Equipment defect prediction and model training method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses an equipment defect prediction and model training method and device, equipment and a storage medium. The method comprises the following steps: inputting a sample monitoring sequence into a to-be-trained equipment defect prediction model; the equipment defect prediction model comprises a learning layer, a prediction layer and an output layer; learning time-dependent characteristics of the sample monitoring sequence through a learning layer, and generating an adjacent matrix of the sample monitoring sequence based on the time-dependent characteristics; graph convolution features of the sample monitoring sequence are obtained through the prediction layer according to the adjacent matrix based on graph convolution operation; obtaining a training prediction result of the equipment defect based on the graph convolution features through an output layer; and adjusting model parameters of the equipment defect prediction model based on the training prediction result of the equipment defect. According to the embodiment of the invention, the equipment defect prediction model for performing high-precision prediction on different types of defects of the equipment can be obtained through training, and the different types of defects of the equipment can be accurately predicted in time.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device, equipment and storage medium for equipment defect prediction and model training. Background Art

[0002] The various industrial equipment used in the production process often has many components and complex structures. Some equipment is even sealed internally, such as gas-insulated metal-enclosed switchgear. When such highly complex equipment has defects or fails, the defect types are diverse and difficult to detect directly and in time, thus affecting the normal and safe operation of production operations. The existing methods for predicting equipment defects include prediction methods that rely on expert experience, prediction methods based on regular inspection results, and prediction methods using convolutional neural networks. However, the prediction methods in the existing technology have poor recognition accuracy when identifying different types of defects. Summary of the invention

[0003] The embodiments of the present invention provide an equipment defect prediction and model training method, apparatus, equipment and storage medium, which can train an equipment defect prediction model that can predict different types of defects of the equipment with high precision, and predict different types of defects of the equipment in a timely and accurate manner.

[0004] In a first aspect, an embodiment of the present invention provides a method for training a device defect prediction model, comprising:

[0005] Inputting the sample monitoring sequence into the equipment defect prediction model to be trained; the equipment defect prediction model comprises: a learning layer, a prediction layer and an output layer, the sample monitoring sequence comprises a plurality of historical monitoring sequences, each of which is a time series of historical monitoring data obtained by continuously monitoring parameters related to equipment defects, and the continuous monitoring time corresponding to each of the historical monitoring sequences is a preset time;

[0006] Learning the time-dependent features of the sample monitoring sequence through the learning layer, and generating an adjacency matrix of the sample monitoring sequence based on the time-dependent features;

[0007] Acquire graph convolution features of the sample monitoring sequence through the prediction layer according to the adjacency matrix based on a graph convolution operation;

[0008] Obtaining a training prediction result of a device defect based on the graph convolution feature through the output layer; and

[0009] The function value of the training loss function is determined based on the training prediction result of the equipment defect, and the model parameters of the equipment defect prediction model are adjusted according to the function value of the training loss function.

[0010] In a second aspect, an embodiment of the present invention provides a method for predicting equipment defects, including:

[0011] Inputting the real-time monitoring sequence of the equipment into the trained equipment defect prediction model; the equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, and the real-time monitoring sequence is a time series of monitoring data obtained by monitoring parameters related to equipment defects from a preset time period before to the current moment;

[0012] Learning the time-dependent features of the real-time monitoring sequence through the learning layer, and generating an adjacency matrix of the real-time monitoring sequence based on the time-dependent features;

[0013] Acquiring graph convolution features of the real-time monitoring sequence through the prediction layer based on the graph convolution operation according to the adjacency matrix; and

[0014] The real-time prediction result of the equipment defect is obtained through the output layer based on the graph convolution feature.

[0015] In a third aspect, an embodiment of the present invention provides a device for training a device defect prediction model, comprising:

[0016] A sample input module is used to input a sample monitoring sequence into the equipment defect prediction model to be trained; the equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, the sample monitoring sequence includes a plurality of historical monitoring sequences, each of which is a time series of historical monitoring data obtained by continuously monitoring parameters related to equipment defects, and the continuous monitoring time corresponding to each historical monitoring sequence is a preset time;

[0017] A sample graph structure learning module, used to learn the time-dependent features of the sample monitoring sequence through the learning layer, and generate an adjacency matrix of the sample monitoring sequence based on the time-dependent features;

[0018] A sample graph convolution module, configured to obtain graph convolution features of the sample monitoring sequence through the prediction layer according to the adjacency matrix based on a graph convolution operation; and

[0019] a sample prediction module, configured to obtain a training prediction result of a device defect based on the graph convolution feature through the output layer; and

[0020] A parameter adjustment module is used to determine the function value of the training loss function based on the training prediction result of the equipment defect, and adjust the model parameters of the equipment defect prediction model according to the function value of the training loss function.

[0021] In a fourth aspect, an embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, an equipment defect prediction model training method or an equipment defect prediction method as described in any one of the embodiments of the present invention is implemented.

[0022] In a fifth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements an equipment defect prediction model training method or an equipment defect prediction method as described in any one of the embodiments of the present invention.

