Elevator fault diagnosis method and system based on graph convolution

Abstract

The invention discloses an elevator fault diagnosis method and system based on graph convolution. The elevator fault diagnosis method comprises the steps that a vibration signal generated when an elevator runs is converted into a horizontal visible graph; inputting the horizontal views into a graph convolutional network to obtain spatial features of the vibration signals; inputting the spatial feature into a gating circulation unit to obtain a time sequence feature of the vibration signal; according to the method, the spatial features and the time sequence features are fused, elevator faults are diagnosed, the fault features of the elevator in the running state are comprehensively captured through an algorithm combining a graph convolutional network and a gating circulation unit, calculation redundancy is reduced, and diagnosis accuracy and real-time performance are improved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an elevator fault diagnosis method and system based on graph convolution. Background Art

[0002] With the acceleration of urbanization and the increase in the number of high-rise buildings, elevators have become an indispensable vertical transportation tool in modern buildings. Elevators are used frequently and involve the safety of a large number of people, so fault diagnosis is particularly important. The maintenance mode of traditional elevator systems mostly relies on regular manual inspections and basic sensor alarms. This method is not only time-consuming and labor-intensive, but also fails to prevent and detect potential faults in a timely manner, which may lead to unnecessary safety hazards during operation. Therefore, the use of modern information technology, especially deep learning technology, to comprehensively analyze and process elevator operation data has attracted more and more attention, aiming to achieve early warning and prediction of faults through real-time data monitoring and analysis, and improve the safety and reliability of elevator systems.

[0003] Existing elevator fault diagnosis methods mainly focus on analyzing the elevator's operating status by monitoring vibration signals through sensors. However, existing technologies often fail to detect potential faults in a timely and accurate manner, which can easily lead to unstable elevator operation and safety hazards.

[0004] Therefore, there is a need for an improved elevator fault diagnosis method and system that can integrate Internet of Things technology and graph neural network algorithm, gradually transform elevator fault diagnosis from the traditional post-repair mode to an intelligent preventive maintenance mode, and realize efficient and safe operation of the elevator system. Summary of the invention

[0005] The present invention aims to provide an elevator fault diagnosis method and system based on graph convolution, which can solve the above technical problems.

[0006] According to one aspect of the present invention, a method for diagnosing elevator faults based on graph convolution is provided, comprising: obtaining a horizontal visual graph of a vibration signal for the vibration condition during elevator operation; inputting the horizontal visual graph into a graph convolutional network to obtain spatial features of the vibration signal; inputting the spatial features into a gated recurrent unit to obtain temporal features of the vibration signal; and fusing the spatial features and the temporal features to diagnose elevator faults.

[0007] Preferably, for the vibration condition of the elevator during operation, obtaining a horizontal visual graph of the vibration signal includes: screening the vibration signal according to sample integrity; performing data processing on the screened vibration signal; and generating the horizontal visual graph for the processed vibration signal.

[0008] Preferably, for the vibration signal, screening according to the sample integrity includes: using the elevator start acceleration time, the elevator end acceleration time, the elevator start acceleration time and the elevator end acceleration time as the elevator operation integrity judgment nodes; judging the sample integrity of each group of the vibration signals according to the integrity judgment nodes; and screening out all the vibration signals with the sample integrity.

[0009] Preferably, for the selected vibration signal, data processing includes: determining all extreme value points of the selected vibration signal x(t); obtaining an upper envelope curve x(t) of the selected vibration signal according to the extreme value points; max (t) and the lower envelope curve x min (t); obtain the mean curve m(t) of the upper envelope curve and the lower envelope curve according to the following formula: The screened vibration signal is de-envelope processed according to the following formula: h(t)=x(t)-m(t), wherein h(t) represents the processed vibration signal and t represents the acquisition time of the screened vibration signal.

[0010] Preferably, for the processed vibration signal, generating the horizontal visibility graph comprises: obtaining vertices of the graph according to the processed vibration signal; constructing an undirected graph according to the visibility relationship between the vertices; generating the horizontal visibility graph G according to the undirected graph, wherein G = (V, E, A), V = (V1, V2, ..., V N ) is a finite set of vertices, N represents the number of nodes, E=(V1V2,V2V3,…,V N-1 V N ) is the finite set of edges connecting the vertex, A is the adjacency matrix of the undirected graph, and the element in the i-th row and j-th column is denoted by a ij , when the vertex V i and V j Adjacent, then V i and V j There are edges between them, or when the vertex V i and V j Meet V i ,V j >V n , Then V i and V j There are edges connecting them.

