Remaining life prediction method and device of rolling bearing, equipment and storage medium

Abstract

The invention provides a residual life prediction method and device of a rolling bearing, equipment and a storage medium. The method is applied to the technical field of prediction of the residual service life of the rotating machinery and comprises the following steps: acquiring performance index data of a rolling bearing, and constructing an undirected graph according to the performance index data; inputting the undirected graph into a feature extraction model, and performing convolution processing and time sequence analysis processing on the input data by the feature extraction model to obtain feature parameters of the rolling bearing; according to the characteristic parameters, the rolling bearing is predicted to obtain a residual life prediction result of the rolling bearing, the manual participation degree of residual life prediction of the rolling bearing is reduced, and the prediction accuracy is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of remaining useful life prediction of rotating machinery, and in particular, relates to a remaining useful life prediction method, device, equipment and storage medium for rolling bearings. Background Art

[0002] Rolling bearings are vital mechanical parts that support and guide the rotation or movement of shafts in equipment. They are widely used in various fields, from automobile manufacturing to aerospace, from heavy machinery to micro equipment. During the operation of the bearing, it is subjected to impact and pressure from the external environment, and friction between the internal structures is also inevitable. As the running time of the bearing increases, these factors will cause the remaining service life of the bearing to gradually decrease, which may eventually lead to equipment failure and downtime.

[0003] The vibration signal of rolling bearing is an important monitoring indicator because it changes as the health of the bearing deteriorates. By analyzing the vibration signal of rolling bearing, key features reflecting the bearing status can be extracted, thereby predicting the remaining service life of the rolling bearing.

[0004] However, in the process of rolling bearing prediction, it is difficult to obtain labeled state data with multiple working conditions and large sample sizes, resulting in high manual participation in the remaining life prediction of rolling bearings and poor prediction accuracy. Summary of the invention

[0005] The present application provides a method, device, equipment and storage medium for predicting the remaining life of a rolling bearing, so as to solve the defects of the prior art in predicting the remaining life of a rolling bearing, such as high degree of manual participation and poor prediction accuracy.

[0006] In a first aspect, the present application provides a method for predicting the remaining life of a rolling bearing, the method comprising:

[0007] Acquire performance index data of the rolling bearing, and construct an undirected graph according to the performance index data, wherein the undirected graph is used to indicate the importance of multiple performance indexes and the correlation relationship between at least two performance indexes;

[0008] Inputting the undirected graph into a feature extraction model to obtain characteristic parameters of the rolling bearing, wherein the feature extraction model is used to perform convolution operation processing and time series analysis processing on the input data;

[0009] According to the characteristic parameters, the rolling bearing is predicted to obtain a prediction result, and the prediction result is used to indicate the remaining life of the rolling bearing.

[0010] Optionally, constructing an undirected graph according to the performance indicator data includes:

[0011] Preprocessing the performance indicator data;

[0012] Analyzing and processing the preprocessed performance indicator data of the multiple different time periods to obtain at least one degradation feature of the multiple different time periods;

[0013] A correlation coefficient between the at least one degradation feature and the health indicator is calculated, and the undirected graph is constructed according to the correlation coefficient.

[0014] Optionally, the calculating a correlation coefficient between the at least one degradation feature and the health indicator, and constructing the undirected graph according to the correlation coefficient, includes:

[0015] Calculating the correlation coefficient between the at least one degradation feature and the health index, and sorting the correlation coefficients in descending order of absolute value, and determining the degradation features corresponding to a preset number of correlation coefficients before sorting as target degradation nodes;

[0016] Calculating the spatial distance between each of the multiple target degenerate nodes and the remaining nodes to obtain the spatial distance value between each target degenerate node and the remaining target degenerate nodes, and sorting the spatial distance values ​​in ascending order, and determining the target degenerate nodes corresponding to the nearest neighbor parameter spatial distance values ​​before sorting as associated nodes;

[0017] The target degenerate node is determined as a point in the undirected graph, and the associated nodes are connected to be determined as edges in the undirected graph.

[0018] Optionally, the feature extraction model includes: a graph convolutional network layer and a long short-term memory network layer, the feature parameters include: relationship features and time series features, and the undirected graph is input into the feature extraction model to obtain the feature parameters of the rolling bearing, including:

[0019] Inputting the undirected graph into the graph convolutional network layer to obtain the relationship features of the rolling bearing, wherein the relationship features are used to indicate the features between each node and adjacent nodes in the undirected graph;

[0020] The relationship feature is input into the long short-term memory network layer to obtain the time series feature of the rolling bearing, and the time series feature is used to indicate the change trend of the rolling bearing at different time points.

