Equipment Cooperative Monitoring System and Method in Bucket Wheel Continuous System

Abstract

A collaborative monitoring system and method for equipment in a continuous bucket wheel system are disclosed. Based on the operating characteristics of the first to fourth belt conveyors and their transmission information, a collaborative monitoring model of the transmission chain composed of the first to fourth belt conveyors is constructed to intelligently determine whether there are any faults in the collaborative conveying between the first to fourth belt conveyors. This ensures that the highly coupled transmission chain composed of the first to fourth belt conveyors does not experience faults during the operation of the continuous bucket wheel system.

Description

Technical Field

[0001] This invention relates to the field of equipment monitoring technology, and more specifically to a collaborative monitoring system and method for equipment in a continuous bucket wheel system. Background Technology

[0002] A coal mine is a rationally excavated space created by humans when mining coal-rich geological strata, typically including roadways, shafts, and mining faces. In the past, coal and soil produced in coal mines were transported out of the depressions by trucks. However, as mines have been dug deeper and deeper, the transportation distances for trucks have become increasingly longer, and the original efficient transportation is no longer possible. As a result, continuous wheel bucket systems for coal mines have emerged.

[0003] A coal mine continuous conveyor system includes a bucket wheel excavator, a transfer conveyor, a cable hopper car, L1 belt conveyors, L2 belt conveyors, L3 belt conveyors, L4 belt conveyors, an unloading car, and a spoil disposal machine. Among these, the coordinated monitoring of equipment within the continuous conveyor system is particularly important, as it affects not only the normal operation of the system but also its working efficiency. However, since a coal mine continuous conveyor system comprises multiple pieces of equipment, and the coordination patterns between these equipment change with operating conditions, coordinated monitoring of equipment within the continuous conveyor system is technically challenging.

[0004] Therefore, a collaborative monitoring system for equipment in a continuous bucket wheel system is needed. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide an equipment coordination monitoring system and method for a continuous bucket wheel system. Based on the operating state characteristics of the first to fourth belt conveyors and the transmission information of each belt conveyor, it constructs an equipment coordination monitoring model of the transmission chain composed of the first to fourth belt conveyors to intelligently determine whether there is a fault in the coordination of the conveying between the first to fourth belt conveyors. This ensures that the highly coupled transmission chain composed of the first to fourth belt conveyors does not experience faults during the operation of the continuous bucket wheel system.

[0006] According to one aspect of this application, a collaborative monitoring system for equipment in a bucket wheel continuous system is provided, comprising:

[0007] The working status data acquisition unit is used to acquire the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period.

[0008] The working state feature extraction unit is used to arrange the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into power input vectors according to the time dimension, and then use the multi-scale neighborhood extraction module to obtain the first to fourth power feature vectors.

[0009] A globalization unit is used to arrange the first to fourth power feature vectors in a two-dimensional arrangement to obtain a global power feature matrix.

[0010] The transfer unit is used to calculate the transfer matrix between every two power feature vectors in the first to fourth power feature vectors to obtain multiple transfer matrices;

[0011] A transfer topology construction unit is used to calculate the global mean of each of the multiple transfer matrices to obtain multiple transfer feature values, and to arrange the multiple transfer feature values ​​in a two-dimensional manner to obtain a transfer topology matrix;

[0012] A transfer topology feature extraction unit is used to pass the transfer topology matrix through a convolutional neural network model as a feature extractor to obtain a transfer topology feature matrix.

[0013] A graph neural network encoding unit is used to pass the global power feature matrix and the transfer topology feature matrix through a graph neural network model to obtain a transfer topology global power feature matrix; and

[0014] The equipment collaborative monitoring result generation unit is used to pass the global power feature matrix of the transfer topology through a classifier to obtain a classification result, which is used to indicate whether there is a fault in the conveying collaboration between the first to the fourth belt conveyors.

[0015] In the aforementioned equipment collaborative monitoring system of the continuous bucket wheel system, the working state feature extraction unit includes: a first-scale feature extraction subunit, used to input the power input vector of the first belt conveyor into the first convolutional layer of the multi-scale neighborhood feature extraction module to obtain a first-scale power vector, wherein the first convolutional layer has a first one-dimensional convolutional kernel of a first length; a second-scale feature extraction subunit, used to input the power input vector of the first belt conveyor into the second convolutional layer of the multi-scale neighborhood feature extraction module to obtain a second-scale power feature vector, wherein the second convolutional layer has a second one-dimensional convolutional kernel of a second length, the first length being different from the second length; and a multi-scale cascading subunit, used to cascade the first-scale power vector and the second-scale power vector to obtain the first power feature vector.

[0016] In the aforementioned equipment collaborative monitoring system of the continuous bucket wheel system, the first-scale feature extraction subunit is further configured to: use the first convolutional layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolutional encoding on the power input vector of the first belt conveyor to obtain the first-scale power feature vector using the following formula; wherein, the formula is:

[0017]

[0018] Where a is the width of the first convolution kernel in the x direction, F(a) is the parameter vector of the first convolution kernel, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the power input vector of the first belt conveyor.

[0019] In the aforementioned equipment collaborative monitoring system of the continuous bucket wheel system, the second-scale feature extraction subunit is further configured to: use the second convolutional layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolutional encoding on the power input vector of the first belt conveyor to obtain the second-scale power feature vector using the following formula; wherein, the formula is:

[0020]

[0021] Where b is the width of the second convolution kernel in the x direction, F(b) is the parameter vector of the second convolution kernel, G(xb) is the local vector matrix of the operation with the convolution kernel function, m is the size of the second convolution kernel, and X represents the power input vector of the first belt conveyor.