[0011] Preferably, inputting the horizontal visibility graph into a graph convolution network to obtain the spatial feature of the vibration signal comprises: inputting the horizontal visibility graph into a first graph convolution layer according to the following formula to obtain a first layer output H1: Among them, Relu() means using the ReLu function as the activation function of the first graph convolutional layer. represents the adjacency matrix A after preprocessing, A matrix with a self-connected structure, is the degree matrix, I N represents the unit matrix, X represents the feature matrix of the horizontal visibility graph, and W0 represents the weight matrix of the first graph convolutional layer; the first layer output H1 is input into the second graph convolutional layer according to the following formula to obtain the spatial feature f(A,X): Among them, H2 represents the output of the second layer, σ() represents the use of Sigmoid function as the activation function of the second graph convolutional layer, and W1 represents the weight matrix of the second graph convolutional layer.

[0012] Preferably, the spatial feature is input into the gated cycle unit to obtain the time series feature of the vibration signal, including: according to the spatial feature f(A, X t ) and the vibration signal h(t-1) processed at the previous moment to obtain the reset gate r in the timing feature t :r t =σ(W r ·[h(t-1),f(A,X t )]+b r ), where σ() represents the Sigmoid function, W r 、b r Represent the reset gate weight and reset gate bias respectively; according to the spatial feature f(A,X t ) and the vibration signal h(t-1) processed at the previous moment to obtain the update gate z in the timing feature t :z t =σ(W z ·[h(t-1),f(A,X t )]+b z ), where σ() represents the Sigmoid function, W z 、b z Represent the update gate weight and update gate bias respectively.

[0013] Preferably, the fusion of the spatial feature and the temporal feature to diagnose the elevator fault includes: resetting the door r t Update the status of the vibration signal: in, Indicates that after the reset gate r t The updated vibration signal, tanh means using the hyperbolic tangent function as the activation function, W, Represent the weight and bias of the reset gate update process respectively; according to the update gate z t Predict the vibration signal h(t) at this moment:

[0014] According to another aspect of the present invention, there is provided an elevator fault diagnosis system based on graph convolution, including: an acquisition module, used to acquire a horizontal visible graph of a vibration signal according to the vibration condition during the operation of the elevator; a first input module, used to input the horizontal visible graph into a graph convolutional network to obtain the spatial features of the vibration signal; a second input module, used to input the spatial features into a gated recurrent unit to obtain the temporal features of the vibration signal; and a diagnosis module, used to fuse the spatial features and the temporal features and diagnose the elevator fault.

[0015] Preferably, the acquisition module includes: a screening submodule for screening the vibration signal according to sample integrity; a processing submodule for performing data processing on the screened vibration signal; and a generating submodule for generating the horizontal visual graph for the processed vibration signal.

[0016] The present invention discloses a method and system for diagnosing elevator faults based on graph convolution, comprising: obtaining a horizontal visual graph of a vibration signal for the vibration condition of an elevator during operation; inputting the horizontal visual graph into a graph convolutional network to obtain spatial features of the vibration signal; inputting the spatial features into a gated recurrent unit to obtain temporal features of the vibration signal; fusing the spatial features with the temporal features to diagnose the elevator fault, and by combining the graph convolutional network and the algorithm of the gated recurrent unit, comprehensively capturing the fault features of the elevator under operation, reducing computational redundancy, and improving the accuracy and real-time performance of the diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0018] Figure 1 is a flow chart of an elevator fault diagnosis method based on graph convolution according to an embodiment of the present invention;

[0019] Figure 2 Schematic diagram of the overall technical solution for elevator fault diagnosis based on graph convolution according to an embodiment of the present invention

[0020] Figure 3 4 is a diagram showing the structure of an elevator fault diagnosis system based on graph convolution according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0022] Figure 1-Figure 2 A flow chart of an elevator fault diagnosis method according to an embodiment of the present invention and a diagram of the overall technical solution for elevator fault diagnosis are shown. As shown in the figure, an embodiment of the present invention provides an elevator fault diagnosis method based on graph convolution, including: obtaining a horizontal visual graph of a vibration signal for the vibration condition during elevator operation; inputting the horizontal visual graph into a graph convolution network to obtain spatial features of the vibration signal; inputting the spatial features into a gated recurrent unit to obtain temporal features of the vibration signal; and fusing the spatial features and the temporal features to diagnose elevator faults.