[0021] Optionally, the step of inputting the undirected graph into the graph convolutional network layer to obtain the relationship features of the rolling bearing, wherein the relationship features are used to indicate features between each node and adjacent nodes in the undirected graph, including:

[0022] Performing conversion processing on the undirected graph, converting the undirected graph into a neighbor matrix and a node feature matrix, and inputting the neighbor matrix and the node feature matrix into a graph convolutional network layer as sample data;

[0023] The graph convolution network layer performs convolution processing on the sample data to obtain relationship features between nodes and adjacent nodes;

[0024] After being processed by multiple convolutional network layers, the target node features are obtained, and the multi-source signal relationship features are extracted from the target node features.

[0025] Optionally, the long short-term memory network layer includes a forward unit and a backward unit, and the relationship feature is input into the long short-term memory network layer to obtain the time series feature of the rolling bearing, and the time series feature is used to indicate the change trend of the rolling bearing at different time points, including:

[0026] Sorting the relationship features of the multi-source signals in chronological order to obtain a multi-source signal relationship time series;

[0027] Inputting the multi-source signal relationship time series into the long short-term memory network layer;

[0028] The forward unit processes the multi-source signal relationship time series from a forward time order to obtain a first causal relationship of the series;

[0029] The backward unit processes the multi-source signal relationship time series from a backward time order to obtain a second causal relationship of the series;

[0030] According to the first causal relationship and the second causal relationship, a time series feature of the sample data is determined.

[0031] In a second aspect, the present application provides a device for predicting the remaining life of a rolling bearing, the device comprising:

[0032] An acquisition module is used to acquire performance index data of rolling bearings;

[0033] A processing module, used to construct an undirected graph according to the performance indicator data, wherein the undirected graph is used to indicate the importance of multiple performance indicators and the correlation relationship between at least two performance indicators;

[0034] The processing module is further used to input the undirected graph into a feature extraction model to obtain characteristic parameters of the rolling bearing, wherein the feature extraction model is used to perform convolution operation processing and time series analysis processing on the input data;

[0035] The processing module is further used to perform prediction processing on the rolling bearing according to the characteristic parameters to obtain a prediction result, and the prediction result is used to indicate the remaining life of the rolling bearing.

[0036] Optionally, the device further comprises: a calculation module;

[0037] The processing module is further used to pre-process the performance indicator data;

[0038] The processing module is further used to analyze and process the pre-processed performance indicator data of the multiple different time periods to obtain at least one degradation feature of the multiple different time periods;

[0039] The calculation module is further used to calculate the correlation coefficient between the at least one degradation feature and the health indicator, and construct the undirected graph according to the correlation coefficient.

[0040] Optionally, the calculation module is further used to calculate the correlation coefficient between the at least one degradation feature and the health index, and sort the correlation coefficients in descending order of absolute value, and determine the degradation features corresponding to a preset number of correlation coefficients before sorting as target degradation nodes;

[0041] The calculation module is further used to calculate the spatial distance between each of the multiple target degenerate nodes and the remaining nodes, obtain the spatial distance value between each target degenerate node and the remaining target degenerate nodes, and sort the spatial distance values ​​in ascending order, and determine the target degenerate nodes corresponding to the nearest neighbor parameter spatial distance values ​​before sorting as associated nodes;

[0042] The processing module is further used to determine the target degenerate node as a point in the undirected graph, and perform connection processing on the associated nodes to determine them as edges in the undirected graph.

[0043] Optionally, the processing module is further used to input the undirected graph into the graph convolutional network layer to obtain the relationship feature of the rolling bearing, where the relationship feature is used to indicate the features between each node and adjacent nodes in the undirected graph;

[0044] The processing module is also used to input the relationship features into the long short-term memory network layer to obtain the time series features of the rolling bearing, and the time series features are used to indicate the change trend of the rolling bearing at different time points.

[0045] Optionally, the processing module is further used to convert the undirected graph into a neighbor matrix and a node feature matrix, and input the neighbor matrix and the node feature matrix into a graph convolutional network layer as sample data;

[0046] The processing module is further used for the graph convolution network layer to perform convolution processing on the sample data to obtain the relationship characteristics between the node and the adjacent nodes;

[0047] The processing module is also used to obtain target node features after processing through multiple convolutional network layers, and extract multi-source signal relationship features from the target node features.

[0048] Optionally, the processing module is further used to sort the relationship features of the multi-source signals in chronological order to obtain a multi-source signal relationship time series;

[0049] The processing module is further used to input the multi-source signal relationship time series into the long short-term memory network layer;

[0050] The processing module is also used for the forward unit to process the multi-source signal relationship time series from the forward time sequence to obtain the first causal relationship of the sequence;

[0051] The processing module is further used for the backward unit to process the multi-source signal relationship time series from a backward time sequence to obtain a second causal relationship of the sequence;

[0052] The processing module is further used to determine the time series characteristics of the sample data according to the first causal relationship and the second causal relationship.