[0023] The embodiment of the present invention provides an equipment defect prediction and model training method, device, equipment and storage medium. When training the equipment defect prediction model, the time-dependent characteristics of the sample monitoring sequence are learned and an adjacency matrix is ​​generated based on the time-dependent characteristics. Then, the graph convolution characteristics of the sample monitoring sequence are obtained based on the adjacency matrix based on the graph convolution operation, and the training prediction results of the equipment defects are further determined based on the graph convolution characteristics. This can facilitate the determination of the complex and irregular relationship that may exist between different types of equipment defects and time series of parameters based on the time-dependent characteristics of the sample monitoring sequence through the graph convolution operation, thereby improving the prediction accuracy of the equipment defect prediction model; by using the equipment defect prediction model trained by the embodiment of the present invention to obtain real-time prediction results of equipment defects, it is possible to provide high accuracy in real-time prediction of different types of equipment defects based on the relationship between different types of equipment defects and time series of parameters, thereby timely preventing or discovering various equipment defects and failures, and avoiding affecting the normal and safe operation of production operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 It is a flow chart of a method for training a device defect prediction model provided by an embodiment of the present invention;

[0026] Figure 2 is another flow chart of the equipment defect prediction model training method provided by an embodiment of the present invention;

[0027] Figure 3 is another flow chart of the equipment defect prediction model training method provided by an embodiment of the present invention;

[0028] Figure 4It is a flow chart of a method for predicting equipment defects provided by an embodiment of the present invention;

[0029] Figure 5 It is a structural schematic diagram of the equipment defect prediction model training device provided by an embodiment of the present invention;

[0030] Figure 6 is a structural schematic diagram of a device for predicting equipment defects provided by an embodiment of the present invention;

[0031] Figure 7 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] Figure 1 A flow chart of a method for training a device defect prediction model provided in an embodiment of the present invention is provided. The method can be performed by a device for training a device defect prediction model provided in an embodiment of the present invention. The device can be implemented in software and / or hardware. In a specific embodiment, the device can be integrated into an electronic device, such as a computer, a server, etc. The following embodiments will be described by taking the device integrated into an electronic device as an example. Reference Figure 1 , the method may specifically include the following steps:

[0035] Step 101, input the sample monitoring sequence into the equipment defect prediction model to be trained; the equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, the sample monitoring sequence includes multiple historical monitoring sequences, each historical monitoring sequence is a time series of historical monitoring data obtained by continuously monitoring parameters related to equipment defects, and the continuous monitoring time corresponding to each historical monitoring sequence is a preset time. This step can facilitate learning the adjacency matrix of the sample monitoring sequence through the learning layer.

[0036] Specifically, the above-mentioned equipment defect-related parameters can be understood as operating parameters related to the equipment defects of the equipment, and the above-mentioned equipment can be understood as highly complex industrial equipment, specifically internally sealed equipment, such as gas insulated metal enclosed switchgear (GIS).

[0037] Optionally, the device defect-related parameters include multiple operating parameters of the device, which may specifically include: temperature, humidity, operating voltage, and operating current when the device is running.

[0038] Specifically, the above-mentioned equipment defect-related parameters may also include only one operating parameter.

[0039] Specifically, the preset duration can be determined according to the monitoring and collection frequency of the equipment defect-related parameters. For example, when the monitoring and collection frequency of the equipment defect-related parameters is high, the preset duration can be set to a smaller value to avoid excessive monitoring data in the sample monitoring sequence, which affects the accuracy of model training; when the collection frequency of equipment-related parameters is low, the preset duration can be set to a larger value to obtain sufficient sample monitoring sequences.

[0040] Optionally, a sample monitoring sequence corresponding to a continuous monitoring period includes multiple simultaneous period sequences corresponding one-to-one to multiple operating parameters.

[0041] Specifically, the sample monitoring sequence corresponding to a continuous monitoring period may also include a monitoring sequence that integrates the information of the above-mentioned multiple operating parameters, and specifically, for example, may be a time series composed of matrices.

[0042] Optionally, before step 101, original historical monitoring data is obtained and preprocessed such as deduplication and anomaly removal is performed on the original historical data to obtain preprocessed historical data, and the preprocessed historical data is used to generate the above-mentioned sample monitoring sequence.

[0043] Step 102, learning the time-dependent features of the sample monitoring sequence through the learning layer, and generating an adjacency matrix of the sample monitoring sequence based on the time-dependent features. This step determines the graph structure of the sample monitoring sequence by learning the time-dependent features of the sample monitoring sequence, and represents it with an adjacency matrix, which can facilitate obtaining the graph convolution features of the sample monitoring sequence based on the graph convolution operation according to the adjacency matrix through the prediction layer, thereby determining the complex and irregular relationship between the time series of different types of equipment defects and parameters based on the time-dependent features of the sample monitoring sequence.

