, etc.). The dynamic monitoring interface should have good visualization effects to facilitate operators to quickly understand the equipment status.

[0116] The dynamic monitoring interface acquires the operation data and fault diagnosis results of the industrial equipment in real time, ensuring that the information on the monitoring interface can be updated in time to reflect the latest status of the equipment.

[0117] Set status indicators or icons on the monitoring interface to intuitively display the operating status of the equipment (such as normal, faulty, waiting for maintenance, etc.). For example, use green to indicate normal status, red to indicate faulty status, and yellow to indicate waiting for maintenance status.

[0118] Integrated alarm function: when the equipment status changes or the fault diagnosis result exceeds the preset threshold, the system can automatically trigger the alarm and notify the operator to handle it in time.

[0119] Provides historical data query function, allowing operators to view the historical operation data and fault diagnosis records of the equipment for fault analysis and decision support.

[0120] See also Figure 2The present invention also provides a real-time fault diagnosis system for industrial equipment based on a generative model, the system comprising:

[0121] A data processing module, used to collect raw operating data of industrial equipment in real time through multiple sensors, and pre-process the raw operating data to obtain real-time operating data, wherein the pre-processing includes outlier detection and standardization;

[0122] A sample integration module, used to build a generative model based on the real-time operation data, generate a normal data sample of the industrial equipment in a normal state through the generative model, and integrate the real-time operation data and the normal data sample to obtain a multi-dimensional data sample;

[0123] A feature extraction module, used to extract features from the multidimensional data sample using a deep learning model to obtain a fault diagnosis vector;

[0124] A fault identification and state judgment module is used to perform fault pattern recognition on the fault diagnosis vector through a classification model to obtain a fault diagnosis result, and perform state judgment on the fault diagnosis result according to a preset threshold to obtain the state of the industrial equipment;

[0125] The monitoring and display module is used to display the industrial equipment status and the fault diagnosis result in real time through a dynamic monitoring interface.

[0126] Specifically, this embodiment collects and pre-processes the original operation data of the industrial equipment in real time through the data processing module, so as to ensure that the operation status of the equipment can be obtained in time.

[0127] The system uses a generative model to generate normal sample data, and combines it with a deep learning model to perform feature extraction and fault pattern recognition, which can effectively identify the fault status of the equipment. The sample integration module integrates the real-time operation data and the generated normal sample data into a multi-dimensional data sample, enhancing the richness and diversity of the data. The monitoring and display module displays the equipment status and fault diagnosis results in real time through a dynamic monitoring interface, and can notify the operator in time in combination with the alarm system.

[0128] The system can adapt to different types of industrial equipment and complex operating conditions. By combining generative models and deep learning models, the system's versatility is enhanced and it is suitable for a variety of industrial scenarios. The system can store and trace back historical fault diagnosis data, provide fault analysis and decision support for operators, and help them formulate effective maintenance and management strategies. Through real-time monitoring and fault diagnosis, it can promptly detect and handle potential faults, reduce equipment downtime and maintenance costs, and improve production efficiency.

[0129] The present invention also discloses an electronic device, comprising: at least one processor, at least one memory communication interface and a bus: wherein the processor, memory and communication interface communicate with each other through the bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement a real-time fault diagnosis method for industrial equipment based on a generative model.

[0130] The present invention also discloses a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to implement all or part of the steps of a real-time fault diagnosis method for industrial equipment based on a generative model according to an embodiment of the present invention. The storage medium includes: a U disk, a mobile hard disk, a read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk, and other media that can store program codes.

[0131] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A real-time fault diagnosis method for industrial equipment based on a generative model, characterized in that: The following steps are involved: S1, collecting original operation data of industrial equipment in real time through multiple sensors, preprocessing the original operation data to obtain real-time operation data, wherein the preprocessing includes outlier detection and standardization; S2, constructing a generative model based on the real-time operation data, generating a normal data sample of the industrial equipment in a normal state through the generative model, and integrating the real-time operation data and the normal data sample to obtain a multi-dimensional data sample; S3, extracting features from the multidimensional data sample using a deep learning model to obtain a fault diagnosis vector; S4, performing fault pattern recognition on the fault diagnosis vector through a classification model to obtain a fault diagnosis result, and performing state judgment on the fault diagnosis result according to a preset threshold to obtain an industrial equipment state; S5, displaying the industrial equipment status and the fault diagnosis result in real time through a dynamic monitoring interface.

2. A method for real-time fault diagnosis of industrial equipment based on a generative model as claimed in claim 1, characterized in that: Step S1 includes: S11, setting a data collection frequency, and collecting original operating data of the industrial equipment in real time at the data collection frequency through multiple sensors, wherein the original operating data includes operating temperature, operating voltage, operating power, vibration frequency and operating current; S12, performing outlier detection on the original operation data by statistical methods, identifying and eliminating outliers beyond a preset range, obtaining outlier-eliminated operation data, and unifying the data scale of the original operation data by Z-score standardization to obtain real-time operation data.