[0053] In a third aspect, the present application provides a device for predicting the remaining life of a rolling bearing, comprising:

[0054] Memory;

[0055] processor;

[0056] Wherein, the memory stores computer-executable instructions;

[0057] The processor executes the computer-executable instructions stored in the memory to implement the remaining life prediction method for the rolling bearing as described in the first aspect and various possible implementations of the first aspect.

[0058] In a fourth aspect, the present application provides a computer storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the method for predicting the remaining life of a rolling bearing as described in the first aspect and various possible implementations of the first aspect.

[0059] The present application provides a method, device, equipment and storage medium for predicting the remaining life of a rolling bearing. The method obtains performance index data of the rolling bearing and constructs an undirected graph based on the performance index data; the undirected graph is input into a feature extraction model, and the feature extraction model performs convolution processing and time series analysis processing on the input data to obtain characteristic parameters of the rolling bearing; according to the characteristic parameters, the rolling bearing is predicted to obtain a prediction result of the remaining life of the rolling bearing, thereby reducing the manual participation in the prediction of the remaining life of the rolling bearing and improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0061] Figure 1 A schematic diagram of a method for predicting the remaining life of a rolling bearing provided in this application Figure 1 ;

[0062] Figure 2 A schematic diagram of a method for predicting the remaining life of a rolling bearing provided in this application Figure 2 ;

[0063] Figure 3 A schematic diagram of a method for predicting the remaining life of a rolling bearing provided in this application Figure 3 ;

[0064] Figure 4 A schematic diagram of the structure of a device for predicting the remaining life of a rolling bearing provided in the present application;

[0065] Figure 5 A schematic diagram of the structure of a rolling bearing remaining life prediction device provided in this application.

[0066] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0068] The terms "first", "second", "third", "fourth", etc. (if any) 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 numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein.

[0069] In the embodiments of the present application, the words "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0070] First, the terms involved in this application are explained.

[0071] Undirected graph: A basic concept in graph theory, it is a graph consisting of vertices (or nodes) and edges, where the edges have no direction. That is, an edge connects two vertices, but does not specify a specific direction from one vertex to another. In other words, if there is an edge connecting vertex A and vertex B in the graph, then this edge can point from A to B or from B to A. Undirected graphs can be used to represent various relationships, such as friend relationships in social networks, traffic connections in road networks, and connections between atoms in molecular structures. In an undirected graph, vertices represent entities or nodes, and edges represent relationships or connections between nodes.

[0072] Graph Convolutional Network (GCN) is a neural network layer used to process graph structured data. Its core idea is to extend the convolution operation to graph data and capture the local connection pattern between nodes by learning the node feature representation on the graph.

[0073] Bidirectional Long Short-Term Memory (Bi-LSTM) is a deep learning model and a variant of the Long Short-Term Memory (LSTM). Bi-LSTM combines information flows in two directions, starting from the beginning and end of the sequence, to better capture contextual information in the sequence data. In the traditional unidirectional LSTM model, information can only propagate in one direction from the past to the future, while Bi-LSTM allows information to propagate in two directions, enabling a more comprehensive understanding of sequence data. The Bi-LSTM model consists of two independent LSTM networks, one responsible for forward propagation (from the beginning to the end of the sequence) and the other responsible for backward propagation (from the end of the sequence to the beginning). In this way, Bi-LSTM can capture both past and future contextual information at the same time, which helps to improve the modeling ability of sequence data.

[0074] As a vital mechanical part that supports and guides the rotation or movement of shafts in equipment, rolling bearings play a key role in various fields. From automobile manufacturing to aerospace, from heavy machinery to micro equipment, the application range of rolling bearings is wide and diverse.

[0075] As the running time of the bearing increases, the bearing will be subjected to impact and pressure from the external environment, and the friction between the internal structures will gradually increase. These factors will cause wear and damage to the bearing, thereby shortening the remaining service life of the bearing. If not detected and maintained in time, the bearing may fail, causing equipment downtime, production interruption and loss. Therefore, real-time monitoring and prediction of the health status of rolling bearings has become crucial.

[0076] By analyzing the vibration signal of rolling bearings, key features reflecting the bearing status can be extracted, thereby realizing the prediction of the remaining service life of rolling bearings. However, in the process of rolling bearing prediction, it is difficult to obtain labeled status data with multiple working conditions and large sample sizes, resulting in high manual participation in the prediction of the remaining life of rolling bearings and poor prediction accuracy.

[0077] In response to the above problems, the present application proposes a method for predicting the remaining life of a rolling bearing. The method constructs an undirected graph based on the performance indicator data of the rolling bearings for multiple time periods, extracts the relationship features and time series features of the undirected graph through a feature extraction model, and predicts the remaining service life of the rolling bearing based on the relationship features and time series features, thereby reducing the degree of manual participation in the prediction of the remaining life of the rolling bearing and improving the prediction accuracy.