[0044] In an optional specific embodiment of the present invention, the above-mentioned process of generating an adjacency matrix of a sample monitoring sequence based on time-dependent features includes: learning the time-dependent features of the sample monitoring sequence through a gated recurrent unit, and adjusting the self-attention of the time-dependent features through a self-attention mechanism to obtain an adjacency matrix.

[0045] Optionally, the time-dependent features may be processed based on other processing methods in the prior art, such as being corrected using a correction coefficient, to obtain the above adjacency matrix.

[0046] Specifically, the dependency characteristics of sample monitoring sequences can also be learned through long short-term memory networks.

[0047] In an optional example, the process of learning the time-dependent features of the sample monitoring sequence through the gated recurrent unit includes: obtaining the last hidden state h of the gated recurrent unit corresponding to the sample monitoring sequence f , and obtain the dependency characteristics of the above sample monitoring sequence.

[0048] Optionally, the process of adjusting the self-attention of the time-dependent feature through the self-attention mechanism to obtain the adjacency matrix can be performed based on the following formula:

[0049]

[0050] Among them, A represents the adjacency matrix, Q represents the query vector, and K represents the key vector. Q and K can correspond to the learnable parameters A in the attention mechanism. Q and A K Linear projection is used to generate the hidden dimensions of Q and K.

[0051] Step 103, obtaining the graph convolution features of the sample monitoring sequence based on the graph convolution operation according to the adjacency matrix through the prediction layer. This step obtains the graph convolution features of the sample monitoring sequence based on the adjacency matrix determined by the time-dependent features of the sample monitoring sequence, and can determine the possible complex and irregular relationship between the time series of different types of equipment defects and parameters based on the time-dependent features of the sample monitoring sequence, and is conducive to obtaining the training prediction results of equipment defects based on the graph convolution features through the output layer.

[0052] Optionally, the prediction layer may include one graph convolution module or multiple graph convolution modules.

[0053] Optionally, the above-mentioned multiple graph convolution modules correspond one-to-one to the above-mentioned multiple operating parameters, and the above-mentioned multiple graph convolution modules can be superimposed in a first-order approximation.

[0054] In an optional specific embodiment of the present invention, the process of obtaining the graph convolution features of the sample monitoring sequence through the prediction layer based on the graph convolution operation according to the adjacency matrix includes:

[0055] The Laplace matrix of the sample monitoring sequence is obtained according to the adjacency matrix; the graph Fourier transform of the sample monitoring sequence is performed based on the Laplace matrix through the graph convolution module to obtain the spectrum data corresponding to the sample monitoring sequence; the convolution operation is performed on the spectrum data through the graph convolution module to obtain the convolution spectrum data; and the graph convolution features are obtained based on the convolution spectrum data.

[0056] Optionally, the above process of obtaining graph convolution features based on convolution spectrum data includes: obtaining convolution spectrum data as graph convolution features.

[0057] Optionally, the process of obtaining graph convolution features based on convolution spectrum data includes: performing an inverse graph Fourier transform on the convolution spectrum data according to the Laplace matrix to obtain the graph convolution features.

[0058] Specifically, the graph convolution features of the sample monitoring sequence can be obtained based on the following formula:

[0059] F=ΘLx=Θ(UΛU T )x=UΘ(Λ)U T x

[0060] Among them, F represents the graph convolution feature, Θ represents the convolution kernel, L represents the Laplace matrix, Λ represents the diagonal matrix composed of the eigenvalues ​​of L, U is the orthogonal matrix composed of the eigenvectors of L, and x represents the sample monitoring sequence.

[0061] Optionally, the process of obtaining the graph convolution features of the sample monitoring sequence through the prediction layer based on the graph convolution operation according to the adjacency matrix includes:

[0062] The adjacency matrix is ​​processed into a one-way link form to obtain a one-way link adjacency matrix; and the graph convolution feature of the sample monitoring sequence is obtained based on the one-way link adjacency matrix based on the graph convolution operation.

[0063] Specifically, by processing the adjacency matrix into a unidirectional link form, it is possible to reduce the required computing power and improve efficiency during model training and final prediction.

[0064] Step 104: Obtain the training prediction results of the equipment defects based on the graph convolution features through the output layer. This step can help adjust the model parameters of the equipment defect prediction model based on the training prediction results of the equipment defects.

[0065] Specifically, the training prediction results of the above equipment defects can be directly determined based on the graph convolution features through the fully connected layer.

[0066] Step 105, determining the function value of the training loss function based on the training prediction results of the equipment defects, and adjusting the model parameters of the equipment defect prediction model according to the function value of the training loss function. This step is based on steps 101 to 104, by learning the time-dependent characteristics of the sample monitoring sequence and generating an adjacency matrix based on the time-dependent characteristics, and then obtaining the graph convolution characteristics of the sample monitoring sequence based on the adjacency matrix based on the graph convolution operation, and further determining the training prediction results of the equipment defects based on the graph convolution characteristics, which can be beneficial to determine the complex and irregular relationship between different types of equipment defects and the time series of parameters based on the time-dependent characteristics of the sample monitoring sequence through the graph convolution operation, thereby improving the prediction accuracy of the equipment defect prediction model.