[0023] In the related art, the traditional elevator fault diagnosis method only analyzes the single feature of the vibration signal and cannot handle the spatiotemporal relationship features well, so the diagnosis result is inaccurate. The embodiment of the present invention can comprehensively capture the fault features of the elevator under the running state, reduce computational redundancy, and improve the accuracy and real-time performance of the diagnosis by combining the graph convolutional network and the gated recurrent unit algorithm.

[0024] Specifically, the vibration signal is monitored by a mobile terminal fixed in the elevator car; the collected original vibration signal data is traversed by a sliding window, and the data of each sliding window is converted into a horizontal visible graph, so that the information of the one-dimensional vibration signal is stored in the graph data structure, thereby realizing the analysis of the time series signal in the form of a horizontal visible graph; a dynamic graph convolution-gated recurrent unit network diagnosis model is constructed, and the horizontal visible graph is extracted by using a graph convolution network, and the features extracted by the graph convolution network are input into the gated recurrent unit in the order of the sliding windows to perform the step of extracting the time series features of the vibration signal. According to the horizontal visible graph and the graph convolution network, a key feature vector for fault diagnosis is obtained; and according to the key feature vector, the gated recurrent unit is input in the order of the sliding windows, so as to train the dynamic graph neural network to the time series regression of the dynamic graph, and complete the fault diagnosis of the elevator.

[0025] According to an embodiment of the present invention, for the vibration condition of an elevator during operation, obtaining a horizontal visual graph of a vibration signal includes: screening the vibration signal according to sample integrity; performing data processing on the screened vibration signal; and generating the horizontal visual graph for the processed vibration signal.

[0026] According to an embodiment of the present invention, for the vibration signal, screening according to the sample integrity includes: using the elevator start acceleration time, the elevator end acceleration time, the elevator start acceleration time and the elevator end acceleration time as the elevator operation integrity judgment node; judging the sample integrity of each group of the vibration signal according to the integrity judgment node; and screening out all the vibration signals with the sample integrity.

[0027] In the embodiment of the present invention, a fixed data acquisition device is set in the elevator to obtain data information of the mobile terminal in real time. The collected data is preprocessed, including data format conversion, missing value processing, and abnormal value detection and correction. Advanced data cleaning algorithms are used to remove noise and interference data to ensure the accuracy and integrity of the data. By analyzing and mining the cleaned data, valuable information is extracted to provide reliable data support for elevator operation status monitoring, fault diagnosis and optimization.

[0028] The embodiment of the present invention proposes an algorithm for dividing the operating time period. The key to judging whether the sample is complete is to determine four time points in the sample, namely, the time when the elevator starts accelerating, the time when acceleration ends, the time when deceleration starts, and the time when deceleration ends.

[0029] According to an embodiment of the present invention, for the selected vibration signal, data processing includes: determining all extreme value points of the selected vibration signal x(t); obtaining an upper envelope curve x(t) of the selected vibration signal according to the extreme value points. max (t) and the lower envelope curve x min (t); obtain the mean curve m(t) of the upper envelope curve and the lower envelope curve according to the following formula: The screened vibration signal is de-envelope processed according to the following formula: h(t)=x(t)-m(t), wherein h(t) represents the processed vibration signal and t represents the acquisition time of the screened vibration signal.

[0030] In the embodiment of the present invention, the elevator operation data set is preprocessed, and a strategy of eliminating abnormal values ​​is adopted to ensure the accuracy and uniqueness of the data and improve the accuracy of fault diagnosis.