[0078] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0079] Figure 1 A schematic diagram of a method for predicting the remaining life of a rolling bearing provided in an embodiment of the application Figure 1 .like Figure 1 As shown, the remaining life prediction method of a rolling bearing provided in this embodiment includes:

[0080] S101. Obtain performance index data of rolling bearings.

[0081] Among them, various performance index data reflecting the running status of the rolling bearing are obtained, such as: vibration amplitude, frequency spectrum, temperature, workload, rotation speed, etc.

[0082] A possible implementation method is to use a vibration sensor to regularly record the vibration signal of the rolling bearing under different working conditions, and use a temperature sensor to monitor the temperature changes of the rolling bearing and the surrounding environment, and use the equipment management system to record performance indicator data such as the load and rotation speed of the rolling bearing.

[0083] S102. Construct an undirected graph according to the performance indicator data.

[0084] Among them, the undirected graph is used to indicate the importance of multiple performance indicators and the correlation between at least two performance indicators. According to the correlation coefficient between the performance indicator and the health indicator, the performance indicator that is strongly correlated with the health indicator is selected as the node in the undirected graph; the correlation between the performance indicators is calculated, for example, the Pearson correlation coefficient calculation method or other correlation measurement methods can be used to evaluate the degree of correlation between the performance indicators. According to the size of the correlation, the performance indicators with correlation between the performance indicators are determined; according to the correlation between the performance indicators, the performance indicator nodes with correlation are connected to form the edges of the undirected graph. The edges can be represented by line segments, which represent the correlation between the performance indicators.

[0085] S103, inputting the undirected graph into a feature extraction model to obtain characteristic parameters of the rolling bearing.

[0086] The undirected graph is passed as input data to a feature extraction model, and the feature extraction model is used to perform convolution operation and time series analysis on the input data to obtain characteristic parameters of the rolling bearing.

[0087] Specifically, the undirected graph is converted into a data representation suitable for input into a feature extraction model, and the undirected graph is input into the feature extraction model. The model performs convolution operation on the undirected graph to obtain a convolution operation result. The model performs time series analysis on the undirected graph to obtain a time series analysis result. The convolution operation result and the time series analysis result are used as characteristic parameters of the rolling bearing.

[0088] S104. Perform prediction processing on the rolling bearing according to the characteristic parameters to obtain a prediction result.

[0089] The characteristic parameters obtained by the feature extraction model are used as input data, and the trained model is used to perform prediction processing on the rolling bearing to obtain a prediction result of the remaining life of the rolling bearing, and the prediction result is used to indicate the remaining life of the rolling bearing.

[0090] This embodiment proposes a method for predicting the remaining life of a rolling bearing. The method obtains performance index data of the rolling bearing and constructs an undirected graph based on the performance index data; the undirected graph is input into a feature extraction model, and the feature extraction model performs convolution processing and time series analysis processing on the input data to obtain characteristic parameters of the rolling bearing; based on the characteristic parameters, the rolling bearing is predicted to obtain a prediction result of the remaining life of the rolling bearing, thereby reducing the degree of manual participation in the prediction of the remaining life of the rolling bearing.

[0091] Figure 2 This is a schematic diagram of a method for predicting the remaining life of a rolling bearing provided in an embodiment of the present application. Figure 2 This embodiment is in Figure 1 Based on the embodiment, a possible implementation method of constructing an undirected graph based on performance indicator data is described in detail. Figure 2 As shown, the method comprises:

[0092] S201, preprocessing performance indicator data.

[0093] The acquired rolling bearing vibration signal is extracted and fused through features such as peak value, standard deviation, and test statistics; and the data is checked for missing values, abnormal values, or error values. For missing values, the corresponding data points can be deleted or filled using interpolation methods. For abnormal values ​​or error values, they can be corrected or deleted.

[0094] S202: Analyze and process the preprocessed performance indicator data of the multiple different time periods to obtain at least one degradation feature of the multiple different time periods.

[0095] The performance indicator data is divided into different time periods, such as by month, quarter or year. Ensure that the data in each time period has a certain degree of continuity and consistency; extract various features to describe the degradation of the system for the performance indicator data of each time period. Common features include mean, standard deviation, maximum value, minimum value, trend indicators, etc.; by analyzing the performance indicator data of different time periods, the degradation characteristics of the system can be identified. These characteristics may include trends that increase or decrease over time, periodic changes, etc.; use statistical methods, such as correlation analysis, regression analysis, time series analysis, etc., to explore the relationship between performance indicator data in different time periods to identify at least one degradation feature of multiple different time periods.

[0096] S203, calculating a correlation coefficient between at least one degradation feature and a health indicator, and sorting the correlation coefficients in descending order of absolute value, and determining degradation features corresponding to a preset number of correlation coefficients before sorting as target degradation nodes.