[0067] Optionally, the above training loss function can be a mean square error loss function, which can be expressed as:

[0068]

[0069] Among them, W θ represents the adjustable model parameters, y represents the training prediction result of equipment defects, and x t+1 Indicates the true value of the device defect.

[0070] Specifically, the above training loss function may also be other loss functions in the prior art, for example, a cross entropy loss function.

[0071] The following further introduces the equipment defect prediction model training method provided by an embodiment of the present invention.

[0072] Optionally, the sample monitoring sequence corresponding to a continuous monitoring period includes multiple simultaneous period sequences corresponding to multiple operating parameters one by one, and the above-mentioned prediction layer includes multiple graph convolution modules corresponding to the above-mentioned multiple operating parameters one by one, and when the multiple graph convolution modules are first-order approximated, such as Figure 2 As shown, it may include:

[0073] Step 201 : input multiple simultaneous sequences of equipment into the equipment defect prediction model to be trained.

[0074] Step 202 , learning the correlation between each parameter data and the running time in each simultaneous sequence through a gated recurrent unit, and obtaining the time-dependent features of each simultaneous sequence.

[0075] Step 203, adjusting the time-dependent features of each simultaneous sequence through a self-attention mechanism to obtain an adjacency matrix of each simultaneous sequence.

[0076] Step 204, obtaining the graph convolution features of each simultaneous sequence based on the graph convolution operation according to the corresponding adjacency matrix through the corresponding graph convolution module, and obtaining the graph convolution features of the sample monitoring sequence based on the graph convolution features of each simultaneous sequence.

[0077] Step 205: Obtain training prediction results of equipment defects based on graph convolution features through the output layer.

[0078] Step 206: determine the function value of the training loss function based on the training prediction result of the equipment defect, and adjust the model parameters of the equipment defect prediction model according to the function value of the training loss function.

[0079] Specifically, the embodiment of the present invention can expand the existing equipment defect data by performing graph structure learning and obtaining corresponding graph convolution features for multiple simultaneous sequences corresponding to multiple operating parameters one by one, and then restoring the deep spatial information of the sample monitoring sequence through a first-order approximate superposition graph convolution module, so that a high-precision equipment defect prediction model can be trained without a large amount of equipment defect data, so as to solve the technical problem that the equipment defect data set in the prior art, especially the defect data set of gas-insulated metal-enclosed switchgear, is small and difficult to meet the large quantity requirements of the deep learning model.

[0080] The following further introduces the equipment defect prediction model training method provided by the embodiment of the present invention. Figure 3 As shown, Figure 3 Step 104 in may include the following steps:

[0081] Step 1041, perform discrete Fourier transform on the graph convolution features to obtain frequency basis data.

[0082] Step 1042, learning the feature representation of the frequency-based data through a gated linear unit to obtain the frequency-based data features.

[0083] Step 1043, performing inverse discrete Fourier transform and inverse graph Fourier transform on the frequency-based data features to obtain output layer features.

[0084] Step 1044, determining the training prediction result of the equipment defect based on the output layer features through the fully connected layer.

[0085] Specifically, the feature representation of the above graph convolutional features can be directly learned through a gated linear learning unit to obtain the above output layer features.

[0086] Specifically, by further learning the feature representation in the graph convolution features through the gated linear unit, the time-related features in the graph convolution features can be better learned. After the graph convolution features are discrete Fourier transformed, the transformed data is learned through the gated linear unit, which can accelerate the learning process of the gated linear unit without losing the calculation accuracy.

[0087] Figure 4 A flow chart of a device defect prediction method provided in an embodiment of the present invention. The method can be executed by the device defect prediction method provided in an embodiment of the present invention. The device can be implemented in software and / or hardware. In a specific embodiment, the device can be integrated in an electronic device, such as a computer, a server, etc. The following embodiments will be described by taking the device integrated in an electronic device as an example. Reference Figure 4 , the method may specifically include the following steps:

[0088] Step 401, input the real-time monitoring sequence of the equipment into the trained equipment defect prediction model; the equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, and the real-time monitoring sequence is a time series of monitoring data obtained by monitoring parameters related to equipment defects from a preset time period before to the current moment.

[0089] Optionally, the above-mentioned equipment defect-related parameters include multiple operating parameters of the equipment, which may specifically include: temperature, humidity, operating voltage and operating current when the equipment is running.

[0090] Optionally, the above-mentioned real-time monitoring sequence may include multiple real-time parameter sequences corresponding one-to-one to the above-mentioned multiple operating parameters.

[0091] Step 402 : learning the time-dependent features of the real-time monitoring sequence through the learning layer, and generating an adjacency matrix of the real-time monitoring sequence based on the time-dependent features.

[0092] Optionally, the above-mentioned process of generating an adjacency matrix of a real-time monitoring sequence based on time-dependent features includes: learning the time-dependent features of the real-time monitoring sequence through a gated recurrent unit, and adjusting the self-attention of the time-dependent features through a self-attention mechanism to obtain the adjacency matrix.