[0022] In the aforementioned equipment collaborative monitoring system of the continuous bucket wheel system, the transfer unit is further configured to: calculate the transfer matrix between every two power feature vectors in the first to fourth power feature vectors using the following formula to obtain multiple transfer matrices; wherein, the formula is:

[0023]

[0024] Where V a V represents the first of every two power eigenvectors in the first to fourth power eigenvectors. b This represents the second power eigenvector among every two power eigenvectors in the first to fourth power eigenvectors, and M represents one of the multiple transition matrices. This represents matrix multiplication.

[0025] In the aforementioned equipment collaborative monitoring system of the continuous bucket wheel system, the transfer topology feature extraction unit includes: each layer of the convolutional neural network model performing the following during the forward propagation of the layer: convolution processing on the input data to obtain a convolutional feature map; mean pooling based on the local feature matrix on the convolutional feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the convolutional neural network model is the transfer topology feature matrix, and the input of the first layer of the convolutional neural network model is the transfer topology matrix.

[0026] In the aforementioned equipment collaborative monitoring system of the continuous bucket wheel system, the equipment collaborative monitoring result generation unit is further configured to: process the global power feature matrix of the transfer topology using the classifier according to the following formula to obtain the classification result; wherein, the formula is:

[0027] softmax{(W n B n ):…:(W1,B1)|Project(M)}

[0028] Where Project(M) represents projecting the global power feature matrix of the transfer topology into a vector, W1 to W... n Here are the weight matrices for each fully connected layer, B1 to B... n This represents the bias vector of each fully connected layer.

[0029] In the aforementioned equipment collaborative monitoring system of the continuous bucket wheel system, a training module is also included for training the multi-scale neighborhood extraction module, the convolutional neural network model as a feature extractor, and the classifier. The training module includes: a training data acquisition unit for acquiring training power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period, and the actual value indicating whether there is a fault in the conveying coordination between the first to fourth belt conveyors; a training working state feature extraction unit for arranging the training power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into training power input vectors according to the time dimension, and then passing them through the multi-scale neighborhood extraction module to obtain the first to fourth training power feature vectors; a training globalization unit for arranging the first to fourth training power feature vectors in a two-dimensional manner to obtain a training global power feature matrix; a training transition unit for calculating the training transition matrix between every two training power feature vectors in the first to fourth training power feature vectors to obtain multiple training transition matrices; and a training transition topology construction unit for calculating the... The system comprises: a training transition matrix and a training transition topology matrix; a training transition topology feature extraction unit, which extracts the training transition topology matrix by passing it through the convolutional neural network model (which acts as a feature extractor) to obtain a training transition topology feature matrix; a training graph neural network encoding unit, which extracts the training global power feature matrix and the training transition topology feature matrix through the graph neural network model to obtain a training transition topology global power feature matrix; a classification loss unit, which extracts the transition topology global power feature matrix through the classifier to obtain a classification loss function value; a multi-distribution binary classification quality loss unit, which calculates the multi-distribution binary classification quality loss function value of the row vectors of the transition topology global power feature matrix; and a training unit, which trains the multi-scale neighborhood extraction module, the convolutional neural network model (which acts as a feature extractor), and the classifier using the weighted sum of the multi-distribution binary classification quality loss function value and the classification loss function value as the loss function value.

[0030] In the equipment collaborative monitoring system of the aforementioned bucket wheel continuous system, the multi-distribution binary classification quality loss unit is further used to: calculate the multi-distribution binary classification quality loss function value of the row vector of the global power feature matrix of the transfer topology using the following formula; wherein, the formula is:

[0031]

[0032] Among them, V1 to V n V represents the row vectors of the global power characteristic matrix of the transfer topology. rIt is a reference vector, and The classification result of the feature vector is represented by ||·||1, which represents the 1-norm of the vector, and log represents the logarithmic function operation with base 2.

[0033] According to another aspect of this application, a method for coordinated monitoring of equipment in a continuous bucket wheel system is also provided, comprising:

[0034] Obtain the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period;

[0035] The power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period are arranged into power input vectors according to the time dimension, and then the first to fourth power feature vectors are obtained by the multi-scale neighborhood extraction module.

[0036] The first to fourth power feature vectors are arranged in a two-dimensional arrangement to obtain the global power feature matrix;

[0037] Calculate the transition matrix between every two power eigenvectors in the first to fourth power eigenvectors to obtain multiple transition matrices;

[0038] Calculate the global mean of each of the multiple transition matrices to obtain multiple transition feature values, and arrange the multiple transition feature values ​​in a two-dimensional arrangement to obtain a transition topology matrix;

[0039] The transition topology matrix is ​​passed through a convolutional neural network model as a feature extractor to obtain a transition topology feature matrix;

[0040] The global power feature matrix and the transfer topology feature matrix are processed through a graph neural network model to obtain the transfer topology global power feature matrix; and

[0041] The global power feature matrix of the transfer topology is passed through a classifier to obtain a classification result, which is used to indicate whether there is a fault in the conveying coordination between the first to fourth belt conveyors.