[0031] According to an embodiment of the present invention, for the processed vibration signal, generating the horizontal visibility graph includes: obtaining vertices of the graph according to the processed vibration signal; constructing an undirected graph according to the visibility relationship between the vertices; generating the horizontal visibility graph G according to the undirected graph, wherein G = (V, E, A), V = (V1, V2, ..., V N ) is a finite set of vertices, N represents the number of nodes, E=(V1V2,V2V3,…,V N-1 V N ) is the finite set of edges connecting the vertex, A is the adjacency matrix of the undirected graph, and the element in the i-th row and j-th column is denoted by a ij , when the vertex V i and V j Adjacent, then V i and V j There are edges between them, or when the vertex V i and V jMeet V i ,V j >V n , Then V i and V j There are edges connecting them.

[0032] In an embodiment of the present invention, the horizontal visibility graph is an undirected unweighted connected graph. Each element in the vibration signal is used as a node in the graph. An undirected graph is constructed according to the visibility relationship between the nodes to obtain an adjacency matrix of the graph, which is the horizontal visibility graph. The node visibility relationship of the horizontal visibility graph is used to image the one-dimensional time series signal onto the graph, and combined with graph signal processing, a new approach is provided for multi-scale analysis of one-dimensional time series signals.

[0033] According to an embodiment of the present invention, inputting the horizontal visibility graph into a graph convolution network to obtain the spatial features of the vibration signal includes: inputting the horizontal visibility graph into a first graph convolution layer according to the following formula to obtain a first layer output H1: Among them, Relu() means using the ReLu function as the activation function of the first graph convolutional layer. represents the adjacency matrix A after preprocessing, A matrix with a self-connected structure, is the degree matrix, I N represents the unit matrix, X represents the feature matrix of the horizontal visibility graph, and W0 represents the weight matrix of the first graph convolutional layer; the first layer output H1 is input into the second graph convolutional layer according to the following formula to obtain the spatial feature f(A,X): Among them, H2 represents the output of the second layer, σ() represents the use of Sigmoid function as the activation function of the second graph convolutional layer, and W1 represents the weight matrix of the second graph convolutional layer.

[0034] Specifically, according to an embodiment of the present invention, the graph convolutional network includes, in sequence: a first graph convolutional layer, a fully connected layer, a second graph convolutional layer, and a batch normalization layer.

[0035] The specific process is as follows: first, the input data, i.e., the graph data features of the vibration signal, enters the network and is multiplied by the adjacency matrix, then multiplied by the weight matrix kernel, and finally the output of the first graph convolutional layer is obtained through the activation function. The output of the graph convolutional layer is then subjected to dropout regularization processing and then enters the second graph convolutional layer. The processing process of the second graph convolutional layer is the same as that of the first graph convolutional layer. The output of the second graph convolutional layer enters batch processing to obtain the final output of the graph convolutional network.

[0036] In the embodiment of the present invention, the graph convolutional network based on spectral graph theory can accurately characterize the information aggregation between each element in the one-dimensional vibration signal and reasonably extract its spatial features.

[0037] According to an embodiment of the present invention, the spatial feature is input into a gated cycle unit to obtain the time series feature of the vibration signal, including: according to the spatial feature f(A, X t ) and the vibration signal h(t-1) processed at the previous moment to obtain the reset gate r in the timing feature t :r t =σ(W r ·[h(t-1),f(A,X t )]+b r ), where σ() represents the Sigmoid function, W r 、b r Represent the reset gate weight and reset gate bias respectively; according to the spatial feature f(A,X t ) and the vibration signal h(t-1) processed at the previous moment to obtain the update gate z in the timing feature t :z t =σ(W z ·[h(t-1),f(A,X t )]+b z ), where σ() represents the Sigmoid function, W z 、b z Represent the update gate weight and update gate bias respectively.

[0038] According to an embodiment of the present invention, the spatial feature and the time sequence feature are integrated to diagnose the elevator fault, including: resetting the door r t Update the status of the vibration signal: in, Indicates that after the reset gate r t The updated vibration signal, tanh means using the hyperbolic tangent function as the activation function, W, Represent the weight and bias of the reset gate update process respectively; according to the update gate z t Predict the vibration signal h(t) at this moment:

[0039] According to an embodiment of the present invention, the gated recurrent unit network can efficiently capture important information in the sequence, reduce information loss, and improve the accuracy of feature extraction through the gating mechanism of the gated recurrent unit. The gated recurrent unit is good at processing time series data, and controls the fusion of current information and past information through the update gate, determines the update degree of the current hidden state, and realizes efficient time series processing.