[0097] For each degradation feature, the correlation coefficient between each degradation feature and the health indicator is calculated. For example, the Pearson correlation coefficient, Spearman correlation coefficient or other correlation indicators can be used to calculate the degree of correlation between them; the calculated correlation coefficients are sorted in order from large to small in absolute value, so as to find the degradation features most correlated with the health indicator; according to a preset number, the degradation features corresponding to the top-ranked correlation coefficients are selected as target degradation nodes. The first few or the first several degradation features with the highest correlation can be selected according to the specific situation.

[0098] Analyze the relationship between the selected target degradation nodes and health indicators to understand their impact and interaction. This helps to deeply understand the degradation mechanism of the system and predict the life of the equipment.

[0099] S204, calculating the spatial distance between each of the multiple target degenerate nodes and the remaining nodes, obtaining the spatial distance value between each target degenerate node and the remaining target degenerate nodes, and sorting the spatial distance values ​​in ascending order, and determining the target degenerate nodes corresponding to the spatial distance values ​​of the neighbor parameter before sorting as associated nodes.

[0100] Among them, the target degradation node represents the features extracted from the bearing performance index data and related to the degradation of the bearing health status; in the feature space, each target degradation node can be regarded as a point whose coordinates are determined by its eigenvalue. Spatial distance usually refers to metrics such as Euclidean distance, Manhattan distance or cosine similarity between these points; to calculate the spatial distance between all target degradation nodes, it is necessary to sort them from small to large according to these distance values. This helps to identify which nodes are close to each other in the feature space and may have similar degradation patterns. The neighbor parameter is a preset parameter that determines how many nearest neighbor nodes to consider when constructing an undirected graph; after sorting, the first neighbor parameter nodes of each target degradation node are selected as its associated nodes.

[0101] S205: Determine the target degenerate node as a point in an undirected graph, perform connection processing on the associated nodes, and determine them as edges in the undirected graph.

[0102] Each target degradation node is represented as a node in an undirected graph. These nodes represent different degradation features or nodes in the system. According to the determined associated nodes, these nodes are connected to form edges in the undirected graph. Each edge indicates that there is an association or similarity between two nodes. According to the connection relationship between nodes and edges, an undirected graph is constructed. The constructed undirected graph is visualized through a graph visualization tool. This helps to intuitively understand the association relationship between target degradation nodes.

[0103] The present embodiment provides a method for predicting the remaining life of a rolling bearing. The method preprocesses the performance indicator data to extract at least one degradation indicator from the performance indicator data of multiple different time periods; calculates the correlation coefficient between the degradation characteristics and the health indicator to determine the target degradation node; calculates the spatial distance between multiple target degradation nodes to determine the associated nodes in the target degradation node; represents the target degradation node as a point in an undirected graph, connects the associated nodes and determines them as edges in the undirected graph to show the association relationship between the nodes. Through this method, by constructing an undirected graph to show the association relationship between the nodes, a deeper understanding and guidance is provided for the maintenance and management of the system, and the manual participation in the prediction of the remaining life of the rolling bearing is reduced.

[0104] Figure 3 This is a schematic diagram of a method for predicting the remaining life of a rolling bearing provided in an embodiment of the present application. Figure 3 This embodiment is in Figure 1 Based on the embodiment, a possible implementation method of inputting an undirected graph into a feature extraction model to obtain characteristic parameters of a rolling bearing is described in detail. Figure 3 As shown, the method comprises:

[0105] S301, converting the undirected graph into a neighbor matrix and a node feature matrix, and inputting the neighbor matrix and the node feature matrix into a graph convolutional network layer as sample data.

[0106] Among them, the neighbor matrix is ​​used to represent the connection relationship between nodes in the graph. The neighbor matrix is ​​a square matrix whose rows and columns represent the nodes in the graph. For example, if there is an edge between node i and node j, then the element A in the matrix ij is 1 if there is no edge; otherwise it is 0.

[0107] The node feature matrix provides the feature information of each node in the graph. The node feature matrix is ​​a two-dimensional matrix, where the rows represent the nodes in the graph and the columns represent the feature values ​​of each node.

[0108] The undirected graph is converted into a neighbor matrix and node feature matrix suitable for the graph convolutional network, and the neighbor matrix and node feature matrix are provided as input to the graph convolutional network layer. Through the graph convolution operation, the local connection pattern and feature relationship between nodes can be captured, so as to learn richer node information.

[0109] S302, the graph convolutional network layer performs convolution processing on the sample data to obtain the relationship characteristics between the node and the adjacent nodes.

[0110] Among them, the node feature matrix contains the feature representation of each node, and the neighbor matrix describes the connection relationship between nodes. The graph convolutional network layer convolves the neighbor matrix and the node feature matrix. For each node, the graph convolutional network layer considers the neighbor nodes of the node and combines their feature information to update the representation of the node. In the convolution operation, each node will perform a weighted sum of its features with the features of its neighbor nodes, and the weight is determined by the neighbor matrix. In this way, each node will get a new feature representation that contains information about its neighboring nodes. Through the convolution operation, the graph convolutional network layer can extract the relationship features between the node and its neighboring nodes. These features can capture the position of the node in the graph structure, the interaction with the neighboring nodes, and the topological structure of the overall graph.