[0042] The equipment collaborative monitoring method in the aforementioned bucket wheel continuous system further includes training the multi-scale neighborhood extraction module, the convolutional neural network model serving as the feature extractor, and the classifier. The training process includes: acquiring training power values ​​for the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period, and the actual values ​​indicating whether there is a fault in the conveyor coordination between the first to fourth belt conveyors; arranging the training power values ​​for the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into training power input vectors according to the time dimension, and then passing them through the multi-scale neighborhood extraction module to obtain first to fourth training power feature vectors; arranging the first to fourth training power feature vectors in a two-dimensional arrangement to obtain a training global power feature matrix; and calculating the training transition between every two training power feature vectors in the first to fourth training power feature vectors. The process involves: obtaining multiple training transition matrices; calculating the global mean of each training transition matrix to obtain multiple training transition feature values; arranging these multiple training transition feature values ​​in a two-dimensional manner to obtain a training transition topology matrix; passing the training transition topology matrix through the convolutional neural network model acting as a feature extractor to obtain a training transition topology feature matrix; passing the training global power feature matrix and the training transition topology feature matrix through the graph neural network model to obtain a training transition topology global power feature matrix; passing the transition topology global power feature matrix through the classifier to obtain a classification loss function value; calculating the multi-distribution binary classification quality loss function value of the row vectors of the transition topology global power feature matrix; and training the multi-scale neighborhood extraction module, the convolutional neural network model acting as a feature extractor, and the classifier using the weighted sum of the multi-distribution binary classification quality loss function value and the classification loss function value as the loss function value.

[0043] Compared with existing technologies, the equipment coordination monitoring system and method in the continuous bucket wheel system provided in this application constructs an equipment coordination monitoring model of the transmission chain composed of the first to fourth belt conveyors based on the working state characteristics of the first to fourth belt conveyors and the transmission information of each belt conveyor. This intelligently determines whether there is a fault in the coordination of the conveying between the first to fourth belt conveyors. In this way, it ensures that the highly coupled transmission chain composed of the first to fourth belt conveyors does not experience faults during the operation of the continuous bucket wheel system. Attached Figure Description

[0044] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0045] Figure 1 This is a schematic diagram of a scenario for a collaborative monitoring system for equipment in a continuous bucket wheel system according to an embodiment of this application.

[0046] Figure 2 This is a block diagram of the equipment collaborative monitoring system in a continuous bucket wheel system according to an embodiment of this application.

[0047] Figure 3 This is a system architecture diagram of the equipment collaborative monitoring system in a continuous bucket wheel system according to an embodiment of this application.

[0048] Figure 4 This is a block diagram of the working status feature extraction unit in the equipment collaborative monitoring system of the bucket wheel continuous system according to an embodiment of this application.

[0049] Figure 5 This is a block diagram of the training module in the equipment collaborative monitoring system of the bucket wheel continuous system according to an embodiment of this application.

[0050] Figure 6 This is a flowchart of a method for coordinated monitoring of equipment in a continuous bucket wheel system according to an embodiment of this application.

[0051] Figure 7 This is a flowchart illustrating the training of the multi-scale neighborhood extraction module, the convolutional neural network model serving as a feature extractor, and the classifier in the equipment collaborative monitoring method of a continuous bucket wheel system according to an embodiment of this application. Detailed Implementation

[0052] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0053] Application Overview

[0054] As mentioned in the background section, the coordinated monitoring of equipment within the continuous bucket wheel system is particularly important during its operation, as it affects not only the system's normal operation but also its working efficiency. However, a coal mine continuous bucket wheel system comprises multiple pieces of equipment, and the coordination patterns between these devices change with operating conditions. Therefore, coordinated monitoring of equipment within the continuous bucket wheel system is technically challenging. Consequently, a coordinated monitoring system for equipment within a continuous bucket wheel system is desired.

[0055] In the technical solution of this application, the applicant notes that the entire continuous bucket wheel system comprises three stages: coal mining, coal conveying, and coal unloading. Coal conveying is primarily achieved by belt conveyors L1, L2, L3, and L4, which form a highly coupled conveyor chain. Therefore, in the technical solution of this application, the applicant attempts to implement coordinated equipment monitoring of the conveyor chain portion of the continuous bucket wheel system to reduce monitoring difficulty and ensure the coal conveying process.

[0056] For ease of explanation, in the technical solution of this application, belt conveyors L1, L2, L3, and L4 are defined as the first to fourth belt conveyors. Then, in the equipment collaborative monitoring, the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period are first acquired, and these power values ​​are arranged according to the time dimension to form a power input vector. That is, the working state of the first to fourth belt conveyors is represented by the time-series vector of their power values.

[0057] Next, the power input vector is processed by a multi-scale neighborhood extraction module to obtain the first to fourth power feature vectors. That is, the multi-scale feature extraction module, which includes multiple parallel one-dimensional convolutional layers, extracts the working feature vectors of the first to fourth belt conveyors. Here, during encoding, the multi-scale neighborhood feature extraction module uses one-dimensional convolutional kernels of different scales to perform multi-scale one-dimensional convolutional encoding on the time-series vector of the power values ​​to extract the distribution features of the power values ​​within different time spans, i.e., the working state features of the first to fourth belt conveyors within different time spans. Furthermore, the first to fourth power feature vectors are arranged in a two-dimensional array to obtain a global power feature matrix; that is, the first to fourth power feature vectors are structurally integrated in a high-dimensional feature space to obtain the global power feature matrix.

[0058] Furthermore, in the technical solution of this application, the first to fourth belt conveyors operate under specific coal transport rules, which can be represented by the transition probabilities between the working state characteristics of the first to fourth belt conveyors. Moreover, integrating the above information into equipment collaborative monitoring would obviously help improve the accuracy of equipment collaborative monitoring.