[0093] Optionally, the above process of learning the time-dependent features of the real-time monitoring sequence through the gated recurrent unit and adjusting the self-attention of the time-dependent features through the self-attention mechanism to obtain the adjacency matrix includes: learning the correlation between each parameter data and the running time in each simultaneous segment sequence through the gated recurrent unit to obtain the time-dependent features of each simultaneous segment sequence, and adjusting the time-dependent features of each simultaneous segment sequence through the self-attention mechanism to obtain the adjacency matrix of each simultaneous segment sequence.

[0094] Step 403, obtaining graph convolution features of the real-time monitoring sequence through the prediction layer based on the adjacency matrix and the graph convolution operation.

[0095] Optionally, the prediction layer may include one graph convolution module or multiple graph convolution modules.

[0096] Optionally, the above-mentioned multiple graph convolution modules correspond one-to-one to the above-mentioned multiple operating parameters, and the above-mentioned multiple graph convolution modules can be superimposed in a first-order approximation.

[0097] Optionally, the above-mentioned process of obtaining the graph convolution features of the real-time monitoring sequence based on the graph convolution operation according to the adjacency matrix through the prediction layer includes: obtaining the Laplacian matrix of the real-time monitoring sequence according to the adjacency matrix; performing a graph Fourier transform on the real-time monitoring sequence based on the Laplacian matrix through the graph convolution module to obtain spectrum data corresponding to the real-time monitoring sequence; performing a convolution operation on the spectrum data through the graph convolution module to obtain convolution spectrum data; and obtaining graph convolution features based on the convolution spectrum data.

[0098] Optionally, the above process of obtaining graph convolution features based on convolution spectrum data includes: obtaining convolution spectrum data as graph convolution features.

[0099] Optionally, the above process of obtaining the graph convolution features of the real-time monitoring sequence through the prediction layer based on the adjacency matrix based on the graph convolution operation includes: processing the adjacency matrix into a unidirectional link form to obtain a unidirectional link adjacency matrix, and obtaining the graph convolution features of the real-time monitoring sequence based on the graph convolution operation according to the unidirectional link adjacency matrix.

[0100] Optionally, the above-mentioned process of obtaining the graph convolution features of the real-time monitoring sequence based on the graph convolution operation according to the adjacency matrix through the prediction layer includes: obtaining the graph convolution features of each simultaneous sequence based on the graph convolution operation according to the corresponding adjacency matrix through the corresponding graph convolution module, and obtaining the graph convolution features of the real-time monitoring sequence based on the graph convolution features of each simultaneous sequence.

[0101] Step 404: Obtain real-time prediction results of equipment defects based on graph convolution features through the output layer.

[0102] Optionally, the above process of obtaining real-time prediction results of equipment defects based on graph convolution features through the output layer includes: performing discrete Fourier transform on the graph convolution features to obtain frequency basis data; learning the feature representation of the frequency basis data through a gated linear unit to obtain frequency basis data features; performing inverse discrete Fourier transform and inverse graph Fourier transform on the frequency basis data features to obtain output layer features; and determining the training prediction results of equipment defects based on the output layer features through a fully connected layer.

[0103] The equipment defect prediction method provided by the embodiment of the present invention obtains real-time prediction results of equipment defects by utilizing the equipment defect prediction model trained by the embodiment of the present invention. It can provide high accuracy of real-time prediction of different types of equipment defects based on the relationship between different types of equipment defects and time series of parameters, thereby timely preventing or discovering various equipment defects and failures, and avoiding affecting the normal and safe operation of production operations.

[0104] Figure 5 is a structural diagram of a device for training a device defect prediction model provided by an embodiment of the present invention, and the device is suitable for executing a device defect prediction model training method provided by an embodiment of the present invention. Figure 5 As shown, the device may specifically include:

[0105] The sample input module 501 is used to input the sample monitoring sequence into the equipment defect prediction model to be trained; the equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, the sample monitoring sequence includes multiple historical monitoring sequences, each historical monitoring sequence is a time series of historical monitoring data obtained by continuously monitoring parameters related to equipment defects, and the continuous monitoring time corresponding to each historical monitoring sequence is a preset time. This module can facilitate learning the adjacency matrix of the sample monitoring sequence through the learning layer.

[0106] Optionally, the above-mentioned equipment defect-related parameters include multiple operating parameters of the equipment, which may specifically include: temperature, humidity, operating voltage and operating current when the equipment is running.

[0107] Optionally, the above-mentioned real-time monitoring sequence may include multiple real-time parameter sequences corresponding one-to-one to the above-mentioned multiple operating parameters.

[0108] Optionally, a sample monitoring sequence corresponding to a continuous monitoring period includes multiple simultaneous period sequences corresponding one-to-one to multiple operating parameters.