[0111] S303, after being processed by the multi-layer convolutional network layer, the target node features are obtained, and the multi-source signal relationship features are extracted from the target node features.

[0112] The graph convolutional network will contain multiple graph convolutional network layers, and each layer will further update the feature representation of the node. After being processed by multiple layers of convolutional networks, the target node features are obtained, and each node will have a feature vector representation, which contains the relationship features between the node and its neighboring nodes. These feature vectors capture the position of the node in the graph structure and the interaction with the neighboring nodes.

[0113] Fusion of target node features with other related features. Use feature extraction methods, such as dimensionality reduction technology, clustering algorithms, etc. to extract multi-source signal relationship features from target node features to reveal the correlation and influence between different signals. Establish a relationship model between multi-source signals, such as a linear model, nonlinear model or deep learning model, to better understand the interaction between different signals.

[0114] S304: sort the relationship features of the multi-source signals in chronological order to obtain a multi-source signal relationship time series.

[0115] Among them, the relationship characteristics of multi-source signals represent the interrelationship characteristics between multiple key signals, such as vibration, temperature, load, etc. These characteristics reveal how different monitoring parameters interact and influence each other during the bearing operation cycle. In order to better understand the trend of these relationships over time, the relationship characteristics of multi-source signals are arranged in the order of their occurrence to form a multi-source signal relationship time series. Time series helps analyze the trend of feature changes over time.

[0116] S305. The long short-term memory network layer includes a forward unit and a backward unit, and the multi-source signal relationship time series is input into the long short-term memory network layer.

[0117] Among them, the long short-term memory network is a special recurrent neural network that is particularly suitable for the analysis of long sequence data. It can solve the problem of gradient disappearance in traditional recurrent neural networks and better capture long-term dependencies. The long short-term memory network layer contains forward units and backward units. These two units work together to provide a capability to analyze sequence data from two opposite time flows.

[0118] S306 , the forward unit processes the multi-source signal relationship time series from the forward time order to obtain the first causal relationship of the sequence.

[0119] S307 , the backward unit processes the multi-source signal relationship time series from a backward time order to obtain a second causal relationship of the sequence.

[0120] Among them, the multi-source signal relationship time series is input into the long short-term memory network layer, and the forward unit processes information from the beginning to the end of the multi-source signal relationship time series, capturing the "cause to effect" relationship of events in the sequence, and obtaining the first causal relationship of the sequence, that is, how the characteristics of a certain moment affect the subsequent state. For example, in the analysis of bearing data, the forward unit revealed how the early increase in small vibrations gradually led to the deterioration of the overall health of the bearing.

[0121] In contrast to the forward unit, the backward unit traces back from the end of the sequence to the beginning, capturing the "result to cause" relationship of events in the sequence, and obtaining the second causal relationship of the sequence, that is, how the current health state is determined by previous events. For example, in the analysis of bearing data, the backward unit identified that the recent severe wear was caused by a minor fault that was not intervened in time earlier.

[0122] S308: Determine the time series characteristics of the sample data according to the first causal relationship and the second causal relationship.

[0123] Among them, the outputs of the forward and backward units of the long short-term memory network layer are combined to obtain the first causal relationship and the second causal relationship of the multi-source signal relationship time series in the forward and backward perspectives. The first causal relationship is the evolution logic in the forward direction of time, and the second causal relationship is the cause chain inferred from the result. Fusion determines a more comprehensive time series feature of the sample data based on the first causal relationship and the second causal relationship. The time series feature reveals the key time points and influencing factors in the bearing degradation process, which is difficult to capture in a one-way analysis.

[0124] This embodiment provides a method for predicting the remaining life of a rolling bearing. The method extracts the relationship characteristics and time series characteristics between the data of each node from an undirected graph by combining a convolutional graph network layer and a long short-term memory network layer, and provides a more comprehensive data basis for further analysis and prediction processing, thereby reducing the manual participation in the prediction of the remaining life of the rolling bearing and improving the prediction accuracy.

[0125] Figure 4 A schematic diagram of the structure of a rolling bearing remaining life prediction device provided in this application, such as Figure 4 As shown, the remaining life prediction device 400 of a rolling bearing provided in this embodiment includes:

[0126] An acquisition module 401 is used to acquire performance index data of a rolling bearing;

[0127] A processing module 402 is used to construct an undirected graph according to the performance indicator data, wherein the undirected graph is used to indicate the importance of multiple performance indicators and the correlation relationship between at least two performance indicators;

[0128] The processing module 402 is further used to input the undirected graph into a feature extraction model to obtain characteristic parameters of the rolling bearing, wherein the feature extraction model is used to perform convolution operation processing and time series analysis processing on the input data;

[0129] The processing module 402 is further used to perform prediction processing on the rolling bearing according to the characteristic parameters to obtain a prediction result, and the prediction result is used to indicate the remaining life of the rolling bearing.