[0059] Specifically, in the technical solution of this application, a transfer matrix between every two power eigenvectors in the first to fourth power eigenvectors is calculated to obtain multiple transfer matrices. That is, in the technical solution of this application, the transfer matrix between every two power eigenvectors represents the coal transfer information between two belt conveyors. Furthermore, the global mean of each of the multiple transfer matrices is calculated to obtain multiple transfer eigenvalues, and the multiple transfer eigenvalues ​​are arranged in a two-dimensional arrangement to obtain a transfer topology matrix. That is, the global mean of each transfer matrix represents the coal transfer probability between the two belt conveyors, and the obtained coal transfer probabilities are arranged in a two-dimensional arrangement to obtain the transfer topology matrix.

[0060] Next, the transition topology matrix is ​​processed by a convolutional neural network model acting as a feature extractor to obtain a transition topology feature matrix. Those skilled in the art will know that convolutional neural network models excel at extracting locally correlated features; therefore, processing the transition topology matrix using the convolutional neural network model as a feature extractor can extract the correlation information between coal transfer characteristics between belt conveyors.

[0061] Furthermore, the first to fourth power feature vectors are used as the node feature representations of each belt conveyor, and the transfer topology feature matrix is ​​used as the feature representation of the edges between nodes. A graph neural network model is used for encoding to extract a global power feature matrix for the transfer topology, which includes irregular transfer topology information and high-dimensional operating state features of each belt conveyor. That is, the global power feature matrix and the transfer topology feature matrix are processed through a graph neural network model to obtain the global power feature matrix for the transfer topology. After obtaining the global power feature matrix for the transfer topology, it is classified using a classifier to obtain a classification result indicating whether there is a fault in the conveying coordination between the first to fourth belt conveyors.

[0062] In summary, based on the working status characteristics and transmission information of each belt conveyor, a device collaborative monitoring model is constructed for the transmission chain consisting of the first to fourth belt conveyors, so as to intelligently determine whether there is a fault in the transmission collaboration between the first to fourth belt conveyors.

[0063] Specifically, in the technical solution of this application, when obtaining the global power feature matrix and the transfer topology feature matrix through a graph neural network model, the dimension of the transfer topology feature matrix is ​​l×l, where l is the length of the power feature vector. However, to obtain the global power feature matrix by arranging the first to fourth power feature vectors in two dimensions, it is necessary to perform low-cost augmentation on the power feature vectors in the sample dimension. This makes the correlation between each row vector of the global power feature matrix (assuming the first to fourth power feature vectors are row vectors) worse. Correspondingly, the correlation between each row vector of the transfer topology global power feature matrix also worsens, affecting the overall feature representation capability of the transfer topology global power feature matrix.

[0064] Therefore, it is necessary to assign weights to each row vector of the global power feature matrix of the transfer topology to solve this problem. However, since the weights need to be obtained through training as hyperparameters, this will increase the training burden of the model. Therefore, the applicant of this application introduces a multi-distribution binary classification quality loss function for the row vectors of the global power feature matrix of the transfer topology, expressed as:

[0065]

[0066] Among them, V1 to V n V represents the row vectors of the global power characteristic matrix of the transfer topology. r It is a reference vector, preferably set as the mean eigenvector of all row vectors of the global power characteristic matrix of the transfer topology, and The classification result of the feature vector is represented by ||·||1, which represents the 1-norm of the vector.

[0067] Here, to avoid difficulties in convergence to the target class domain due to excessive fragmentation of the decision boundaries corresponding to the local feature distributions of each row vector in the global power feature matrix of the transfer topology during multi-distribution classification, the continuity quality factor of the local feature distribution of each row vector relative to the global feature distribution of the global power feature matrix of the transfer topology can be predicted as a loss function by averaging the global offset class probability information of the local feature distribution of each row vector based on the binary classification of the predetermined label. By training the model in this way, the optimization of hyperparameters during training can be transformed from backpropagation into a classification problem based on multi-binary classification, and the overall feature representation capability of the global power feature matrix of the transfer topology can be improved. This improves the accuracy of classifying whether there is a fault in the conveyor coordination between the first to fourth belt conveyors.

[0068] Based on this, this application proposes a collaborative monitoring system for equipment in a continuous bucket wheel system, comprising: a working status data acquisition unit for acquiring power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period; a working status feature extraction unit for arranging the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into power input vectors according to the time dimension, and then obtaining first to fourth power feature vectors through a multi-scale neighborhood extraction module; a globalization unit for arranging the first to fourth power feature vectors in a two-dimensional manner to obtain a global power feature matrix; a transfer unit for calculating the transfer matrix between every two power feature vectors in the first to fourth power feature vectors to obtain multiple transfer matrices; and a transfer topology. The system comprises: a topology generation unit, used to calculate the global mean of each of the plurality of transfer matrices to obtain a plurality of transfer feature values, and to arrange the plurality of transfer feature values ​​in a two-dimensional manner to obtain a transfer topology matrix; a transfer topology feature extraction unit, used to pass the transfer topology matrix through a convolutional neural network model as a feature extractor to obtain a transfer topology feature matrix; a graph neural network encoding unit, used to pass the global power feature matrix and the transfer topology feature matrix through a graph neural network model to obtain a transfer topology global power feature matrix; and a device collaborative monitoring result generation unit, used to pass the transfer topology global power feature matrix through a classifier to obtain a classification result, the classification result being used to indicate whether there is a fault in the conveying collaboration between the first to fourth belt conveyors.

[0069] Figure 1 This is a schematic diagram of a scenario involving a collaborative monitoring system for equipment in a continuous bucket wheel system according to an embodiment of this application. Figure 1 As shown, in this application scenario, firstly, the first to fourth power sensors (e.g., such as...) are used. Figure 1 Se1 to Se4 (as shown) respectively acquire the first to fourth belt conveyors (e.g., such as Figure 1 The power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period (C1 to C4) are shown. Then, the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within the predetermined time period are input to a server (e.g., a server equipped with a collaborative monitoring algorithm for equipment in a bucket wheel continuous system) that is configured with such an algorithm. Figure 1 In the illustrated S), the server is able to process the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period based on the equipment collaborative monitoring algorithm in the bucket wheel continuous system, so as to obtain a classification result indicating whether there is a fault in the conveying coordination between the first to fourth belt conveyors.