[0109] The sample graph structure learning module 502 is used to learn the time-dependent features of the sample monitoring sequence through the learning layer, and generate the adjacency matrix of the sample monitoring sequence based on the time-dependent features. This module can determine the graph structure of the sample monitoring sequence by learning the time-dependent features of the sample monitoring sequence, and represent it with an adjacency matrix, which can facilitate obtaining the graph convolution features of the sample monitoring sequence based on the graph convolution operation according to the adjacency matrix through the prediction layer, thereby determining the complex and irregular relationship between the time series of different types of equipment defects and parameters based on the time-dependent features of the sample monitoring sequence.

[0110] Optionally, the sample graph structure learning module 502 can be specifically used to learn the time-dependent features of the sample monitoring sequence through a gated recurrent unit, and adjust the self-attention of the time-dependent features through a self-attention mechanism to obtain an adjacency matrix.

[0111] Optionally, the above-mentioned sample graph structure learning module 502 can be specifically used to learn the correlation between each parameter data and the running time in each simultaneous segment sequence through a gated recurrent unit to obtain the time-dependent characteristics of each simultaneous segment sequence; and to adjust the time-dependent characteristics of each simultaneous segment sequence through a self-attention mechanism to obtain the adjacency matrix of each simultaneous segment sequence.

[0112] The sample graph convolution module 503 is used to obtain the graph convolution features of the sample monitoring sequence based on the graph convolution operation according to the adjacency matrix through the prediction layer. This module can obtain the graph convolution features of the sample monitoring sequence according to the adjacency matrix determined by the time-dependent features of the sample monitoring sequence, and can determine the complex and irregular relationship between the time series of different types of equipment defects and parameters based on the time-dependent features of the sample monitoring sequence, and is conducive to obtaining the training prediction results of equipment defects based on the graph convolution features through the output layer.

[0113] Optionally, the above-mentioned sample graph convolution module 503 can be specifically used to obtain the Laplace matrix of the sample monitoring sequence according to the adjacency matrix; obtain the spectrum data corresponding to the sample monitoring sequence by performing a graph Fourier transform on the sample monitoring sequence based on the Laplace matrix through the graph convolution module; obtain the convolution spectrum data by performing a convolution operation on the spectrum data through the graph convolution module; and obtain the graph convolution features based on the convolution spectrum data.

[0114] Optionally, the sample graph convolution module 503 can be specifically used to obtain convolution spectrum data as graph convolution features.

[0115] Optionally, the prediction layer may include one graph convolution module or multiple graph convolution modules.

[0116] Optionally, the above-mentioned multiple graph convolution modules correspond one-to-one to the above-mentioned multiple operating parameters, and the above-mentioned multiple graph convolution modules can be superimposed in a first-order approximation.

[0117] Optionally, the sample graph convolution module 503 can be specifically used to process the adjacency matrix into a unidirectional link form to obtain a unidirectional link adjacency matrix; and obtain graph convolution features of the sample monitoring sequence based on the unidirectional link adjacency matrix based on the graph convolution operation.

[0118] Optionally, the above-mentioned sample graph convolution module 503 can be specifically used to obtain the graph convolution features of each simultaneous sequence based on the graph convolution operation according to the corresponding adjacency matrix through the corresponding graph convolution module, and obtain the graph convolution features of the sample monitoring sequence based on the graph convolution features of each simultaneous sequence.

[0119] The sample prediction module 504 is used to obtain the training prediction results of the equipment defects based on the graph convolution features through the output layer. This module can be helpful in adjusting the model parameters of the equipment defect prediction model based on the training prediction results of the equipment defects.

[0120] Optionally, the above-mentioned sample prediction module 504 can be specifically used to perform discrete Fourier transform on the graph convolution features to obtain frequency basis data; learn the feature representation of the frequency basis data through a gated linear unit to obtain frequency basis data features; perform inverse discrete Fourier transform and inverse graph Fourier transform on the frequency basis data features to obtain output layer features; and determine the training prediction results of equipment defects based on the output layer features through a fully connected layer.

[0121] Parameter adjustment module 505 is used to determine the function value of the training loss function based on the training prediction results of the equipment defects, and adjust the model parameters of the equipment defect prediction model according to the function value of the training loss function. This module is combined with modules 501 to 503, by learning the time-dependent characteristics of the sample monitoring sequence and generating an adjacency matrix based on the time-dependent characteristics, and then obtaining the graph convolution characteristics of the sample monitoring sequence based on the adjacency matrix based on the graph convolution operation, and further determining the training prediction results of the equipment defects based on the graph convolution characteristics, which can be beneficial to determine the complex and irregular relationship between different types of equipment defects and the time series of parameters based on the time-dependent characteristics of the sample monitoring sequence through the graph convolution operation, thereby improving the prediction accuracy of the equipment defect prediction model.