[0040] According to an embodiment of the present invention, the graph convolutional network elevator fault diagnosis module is used to extract spatial features, and the extracted spatial features are fused as input variables of the gated recurrent unit elevator fault diagnosis module to construct the elevator fault diagnosis model, and finally output the elevator fault diagnosis prediction. In an embodiment of the present invention, by combining the graph convolutional network with the gated recurrent unit, the spatial and temporal characteristics of the elevator vibration signal are comprehensively considered, so that the model can not only capture the temporal dynamics of the vibration signal, but also deeply extract the spatial characteristics of the vibration signal of different faults, and can better cope with complex situations in practical applications.

[0041] In the embodiment of the present invention, a new feature extraction approach is proposed to address the problem of low accuracy in vibration signal feature extraction in the prior art, in which the features defined on a one-dimensional time series signal are extended to a graph structure; the embodiment of the present invention can effectively diagnose different types of elevator faults, and has high robustness, can provide a higher accuracy rate for the fault diagnosis results of the elevator, realizes accurate diagnosis of elevator fault types, can effectively monitor elevator faults, and improves the safety and economy of elevator operation.

[0042] Figure 3 FIG. 2 shows a structure diagram of an elevator fault diagnosis system according to an embodiment of the present invention. Figure 3 As shown, another embodiment of the present invention provides an elevator fault diagnosis system, including: an acquisition module, used to acquire a horizontal visual graph of a vibration signal according to the vibration condition during the operation of the elevator; a first input module, used to input the horizontal visual graph into a graph convolutional network to obtain the spatial features of the vibration signal; a second input module, used to input the spatial features into a gated recurrent unit to obtain the temporal features of the vibration signal; and a diagnosis module, used to fuse the spatial features and the temporal features and diagnose the elevator fault.

[0043] The elevator system fault diagnosis system based on graph convolution in the embodiment of the present invention comprises an acquisition module, which is used to monitor the vibration signal through a fixed mobile terminal in the elevator when the elevator is started, accelerated, uniformly accelerated, decelerated and stopped, and to traverse the collected original vibration signal data using a sliding window, and to convert the data of each sliding window into a horizontal visible graph, so that the information of the one-dimensional vibration signal is stored in the graph data structure, thereby realizing the analysis of the time series signal in the form of a horizontal visible graph; a first input module, which uses a graph convolution network to extract features of the horizontal visible graph, obtains a graph matrix representation based on spectral graph theory, including an adjacency matrix, a degree diagonal matrix and a Laplace matrix, and obtains the eigenvalues ​​and eigenvectors of the normalized Laplace matrix for extracting the spatial features of the vibration signal; a second input module, which is used to input the features extracted by the graph convolution network into a gated recurrent unit in the order of the sliding windows to perform the extraction of the time series features of the vibration signal; a diagnosis module, i.e., an output module, fuses the time series features with the spatial features to construct a composite feature matrix, which is a set of time series with spatial associations, and performs fault diagnosis on the elevator based on the constructed fault diagnosis module.

[0044] In the related art, the traditional elevator system fault diagnosis system only analyzes the single feature of the vibration signal and cannot handle the spatiotemporal relationship features well, so the diagnosis result is inaccurate. In the embodiment of the present invention, by monitoring and analyzing the vibration signal, and then combining the horizontal visibility graph method, it is possible to fully capture the fault characteristics under the operation state of the elevator, effectively screen the key features, and further combine the graph convolution network-gated recurrent unit model to reduce computational redundancy and improve the accuracy and real-time performance of the diagnosis.

[0045] According to an embodiment of the present invention, based on a series of methods for extracting time series and space series features of vibration signals, it is found that the data of vibration signals constitute a set of time series with spatial correlation, which has both time attributes and space attributes, and has space-time characteristics with strong space-time coupling.

[0046] According to an embodiment of the present invention, the acquisition module includes: a screening submodule, used to screen the vibration signal according to sample integrity; a processing submodule, used to perform data processing on the screened vibration signal; and a generating submodule, used to generate the horizontal visual graph for the processed vibration signal.