[0130] Optionally, the device further includes: a calculation module 403;

[0131] The processing module 402 is also used to pre-process the performance indicator data;

[0132] The processing module 402 is further used to analyze and process the pre-processed performance indicator data of the multiple different time periods to obtain at least one degradation feature of the multiple different time periods;

[0133] The calculation module 403 is further used to calculate the correlation coefficient between the at least one degradation feature and the health index, and construct the undirected graph according to the correlation coefficient.

[0134] Optionally, the calculation module 403 is further used to calculate the correlation coefficient between the at least one degradation feature and the health index, and sort the correlation coefficients in descending order of absolute value, and determine the degradation features corresponding to a preset number of correlation coefficients before sorting as target degradation nodes;

[0135] The calculation module 403 is further used to calculate the spatial distance between each of the multiple target degraded nodes and the remaining nodes, obtain the spatial distance value between each target degraded node and the remaining target degraded nodes, and sort the spatial distance values ​​in ascending order, and determine the target degraded nodes corresponding to the neighbor parameter spatial distance values ​​before sorting as associated nodes;

[0136] The processing module 402 is further configured to determine the target degraded node as a point in the undirected graph, and perform connection processing on the associated nodes to determine them as edges in the undirected graph.

[0137] Optionally, the processing module 402 is further used to input the undirected graph into the graph convolutional network layer to obtain the relationship feature of the rolling bearing, where the relationship feature is used to indicate the feature between each node and adjacent nodes in the undirected graph;

[0138] The processing module 402 is further used to input the relationship features into the long short-term memory network layer to obtain the time series features of the rolling bearing, and the time series features are used to indicate the change trend of the rolling bearing at different time points.

[0139] Optionally, the processing module 402 is further used to convert the undirected graph into a neighbor matrix and a node feature matrix, and input the neighbor matrix and the node feature matrix into a graph convolutional network layer as sample data;

[0140] The processing module 402 is also used for the graph convolution network layer to perform convolution processing on the sample data to obtain the relationship characteristics between the node and the adjacent nodes;

[0141] The processing module 402 is also used to obtain target node features after processing through multiple convolutional network layers, and extract multi-source signal relationship features from the target node features.

[0142] Optionally, the processing module 402 is further configured to sort the relationship features of the multi-source signals in time order to obtain a multi-source signal relationship time series;

[0143] The processing module 402 is further used to input the multi-source signal relationship time series into the long short-term memory network layer;

[0144] The processing module 402 is also used for the forward unit to process the multi-source signal relationship time series from the forward time sequence to obtain the first causal relationship of the sequence;

[0145] The processing module 402 is also used for the backward unit to process the multi-source signal relationship time series from a backward time sequence to obtain a second causal relationship of the sequence;

[0146] The processing module 402 is further configured to determine a time series feature of the sample data according to the first causal relationship and the second causal relationship.

[0147] Figure 5 This is a schematic diagram of the structure of a rolling bearing remaining life prediction device provided in this application. Figure 5 As shown, the present application provides a device for predicting the remaining life of a rolling bearing. The device 500 for predicting the remaining life of a rolling bearing includes: a receiver 501 , a transmitter 502 , a processor 503 and a memory 504 .

[0148] Receiver 501, used for receiving instructions and data;

[0149] A transmitter 502, used for sending instructions and data;

[0150] Memory 504, used to store computer-executable instructions;

[0151] The processor 503 is used to execute the computer-executable instructions stored in the memory 504 to implement the various steps executed by the method for predicting the remaining life of a rolling bearing in the above embodiment. For details, please refer to the relevant description in the above embodiment of the method for predicting the remaining life of a rolling bearing.

[0152] Optionally, the memory 504 may be independent or integrated with the processor 503 .

[0153] When the memory 504 is independently provided, the electronic device further includes a bus for connecting the memory 504 and the processor 503 .

[0154] The present application also provides a computer storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, a method for predicting the remaining life of a rolling bearing as performed by the above-mentioned remaining life prediction device for a rolling bearing is implemented.

[0155] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0156] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0157] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for predicting the remaining life of a rolling bearing, characterized in that: The method comprises: Acquire performance index data of the rolling bearing, and construct an undirected graph according to the performance index data, wherein the undirected graph is used to indicate the importance of multiple performance indexes and the correlation relationship between at least two performance indexes; Inputting the undirected graph into a feature extraction model to obtain characteristic parameters of the rolling bearing, wherein the feature extraction model is used to perform convolution operation processing and time series analysis processing on the input data; According to the characteristic parameters, the rolling bearing is predicted to obtain a prediction result, and the prediction result is used to indicate the remaining life of the rolling bearing.