[0070] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0071] Exemplary System

[0072] Figure 2 This is a block diagram of the equipment collaborative monitoring system in a bucket wheel continuous system according to an embodiment of this application. Figure 2 As shown, the equipment collaborative monitoring system 100 in the continuous bucket wheel system according to an embodiment of this application includes: a working status data acquisition unit 110, used to acquire the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period; a working status feature extraction unit 120, used to arrange the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into power input vectors according to the time dimension, and then obtain the first to fourth power feature vectors through a multi-scale neighborhood extraction module; a globalization unit 130, used to arrange the first to fourth power feature vectors in a two-dimensional arrangement to obtain a global power feature matrix; a transfer unit 140, used to calculate the transfer matrix between every two power feature vectors in the first to fourth power feature vectors to obtain multiple transfer matrices; and a transfer topology. The system comprises: a construction unit 150, which calculates the global mean of each of the plurality of transfer matrices to obtain a plurality of transfer feature values, and arranges the plurality of transfer feature values ​​in a two-dimensional manner to obtain a transfer topology matrix; a transfer topology feature extraction unit 160, which passes the transfer topology matrix through a convolutional neural network model as a feature extractor to obtain a transfer topology feature matrix; a graph neural network encoding unit 170, which passes the global power feature matrix and the transfer topology feature matrix through a graph neural network model to obtain a transfer topology global power feature matrix; and a device collaborative monitoring result generation unit 180, which passes the transfer topology global power feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether there is a fault in the conveying collaboration between the first to fourth belt conveyors.

[0073] Figure 3 This is a system architecture diagram of the equipment collaborative monitoring system 100 in a continuous bucket wheel system according to an embodiment of this application. Figure 3As shown, in the system architecture of the equipment collaborative monitoring system 100 in the bucket wheel continuous system, firstly, the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period are acquired. Next, the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within the predetermined time period are arranged according to the time dimension into power input vectors, and then processed by a multi-scale neighborhood extraction module to obtain first to fourth power feature vectors. Then, the first to fourth power feature vectors are arranged in a two-dimensional manner to obtain a global power feature matrix. Furthermore, the transition matrix between every two power feature vectors in the first to fourth power feature vectors is calculated to obtain multiple transition matrices. Simultaneously, the global mean of each of the multiple transition matrices is calculated to obtain multiple transition feature values, and these multiple transition feature values ​​are arranged in a two-dimensional manner to obtain a transition topology matrix. Next, the transition topology matrix is ​​processed by a convolutional neural network model as a feature extractor to obtain a transition topology feature matrix. Finally, the global power feature matrix and the transition topology feature matrix are processed by a graph neural network model to obtain a global power feature matrix for the transition topology. Then, the global power feature matrix of the transfer topology is passed through a classifier to obtain a classification result, which is used to indicate whether there is a fault in the conveying coordination between the first to fourth belt conveyors.

[0074] In the aforementioned equipment collaborative monitoring system 100 of the bucket wheel continuous conveyor system, the working status data acquisition unit 110 is used to acquire the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period. As mentioned in the background section, equipment collaborative monitoring is particularly important in the operation of the bucket wheel continuous conveyor system, as it not only affects the normal operation of the system but also its working efficiency. However, the coal mine bucket wheel continuous conveyor system contains multiple pieces of equipment, and the collaborative mode between these pieces of equipment changes with the working conditions. Therefore, equipment collaborative monitoring in the bucket wheel continuous conveyor system is technically challenging. Thus, a system for monitoring equipment collaboratively in a bucket wheel continuous conveyor system is desired.

[0075] In the technical solution of this application, the applicant notes that the entire bucket wheel continuous system comprises three stages: coal mining, coal conveying, and coal unloading. Coal conveying is mainly achieved by belt conveyors L1, L2, L3, and L4, which form a highly coupled conveyor chain. Therefore, in the technical solution of this application, the applicant attempts to perform equipment collaborative monitoring on the conveyor chain portion of the bucket wheel continuous system to reduce monitoring difficulty and ensure the coal conveying process. For ease of explanation, in the technical solution of this application, belt conveyors L1, L2, L3, and L4 are defined as the first to fourth belt conveyors. Then, in the equipment collaborative monitoring, the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period are first obtained. The power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period can be collected by power sensors. Since the first to fourth belt conveyors are large in size and come in various forms, the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period are respectively collected by the first to fourth power sensors.

[0076] In the equipment collaborative monitoring system 100 of the aforementioned bucket wheel continuous system, the working state feature extraction unit 120 is used to arrange the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into power input vectors according to the time dimension, and then use a multi-scale neighborhood extraction module to obtain the first to fourth power feature vectors. That is, the working state of the first to fourth belt conveyors is represented by the time-series vectors of the power values ​​of the first to fourth belt conveyors, and the working feature vectors of the first to fourth belt conveyors are extracted by a multi-scale feature extraction module containing multiple parallel one-dimensional convolutional layers.

[0077] Here, the multi-scale neighborhood feature extraction module uses one-dimensional convolution kernels of different scales to perform multi-scale one-dimensional convolution encoding on the time-series vector of the power value during encoding in order to extract the distribution features of the power value within different time spans, that is, the working state features of the first to fourth belt conveyors within different time spans.