[0047] Specifically, the embodiment of the present invention makes full use of the sensing capabilities of modern mobile devices to replace traditional IoT edge devices. The system uses a mobile phone to obtain elevator vibration data with a sampling frequency of up to 500Hz, ensuring high accuracy and comprehensiveness of the data. During the data collection process, the mobile device is fixed inside the elevator to monitor the vibration of the elevator in real time. The collected vibration signals provide a basis for subsequent data processing. The system removes incomplete samples through a series of data cleaning and preprocessing processes to ensure that each sample fully covers all stages of elevator operation, including static, acceleration, uniform speed, deceleration and stop states. In order to improve the effectiveness of data analysis, the present invention conducts in-depth research on feature extraction. First, through the node visibility relationship of the horizontal visibility graph, the one-dimensional time series signal is mapped to the graph, and combined with the graph signal processing, a new approach is provided for the multi-scale analysis of the one-dimensional time series signal. The horizontal visibility graph extracts a variety of statistical features that can reflect the time domain vibration characteristics of the elevator at different operating stages.

[0048] In spatial feature analysis, graph convolutional networks are applied to signal processing to convert vibration signals into graph data. By extracting the spatial features of vibration signal graph data, the system obtains rich spatial features of vibration signals. The present invention uses gated recurrent units for fault classification. The system can make full use of the advantages of gated recurrent units in processing time series data, control the fusion of current information and past information through update gates, determine the update degree of the current hidden state, and achieve efficient time series processing, thereby solving complex classification problems.

[0049] In general, the embodiment of the present invention provides an elevator fault detection and diagnosis system based on mobile devices, which has the advantages of high accuracy, high efficiency and low cost. By comprehensively utilizing the temporal and spatial characteristics of vibration signals, combined with advanced deep learning models and feature selection methods, the system provides strong technical support for the safe operation of elevators.

[0050] The present invention has the following effects:

[0051] (1) The feature enhancement model based on the horizontal visible phantom is capable of extracting spatial features, which can fully express the intrinsic characteristics of the elevator vibration signal and effectively distinguish different types of elevator faults.

[0052] (2) Traditional time series modeling methods cannot handle data with high complexity of spatial and temporal relationships. The present invention optimizes the drawbacks of traditional neural networks in processing non-traditional data structures. By combining graph convolutional networks with gated recurrent units, the spatial and temporal characteristics of elevator vibration signals are comprehensively considered, so that the model can not only capture the temporal dynamics of vibration signals, but also deeply extract the spatial characteristics of vibration signals of different faults, which can significantly improve the accuracy of fault diagnosis.

[0053] (3) The comprehensive fusion technology solution provided by the present invention significantly improves the accuracy and robustness of elevator fault diagnosis prediction, and can better cope with complex situations in practical applications. Through this comprehensive approach, the diagnostic model can provide more reliable decision support and optimize the daily operation and management of elevators.

[0054] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0055] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An elevator fault diagnosis method based on graph convolution, characterized in that: include: For the vibration condition of the elevator during operation, obtain a horizontal visual graph of the vibration signal; Inputting the horizontal visible graph into a graph convolutional network to obtain the spatial features of the vibration signal; Inputting the spatial features into a gated recurrent unit to obtain a temporal feature of the vibration signal; The spatial features and the temporal features are integrated to diagnose elevator faults.

2. The diagnostic method according to claim 1, characterized in that For the vibration of the elevator during operation, the horizontal visual diagram of the vibration signal includes: For the vibration signal, screening is performed according to sample integrity; Performing data processing on the screened vibration signals; For the processed vibration signal, the horizontal visual graph is generated.

3. The diagnostic method according to claim 2, characterized in that For the vibration signal, screening based on sample integrity includes: The elevator acceleration start time, elevator acceleration end time, elevator acceleration start time and elevator acceleration end time are used as elevator operation integrity judgment nodes; Determining the sample integrity of each group of the vibration signals according to the integrity determination node; All the vibration signals having the sample integrity are screened out.