[0122] Figure 6 is a structural diagram of a device for predicting equipment defects provided by an embodiment of the present invention, and the device is suitable for executing the device defect prediction method provided by an embodiment of the present invention. Figure 6 As shown, the device may specifically include:

[0123] The real-time input module 601 is used to input the real-time monitoring sequence of the equipment into the trained equipment defect prediction model; the equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, and the real-time monitoring sequence is a time series of monitoring data obtained by monitoring parameters related to equipment defects from a preset time period before to the current moment;

[0124] A real-time graph structure learning module 602, used to learn the time-dependent features of the real-time monitoring sequence through a learning layer, and generate an adjacency matrix of the real-time monitoring sequence based on the time-dependent features;

[0125] A real-time graph convolution module 603, configured to obtain graph convolution features of a real-time monitoring sequence based on a graph convolution operation according to an adjacency matrix through a prediction layer; and

[0126] The real-time prediction module 604 is used to obtain the real-time prediction results of equipment defects based on the graph convolution features through the output layer.

[0127] The equipment defect prediction device provided by the embodiment of the present invention can obtain real-time prediction results of equipment defects by using the equipment defect prediction model trained by the embodiment of the present invention. It can provide high accuracy of real-time prediction of different types of equipment defects based on the relationship between different types of equipment defects and time series of parameters, thereby timely preventing or discovering various equipment defects and failures, and avoiding affecting the normal and safe operation of production operations.

[0128] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the functional modules described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0129] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the device defect prediction model training method or the device defect prediction method provided in any of the above embodiments is implemented.

[0130] An embodiment of the present invention further provides a computer-readable medium having a computer program stored thereon, and when the program is executed by a processor, the device defect prediction model training method or the device defect prediction method provided in any of the above embodiments is implemented.

[0131] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the equipment defect prediction model training method or equipment defect prediction method as described in any one of the embodiments of the present invention.

[0132] Reference below Figure 7 , which shows a schematic diagram of the structure of a computer system 700 of an electronic device suitable for implementing an embodiment of the present invention. Figure 7 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0133] like Figure 7 As shown, the computer system 700 includes a central processing unit (CPU) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage part 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the system 700 are also stored. The CPU 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0134] The following components are connected to the I / O interface 705: an input section 706 including a keyboard, a mouse, etc.; an output section 707 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, a modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. A removable medium 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 710 as needed, so that a computer program read therefrom is installed into the storage section 708 as needed.

[0135] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 709, and / or installed from the removable medium 711. When the computer program is executed by the central processing unit (CPU) 701, the above-mentioned functions defined in the system of the present invention are executed.

[0136] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0137] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0138] The modules and / or units involved in the embodiments of the present invention may be implemented in software or hardware. The modules and / or units described may also be arranged in a processor, for example, may be described as: a processor including a sample input module, a sample graph structure learning module, a sample graph convolution module, a sample prediction module and a parameter adjustment module; or, may be described as: a processor including a real-time input module, a real-time graph structure learning module, a real-time graph convolution module and a real-time prediction module. The names of these modules do not, in some cases, constitute limitations on the modules themselves.

[0139] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: inputting the sample monitoring sequence into the equipment defect prediction model to be trained; the equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, the sample monitoring sequence includes multiple historical monitoring sequences, each historical monitoring sequence is a time series of historical monitoring data obtained by continuously monitoring parameters related to equipment defects, and the continuous monitoring time corresponding to each historical monitoring sequence is a preset time; learning the time-dependent features of the sample monitoring sequence through the learning layer, and generating an adjacency matrix of the sample monitoring sequence based on the time-dependent features; obtaining the graph convolution features of the sample monitoring sequence based on the graph convolution operation according to the adjacency matrix through the prediction layer; obtaining the training prediction results of the equipment defects based on the graph convolution features through the output layer; and determining the function value of the training loss function based on the training prediction results of the equipment defects, and adjusting the model parameters of the equipment defect prediction model according to the function value of the training loss function.

[0140] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for training a device defect prediction model, characterized in that: include: Inputting the sample monitoring sequence into the equipment defect prediction model to be trained; The equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, the sample monitoring sequence includes a plurality of historical monitoring sequences, each of which is a time series of historical monitoring data obtained by continuously monitoring parameters related to equipment defects, and the continuous monitoring time corresponding to each historical monitoring sequence is a preset time; Learning the time-dependent features of the sample monitoring sequence through the learning layer, and generating an adjacency matrix of the sample monitoring sequence based on the time-dependent features; Obtaining graph convolution features of the sample monitoring sequence through the prediction layer according to the adjacency matrix based on a graph convolution operation; Obtaining a training prediction result of a device defect based on the graph convolution feature through the output layer; and The function value of the training loss function is determined based on the training prediction result of the equipment defect, and the model parameters of the equipment defect prediction model are adjusted according to the function value of the training loss function.

2. The equipment defect prediction model training method according to claim 1, characterized in that: The step of generating the adjacency matrix of the sample monitoring sequence based on the time-dependent feature includes: The time-dependent features of the sample monitoring sequence are learned through a gated recurrent unit, and the self-attention of the time-dependent features is adjusted through a self-attention mechanism to obtain the adjacency matrix.