[0078] Figure 4 This is a block diagram of the working status feature extraction unit in the equipment collaborative monitoring system of the bucket wheel continuous system according to an embodiment of this application. Figure 4As shown, the working state feature extraction unit 120 includes: a first scale feature extraction subunit 121, used to input the power input vector of the first belt conveyor into the first convolutional layer of the multi-scale neighborhood feature extraction module to obtain a first scale power vector, wherein the first convolutional layer has a first one-dimensional convolutional kernel of a first length; a second scale feature extraction subunit 122, used to input the power input vector of the first belt conveyor into the second convolutional layer of the multi-scale neighborhood feature extraction module to obtain a second scale power feature vector, wherein the second convolutional layer has a second one-dimensional convolutional kernel of a second length, and the first length is different from the second length; and a multi-scale cascading subunit 123, used to cascade the first scale power vector and the second scale power vector to obtain the first power feature vector.

[0079] More specifically, in this embodiment, the first scale feature extraction subunit 121 is further configured to: use the first convolutional layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolutional encoding on the power input vector of the first belt conveyor to obtain the first scale power feature vector using the following formula; wherein, the formula is:

[0080]

[0081] Where a is the width of the first convolution kernel in the x direction, F(a) is the parameter vector of the first convolution kernel, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the power input vector of the first belt conveyor.

[0082] More specifically, in this embodiment, the second-scale feature extraction subunit 122 is further configured to: use the second convolutional layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolutional encoding on the power input vector of the first belt conveyor to obtain the second-scale power feature vector using the following formula; wherein, the formula is:

[0083]

[0084] Where b is the width of the second convolution kernel in the x direction, F(b) is the parameter vector of the second convolution kernel, G(xb) is the local vector matrix of the operation with the convolution kernel function, m is the size of the second convolution kernel, and X represents the power input vector of the first belt conveyor.

[0085] In the equipment collaborative monitoring system 100 of the aforementioned bucket wheel continuous system, the globalization unit 130 is used to arrange the first to fourth power feature vectors in a two-dimensional arrangement to obtain a global power feature matrix. That is, the first to fourth power feature vectors are structurally integrated in a high-dimensional feature space to obtain the global power feature matrix.

[0086] In the aforementioned bucket wheel continuous system equipment collaborative monitoring system 100, the transfer unit 140 is used to calculate the transfer matrix between every two power feature vectors in the first to fourth power feature vectors to obtain multiple transfer matrices. In the technical solution of this application, the first to fourth belt conveyors exhibit specific coal transport rules during coal transport, which can be represented based on the transfer probabilities between the working state characteristics of the first to fourth belt conveyors. Furthermore, integrating the above information into equipment collaborative monitoring obviously helps improve the accuracy of equipment collaborative monitoring. Specifically, in the technical solution of this application, the transfer matrix between every two power feature vectors in the first to fourth power feature vectors is calculated to obtain multiple transfer matrices. That is, in the technical solution of this application, the transfer matrix between every two power feature vectors represents the coal transfer information between the two belt conveyors.

[0087] Specifically, in this embodiment, the transfer unit 140 is further configured to: calculate the transfer matrix between every two power feature vectors in the first to fourth power feature vectors using the following formula to obtain multiple transfer matrices; wherein, the formula is:

[0088]

[0089] Where V a V represents the first of every two power eigenvectors in the first to fourth power eigenvectors. b This represents the second power eigenvector among every two power eigenvectors in the first to fourth power eigenvectors, and M represents one of the multiple transition matrices. This represents matrix multiplication.

[0090] In the equipment collaborative monitoring system 100 of the aforementioned bucket wheel continuous system, the transfer topology construction unit 150 is used to calculate the global mean of each of the multiple transfer matrices to obtain multiple transfer feature values, and then arrange the multiple transfer feature values ​​in a two-dimensional manner to obtain a transfer topology matrix. That is, the global mean of each transfer matrix represents the coal transfer probability between two belt conveyors, and the obtained coal transfer probabilities are arranged in a two-dimensional manner to obtain the transfer topology matrix. Here, the transfer topology matrix integrates the coal transfer information between each pair of belt conveyors contained in the multiple transfer feature values.

[0091] In the equipment collaborative monitoring system 100 of the aforementioned bucket wheel continuous system, the transfer topology feature extraction unit 160 is used to obtain a transfer topology feature matrix by passing the transfer topology matrix through a convolutional neural network model as a feature extractor. Those skilled in the art will know that convolutional neural network models have excellent performance in extracting locally correlated features. Therefore, using the convolutional neural network model as a feature extractor to process the transfer topology matrix can extract the correlation information between the coal transfer characteristics between belt conveyors.

[0092] Specifically, in this embodiment, the transfer topology feature extraction unit 160 includes: each layer of the convolutional neural network model performing the following during the forward propagation of the layer: convolving the input data to obtain a convolutional feature map; performing mean pooling based on the local feature matrix on the convolutional feature map to obtain a pooled feature map; and performing nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the convolutional neural network model is the transfer topology feature matrix, and the input of the first layer of the convolutional neural network model is the transfer topology matrix.

[0093] In the equipment collaborative monitoring system 100 of the aforementioned bucket wheel continuous system, the graph neural network encoding unit 170 is used to obtain a transfer topology global power feature matrix by passing the global power feature matrix and the transfer topology feature matrix through a graph neural network model. That is, the first to fourth power feature vectors are used as the node feature representations of each belt conveyor, and the transfer topology feature matrix is ​​used as the feature representation of the edges between nodes. A graph neural network model is used for encoding to extract a transfer topology global power feature matrix containing irregular transfer topology information and high-dimensional working state features of each belt conveyor. In other words, the global power feature matrix and the transfer topology feature matrix are passed through a graph neural network model to obtain the transfer topology global power feature matrix.