4. The diagnostic method according to claim 3, characterized in that: For the screened vibration signal, data processing includes: Determine all extreme points of the screened vibration signal x(t); According to the extreme point, the upper envelope curve x of the selected vibration signal is obtained. max (t) and the lower envelope curve x min (t); The mean curve m(t) of the upper envelope curve and the lower envelope curve is obtained according to the following formula: According to the following formula, the filtered vibration signal is processed to remove the envelope: h(t)=x(t)-m(t) Among them, h(t) represents the processed vibration signal, and t represents the acquisition time of the screened vibration signal.

5. The diagnostic method according to claim 4, characterized in that: For the processed vibration signal, generating the horizontal visual graph comprises: According to the processed vibration signal, the vertex of the spectrum is obtained; Constructing an undirected graph according to the visibility relationship between the vertices; According to the undirected graph, the horizontal visible graph G is generated, Among them, G=(V,E,A), V=(V1,V2,…,V N ) is a finite set of vertices, N represents the number of nodes, E=(V1V2,V2V3,…,V N-1 V N ) is a finite set of edges connecting the vertices, A is the adjacency matrix of the undirected graph, and the element in the i-th row and j-th column is denoted by a ij , when the vertex V i and V j Adjacent, then V i and V j There are edges between them, or when the vertex V i and V j Meet V i ,V j >V n , Then V i and V j There are edges connecting them.

6. The diagnostic method according to claim 5, characterized in that: The horizontal visibility graph is input into a graph convolutional network to obtain the spatial features of the vibration signal, including: The horizontal visibility graph is input into the first graph convolutional layer according to the following formula to obtain the first layer output H1: Wherein, Relu() indicates that the ReLu function is used as the activation function of the first graph convolutional layer. represents the adjacency matrix A after preprocessing, A matrix with a self-connected structure, is the degree matrix, I N represents the identity matrix, X represents the feature matrix of the horizontal visible graph, and W0 represents the weight matrix of the first graph convolutional layer; The first layer output H1 is input into the second graph convolution layer according to the following formula to obtain the spatial feature f(A,X): Among them, H2 represents the second layer output, σ() represents the use of Sigmoid function as the activation function of the second graph convolution layer, and W1 represents the weight matrix of the second graph convolution layer.

7. The diagnostic method according to claim 6, characterized in that: Inputting the spatial features into the gated recurrent unit to obtain the temporal features of the vibration signal includes: According to the spatial feature f(A, X t ) and the vibration signal h(t-1) processed at the previous moment to obtain the reset gate r in the timing characteristics t : r t =σ(W r ·[h(t-1),f(A,X t )]+b r ) Among them, σ() represents the Sigmoid function, W r 、b r Represent the reset gate weight and reset gate bias respectively; According to the spatial feature f(A, X t ) and the vibration signal h(t-1) processed at the previous moment to obtain the update gate z in the timing feature t : z t =σ(W z ·[h(t-1),f(A,X t )]+b z ) Among them, σ() represents the Sigmoid function, W z 、b z Represent the update gate weight and update gate bias respectively.

8. The diagnostic method according to claim 7, characterized in that: Combining the spatial features and the temporal features to diagnose elevator faults includes: According to the reset gate r t Update the status of the vibration signal: in, Indicates that the reset gate r t The updated vibration signal, tanh means using the hyperbolic tangent function as the activation function, W, Represent the weight and bias of the reset gate update process respectively; According to the update gate z t Predict the vibration signal h(t) at this moment:

9. An elevator fault diagnosis system based on graph convolution, comprising: An acquisition module, used for acquiring a horizontal visual graph of a vibration signal according to the vibration condition of the elevator during operation; A first input module, used for inputting the horizontal visible graph into a graph convolutional network to obtain the spatial features of the vibration signal; A second input module, used for inputting the spatial feature into a gated recurrent unit to obtain a temporal feature of the vibration signal; The diagnosis module is used to fuse the spatial features and the temporal features and diagnose elevator faults.

10. The diagnostic system according to claim 9, characterized in that The acquisition module comprises: A screening submodule, used for screening the vibration signal according to sample integrity; A processing submodule is used to perform data processing on the screened vibration signal; The generating submodule is used to generate the horizontal visible graph for the processed vibration signal.

Images