2. The method according to claim 1, characterized in that: The constructing an undirected graph according to the performance indicator data includes: Preprocessing the performance indicator data; Analyzing and processing the preprocessed performance indicator data of the multiple different time periods to obtain at least one degradation feature of the multiple different time periods; A correlation coefficient between the at least one degradation feature and the health indicator is calculated, and the undirected graph is constructed according to the correlation coefficient.

3. The method according to claim 2, characterized in that The calculating the correlation coefficient between the at least one degradation feature and the health indicator, and constructing the undirected graph according to the correlation coefficient, comprises: Calculating the correlation coefficient between the at least one degradation feature and the health index, and sorting the correlation coefficients in descending order of absolute value, and determining the degradation features corresponding to a preset number of correlation coefficients before sorting as target degradation nodes; Calculating the spatial distance between each of the multiple target degenerate nodes and the remaining nodes to obtain the spatial distance value between each target degenerate node and the remaining target degenerate nodes, and sorting the spatial distance values ​​in ascending order, and determining the target degenerate nodes corresponding to the nearest neighbor parameter spatial distance values ​​before sorting as associated nodes; The target degenerate node is determined as a point in the undirected graph, and the associated nodes are connected to be determined as edges in the undirected graph.

4. The method according to claim 1, characterized in that: The feature extraction model includes: a graph convolution network layer and a long short-term memory network layer, the feature parameters include: relationship features and time series features, and the undirected graph is input into the feature extraction model to obtain the feature parameters of the rolling bearing, including: Inputting the undirected graph into the graph convolutional network layer to obtain the relationship features of the rolling bearing, wherein the relationship features are used to indicate the features between each node and adjacent nodes in the undirected graph; The relationship feature is input into the long short-term memory network layer to obtain the time series feature of the rolling bearing, and the time series feature is used to indicate the change trend of the rolling bearing at different time points.

5. The method according to claim 4, characterized in that The undirected graph is input into the graph convolutional network layer to obtain the relationship feature of the rolling bearing, wherein the relationship feature is used to indicate the feature between each node and the adjacent node in the undirected graph, including: Performing conversion processing on the undirected graph, converting the undirected graph into a neighbor matrix and a node feature matrix, and inputting the neighbor matrix and the node feature matrix into a graph convolutional network layer as sample data; The graph convolution network layer performs convolution processing on the sample data to obtain relationship features between nodes and adjacent nodes; After being processed by multiple convolutional network layers, the target node features are obtained, and the multi-source signal relationship features are extracted from the target node features.

6. The method according to claim 5, characterized in that The long short-term memory network layer includes a forward unit and a backward unit. The relationship feature is input into the long short-term memory network layer to obtain the time series feature of the rolling bearing. The time series feature is used to indicate the change trend of the rolling bearing at different time points, including: Sorting the relationship features of the multi-source signals in chronological order to obtain a multi-source signal relationship time series; Inputting the multi-source signal relationship time series into the long short-term memory network layer; The forward unit processes the multi-source signal relationship time series from a forward time order to obtain a first causal relationship of the series; The backward unit processes the multi-source signal relationship time series from a backward time order to obtain a second causal relationship of the series; According to the first causal relationship and the second causal relationship, a time series feature of the sample data is determined.

7. A device for predicting the remaining life of a rolling bearing, characterized in that: The device comprises: An acquisition module is used to acquire performance index data of rolling bearings; A processing module, used to construct an undirected graph according to the performance indicator data, wherein the undirected graph is used to indicate the importance of multiple performance indicators and the correlation relationship between at least two performance indicators; The processing module is further used to input the undirected graph into a feature extraction model to obtain characteristic parameters of the rolling bearing, wherein the feature extraction model is used to perform convolution operation processing and time series analysis processing on the input data; The processing module is further used to perform prediction processing on the rolling bearing according to the characteristic parameters to obtain a prediction result, and the prediction result is used to indicate the remaining life of the rolling bearing.

8. The device according to claim 7, characterized in that The device further comprises: a calculation module; The processing module is further used to pre-process the performance indicator data; The processing module is further used to analyze and process the pre-processed performance indicator data of the multiple different time periods to obtain at least one degradation feature of the multiple different time periods; The calculation module is further used to calculate the correlation coefficient between the at least one degradation feature and the health indicator, and construct the undirected graph according to the correlation coefficient.

9. A device for predicting the remaining life of a rolling bearing, characterized in that: include: Memory; processor; Wherein, the memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the remaining life prediction method for a rolling bearing according to any one of claims 1 to 6.

10. A computer storage medium, characterized in that: The computer storage medium stores computer execution instructions, which, when executed by a processor, are used to implement the remaining life prediction method for a rolling bearing according to any one of claims 1 to 6.

Images