[0094] In the aforementioned equipment collaborative monitoring system 100 of the continuous bucket wheel system, the equipment collaborative monitoring result generation unit 180 is used to classify the global power feature matrix of the transfer topology through a classifier to obtain a classification result. This classification result indicates whether a fault exists in the collaborative conveying between the first and fourth belt conveyors. Specifically, the classifier performs class boundary partitioning and determination on the high-dimensional data manifold of the global power feature matrix of the transfer topology to obtain a classification result indicating whether a fault exists in the collaborative conveying between the first and fourth belt conveyors. Thus, based on the operating state characteristics and transmission information of each belt conveyor, an equipment collaborative monitoring model of the transmission chain composed of the first to fourth belt conveyors is constructed to intelligently determine whether a fault exists in the collaborative conveying between the first and fourth belt conveyors.

[0095] Specifically, in this embodiment, the device collaborative monitoring result generation unit 180 is further configured to: process the global power feature matrix of the transfer topology using the classifier according to the following formula to obtain the classification result; wherein, the formula is:

[0096] softmax{(W n B n ):…:(W1,B1)|Project(M)}

[0097] Where Project(M) represents projecting the global power feature matrix of the transfer topology into a vector, W1 to W... n Here are the weight matrices for each fully connected layer, B1 to B... n This represents the bias vector of each fully connected layer.

[0098] The equipment collaborative monitoring system 100 in the aforementioned bucket wheel continuous system also includes a training module 200 for training the multi-scale neighborhood extraction module, the convolutional neural network model that serves as a feature extractor, and the classifier.

[0099] Figure 5 This is a block diagram of the training module in the equipment collaborative monitoring system of a continuous bucket wheel system according to an embodiment of this application. Figure 5As shown, the training module 200 includes: a training data acquisition unit 210, used to acquire training power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period, and the actual value of whether there is a fault in the conveying coordination between the first to fourth belt conveyors; a training working state feature extraction unit 220, used to arrange the training power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into training power input vectors according to the time dimension, and then use a multi-scale neighborhood extraction module to obtain the first to fourth training power feature vectors; a training globalization unit 230, used to arrange the first to fourth training power feature vectors in two dimensions to obtain a training global power feature matrix; a training transfer unit 240, used to calculate the training transfer matrix between every two training power feature vectors in the first to fourth training power feature vectors to obtain multiple training transfer matrices; and a training transfer topology construction unit 250, used to calculate the global mean of each of the multiple training transfer matrices to obtain multiple training transfer features. The training transfer feature values ​​are arranged in a two-dimensional manner to obtain a training transfer topology matrix; a training transfer topology feature extraction unit 260 is used to pass the training transfer topology matrix through the convolutional neural network model, which serves as a feature extractor, to obtain a training transfer topology feature matrix; a training graph neural network encoding unit 270 is used to pass the training global power feature matrix and the training transfer topology feature matrix through the graph neural network model to obtain a training transfer topology global power feature matrix; a classification loss unit 280 is used to pass the transfer topology global power feature matrix through the classifier to obtain a classification loss function value; a multi-distribution binary classification quality loss unit 290 is used to calculate the multi-distribution binary classification quality loss function value of the row vectors of the transfer topology global power feature matrix; and a training unit 300 is used to train the multi-scale neighborhood extraction module, the convolutional neural network model, and the classifier using the weighted sum of the multi-distribution binary classification quality loss function value and the classification loss function value as the loss function value.

[0100] Specifically, in the technical solution of this application, when obtaining the global power feature matrix and the transfer topology feature matrix through a graph neural network model, the dimension of the transfer topology feature matrix is ​​l×l, where l is the length of the power feature vector. However, to obtain the global power feature matrix by arranging the first to fourth power feature vectors in two dimensions, it is necessary to perform low-cost augmentation on the power feature vectors in the sample dimension. This makes the correlation between each row vector of the global power feature matrix (assuming the first to fourth power feature vectors are row vectors) worse. Correspondingly, the correlation between each row vector of the transfer topology global power feature matrix also worsens, affecting the overall feature representation capability of the transfer topology global power feature matrix.

[0101] Therefore, it is necessary to assign weights to each row vector of the global power feature matrix of the transition topology to solve this problem. However, since the weights need to be obtained through training as hyperparameters, this will increase the training burden of the model. Therefore, the applicant of this application introduces a multi-distribution binary classification quality loss function for the row vectors of the global power feature matrix of the transition topology.

[0102] Specifically, in this embodiment, the multi-distribution binary classification quality loss unit 290 is further configured to: calculate the multi-distribution binary classification quality loss function value of the row vectors of the global power feature matrix of the transfer topology using the following formula; wherein, the formula is:

[0103]

[0104] Among them, V1 to V n V represents the row vectors of the global power characteristic matrix of the transfer topology. r It is a reference vector, preferably set as the mean eigenvector of all row vectors of the global power characteristic matrix of the transfer topology, and The classification result of the feature vector is represented by ||·||1, which represents the 1-norm of the vector, and log represents the logarithmic function operation with base 2.