3. The equipment defect prediction model training method according to claim 1, characterized in that: The obtaining the graph convolution feature of the sample monitoring sequence based on the graph convolution operation according to the adjacency matrix includes: Processing the adjacency matrix into a one-way link form to obtain a one-way link adjacency matrix; The graph convolution features of the sample monitoring sequence are obtained based on the graph convolution operation according to the unidirectional link adjacency matrix.

4. The equipment defect prediction model training method according to claim 1, characterized in that: The obtaining the graph convolution feature of the sample monitoring sequence through the prediction layer according to the adjacency matrix based on the graph convolution operation includes: Obtaining a Laplacian matrix of the sample monitoring sequence according to the adjacency matrix; Performing a graph Fourier transform on the sample monitoring sequence based on the Laplace matrix through a graph convolution module to obtain spectrum data corresponding to the sample monitoring sequence; Performing a convolution operation on the spectrum data by the graph convolution module to obtain convolved spectrum data; and The graph convolution feature is obtained based on the convolution spectrum data.

5. The equipment defect prediction model training method according to claim 2, characterized in that: The equipment defect related parameters include multiple operating parameters, the sample monitoring sequence corresponding to a continuous monitoring period includes multiple simultaneous period sequences corresponding to the multiple operating parameters one by one, the prediction layer includes multiple graph convolution modules corresponding to the multiple operating parameters one by one, and the first-order approximate superposition of multiple graph convolution modules is included; The learning of the time-dependent features of the sample monitoring sequence by a gated recurrent unit and adjusting the self-attention of the time-dependent features by a self-attention mechanism to obtain the adjacency matrix includes: The gated recurrent unit is used to respectively learn the correlation between each parameter data and the running time in each simultaneous segment sequence, so as to obtain the time-dependent features of each simultaneous segment sequence; and The adjacency matrix of each simultaneous sequence is obtained by adjusting the time-dependent features of each simultaneous sequence through a self-attention mechanism; The obtaining the graph convolution feature of the sample monitoring sequence through the prediction layer according to the adjacency matrix based on the graph convolution operation includes: The graph convolution features of the respective simultaneous time segment sequences are obtained based on the graph convolution operation according to the corresponding adjacency matrix through the corresponding graph convolution module, and the graph convolution features of the sample monitoring sequence are obtained based on the graph convolution features of the respective simultaneous time segment sequences.

6. The equipment defect prediction model training method according to claim 4, characterized in that: The acquiring the graph convolution feature based on the convolution spectrum data includes: Acquire the convolution spectrum data as the graph convolution feature; The obtaining the training prediction result of the equipment defect based on the graph convolution feature through the output layer includes: Performing discrete Fourier transform on the graph convolution features to obtain frequency basis data; Learning the feature representation of the frequency-based data by a gated linear unit to obtain the frequency-based data feature; Performing inverse discrete Fourier transform and inverse graph Fourier transform on the frequency-based data features to obtain output layer features; and The training prediction result of the equipment defect is determined based on the output layer features through a fully connected layer.

7. A method for predicting equipment defects, characterized in that: include: Input the real-time monitoring sequence of the equipment into the trained equipment defect prediction model; The equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, and the real-time monitoring sequence is a time series of monitoring data obtained by monitoring parameters related to equipment defects from a preset time period before to the current moment; Learning the time-dependent features of the real-time monitoring sequence through the learning layer, and generating an adjacency matrix of the real-time monitoring sequence based on the time-dependent features; Acquiring graph convolution features of the real-time monitoring sequence through the prediction layer based on the graph convolution operation according to the adjacency matrix; and The real-time prediction result of the equipment defect is obtained through the output layer based on the graph convolution feature.

8. A device for training equipment defect prediction model, characterized in that: include: A sample input module, used to input the sample monitoring sequence into the equipment defect prediction model to be trained; The equipment defect prediction model includes: a learning layer, a prediction layer and an output layer, the sample monitoring sequence includes a plurality of historical monitoring sequences, each of which is a time series of historical monitoring data obtained by continuously monitoring parameters related to equipment defects, and the continuous monitoring time corresponding to each historical monitoring sequence is a preset time; A sample graph structure learning module, used to learn the time-dependent features of the sample monitoring sequence through the learning layer, and generate an adjacency matrix of the sample monitoring sequence based on the time-dependent features; A sample graph convolution module, configured to obtain graph convolution features of the sample monitoring sequence through the prediction layer according to the adjacency matrix based on a graph convolution operation; and a sample prediction module, configured to obtain a training prediction result of a device defect based on the graph convolution feature through the output layer; and A parameter adjustment module is used to determine the function value of the training loss function based on the training prediction result of the equipment defect, and adjust the model parameters of the equipment defect prediction model according to the function value of the training loss function.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the equipment defect prediction model training method as described in any one of claims 1 to 6 or implements the equipment defect prediction method as described in claim 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the equipment defect prediction model training method as described in any one of claims 1 to 6 or implements the equipment defect prediction method as described in claim 7.

Images