3. A method for real-time fault diagnosis of industrial equipment based on a generative model as claimed in claim 1, characterized in that: Step S2 includes: S21, constructing an initial generative model based on a generative adversarial network, and training the initial generative model through the real-time operation data to obtain a generative model; S22, inputting random noise into the generative model to obtain original sample data similar to a normal state, evaluating the quality of the original sample data, retaining original sample data corresponding to an evaluation value greater than a preset evaluation value, and obtaining normal sample data; S23, arranging the real-time operation data and the normal data samples according to the time series, and performing unified data dimension processing to obtain multi-dimensional data samples.

4. A method for real-time fault diagnosis of industrial equipment based on a generative model as claimed in claim 3, characterized in that: The calculation formula of the generative model includes: Where V(D(x;θ d ),G(z;θ g )) is the objective function of the generative adversarial network, G(z;θ g ) is the generating function, D(x;θ d ) is the discriminant function, θ d is the learnable parameter of the discriminant function, θ g is the learnable parameter of the generating function, x is the real-time running data, z is the random noise, P data (x) is the real data distribution of x, P z (z) is the random noise distribution, is the expected value of the discriminator’s predicted probability of x, is the discriminator pair generating function G(z;θ g ), λ is the regularization parameter, ||G(z;θ g )|| 2 is the generating function G(z;θ g ) is the L2 norm of the output sample, is the expected value of the L2 norm of the sample generating function.

5. The real-time fault diagnosis method for industrial equipment based on a generative model according to claim 1, characterized in that: Step S3 includes: S31, constructing an initial deep learning model, wherein the initial deep learning model adopts a convolutional neural network model, a recurrent neural network model or a long short-term memory network model, obtains annotated historical fault diagnosis data, performs model training on the initial deep learning model through the historical fault diagnosis data, optimizes model parameters, and obtains a trained deep learning model; S32, inputting the multidimensional data sample into the deep learning model, performing forward propagation to obtain feature representation, and generating a fault diagnosis vector through the feature representation.

6. A method for real-time fault diagnosis of industrial equipment based on a generative model as claimed in claim 1, characterized in that: Step S4 includes: S41, constructing an initial classification model through a classification algorithm, wherein the classification algorithm includes a support vector machine algorithm, a decision tree algorithm, a random forest algorithm, and a logistic regression algorithm, obtaining a labeled historical fault diagnosis vector, and training the initial classification model through the labeled historical fault diagnosis vector to optimize model parameters; S42, inputting the fault diagnosis vector into the classification model, and performing forward propagation to obtain a fault diagnosis result; S43, setting a preset threshold, comparing the fault diagnosis result with the preset threshold, and obtaining the industrial equipment status.

7. A method for real-time fault diagnosis of industrial equipment based on a generative model as claimed in claim 6, characterized in that: The calculation formula of the classification model includes: in, is the fault diagnosis result, f(·) is the sigmoid activation function, X is the fault diagnosis vector, θ is the model parameter of the classification model, W is the model weight vector, and b is the bias; The calculation formula for the state judgment includes: Among them, W status For industrial equipment status, W1 status 、W2 status 、W3 status They are the first industrial equipment status, the second industrial equipment status, and the third industrial equipment status. and They are the first preset threshold and the second preset threshold respectively.

8. A real-time fault diagnosis system for industrial equipment based on a generative model, a real-time fault diagnosis method for industrial equipment based on a generative model as claimed in any one of claims 1 to 7, characterized in that: The system comprises: A data processing module, used to collect raw operating data of industrial equipment in real time through multiple sensors, and pre-process the raw operating data to obtain real-time operating data, wherein the pre-processing includes outlier detection and standardization; A sample integration module, used to build a generative model based on the real-time operation data, generate a normal data sample of the industrial equipment in a normal state through the generative model, and integrate the real-time operation data and the normal data sample to obtain a multi-dimensional data sample; A feature extraction module, used to extract features from the multidimensional data sample using a deep learning model to obtain a fault diagnosis vector; A fault identification and state judgment module is used to perform fault pattern recognition on the fault diagnosis vector through a classification model to obtain a fault diagnosis result, and perform state judgment on the fault diagnosis result according to a preset threshold to obtain the state of the industrial equipment; The monitoring and display module is used to display the industrial equipment status and the fault diagnosis result in real time through a dynamic monitoring interface.

9. An electronic device, characterized in that: include: at least one processor, at least one memory, a communication interface, and a bus; The processor, memory, and communication interface communicate with each other via the bus, the memory stores program instructions executable by the processor, and the processor calls the program instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions enable a computer to implement the method according to any one of claims 1 to 7.

Images