[0105] Here, to avoid difficulties in convergence to the target class domain due to excessive fragmentation of the decision boundaries corresponding to the local feature distributions of each row vector in the global power feature matrix of the transfer topology during multi-distribution classification, the continuity quality factor of the local feature distribution of each row vector relative to the global feature distribution of the global power feature matrix of the transfer topology can be predicted as a loss function by averaging the global offset class probability information of the local feature distribution of each row vector based on the binary classification of the predetermined label. By training the model in this way, the optimization of hyperparameters during training can be transformed from backpropagation into a classification problem based on multi-binary classification, and the overall feature representation capability of the global power feature matrix of the transfer topology can be improved. This improves the accuracy of classifying whether there is a fault in the conveyor coordination between the first to fourth belt conveyors.

[0106] In summary, the equipment coordination monitoring system 100 in the continuous bucket wheel system according to the embodiments of this application is explained. It constructs an equipment coordination monitoring model of the transmission chain composed of the first to fourth belt conveyors based on the working state characteristics of the first to fourth belt conveyors and the transmission information of each belt conveyor, so as to intelligently determine whether there is a fault in the coordination of the conveying between the first to fourth belt conveyors. This ensures that no fault occurs in the highly coupled transmission chain composed of the first to fourth belt conveyors during the operation of the continuous bucket wheel system.

[0107] As described above, the equipment collaborative monitoring system 100 in the bucket wheel continuous system according to the embodiments of this application can be implemented in various terminal devices, such as servers with equipment collaborative monitoring functions in the bucket wheel continuous system. In one example, the equipment collaborative monitoring system 100 in the bucket wheel continuous system according to the embodiments of this application can be integrated into the terminal device as a software module and / or a hardware module. For example, the equipment collaborative monitoring system 100 in the bucket wheel continuous system can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the equipment collaborative monitoring system 100 in the bucket wheel continuous system can also be one of many hardware modules of the terminal device.

[0108] Alternatively, in another example, the equipment collaborative monitoring system 100 and the terminal device in the continuous bucket wheel system can also be separate devices, and the equipment collaborative monitoring system 100 in the continuous bucket wheel system can be connected to the terminal device via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0109] Exemplary methods

[0110] Figure 6 This is a flowchart of a collaborative monitoring method for equipment in a continuous bucket wheel system according to an embodiment of this application. Figure 6 As shown, the equipment collaborative monitoring method in the continuous bucket wheel system according to the embodiment of this application includes the following steps: S110, acquiring the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period; S120, arranging the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into power input vectors according to the time dimension, and then obtaining the first to fourth power feature vectors through a multi-scale neighborhood extraction module; S130, arranging the first to fourth power feature vectors in a two-dimensional manner to obtain a global power feature matrix; S140, calculating the transition matrix between every two power feature vectors in the first to fourth power feature vectors to obtain multiple transition matrices. S150, calculate the global mean of each of the plurality of transfer matrices to obtain a plurality of transfer feature values, and arrange the plurality of transfer feature values ​​in a two-dimensional manner to obtain a transfer topology matrix; S160, pass the transfer topology matrix through a convolutional neural network model as a feature extractor to obtain a transfer topology feature matrix; S170, pass the global power feature matrix and the transfer topology feature matrix through a graph neural network model to obtain a transfer topology global power feature matrix; and S180, pass the transfer topology global power feature matrix through a classifier to obtain a classification result, the classification result being used to indicate whether there is a fault in the conveying coordination between the first to fourth belt conveyors.

[0111] In one example, the equipment collaborative monitoring method in the above-mentioned bucket wheel continuous system further includes training the multi-scale neighborhood extraction module, the convolutional neural network model as a feature extractor, and the classifier.

[0112] Figure 7 This is a flowchart illustrating the training of the multi-scale neighborhood extraction module, the convolutional neural network model as a feature extractor, and the classifier in the equipment collaborative monitoring method of a continuous bucket wheel system according to an embodiment of this application. Figure 7As shown, the process of training the multi-scale neighborhood extraction module, the convolutional neural network model as the feature extractor, and the classifier includes: S210, obtaining the training power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period, and the true value of whether there is a fault in the conveying coordination between the first to fourth belt conveyors; S220, arranging the training power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into training power input vectors according to the time dimension, and then passing them through the multi-scale neighborhood extraction module to obtain the first to fourth training power feature vectors; S230, arranging the first to fourth training power feature vectors in a two-dimensional arrangement to obtain a training global power feature matrix; S240, calculating the training transition matrix between every two training power feature vectors in the first to fourth training power feature vectors to obtain multiple training transition matrices; S250, calculating each training power feature vector in the multiple training transition matrices. The global mean of the transition matrix is ​​used to obtain multiple training transition feature values, and the multiple training transition feature values ​​are arranged in two dimensions to obtain a training transition topology matrix; S260, the training transition topology matrix is ​​passed through the convolutional neural network model as a feature extractor to obtain a training transition topology feature matrix; S270, the training global power feature matrix and the training transition topology feature matrix are passed through the graph neural network model to obtain a training transition topology global power feature matrix; S280, the transition topology global power feature matrix is ​​passed through the classifier to obtain a classification loss function value; S290, the multi-distribution binary classification quality loss function value of the row vectors of the transition topology global power feature matrix is ​​calculated; and S300, the multi-scale neighborhood extraction module, the convolutional neural network model as a feature extractor, and the classifier are trained using the weighted sum of the multi-distribution binary classification quality loss function value and the classification loss function value as the loss function value.

[0113] In summary, the equipment coordination monitoring method in the continuous bucket wheel system according to the embodiments of this application is explained. It constructs an equipment coordination monitoring model of the transmission chain composed of the first to fourth belt conveyors based on the working state characteristics of the first to fourth belt conveyors and the transmission information of each belt conveyor, so as to intelligently determine whether there is a fault in the transmission coordination between the first to fourth belt conveyors. This ensures that the highly coupled transmission chain composed of the first to fourth belt conveyors does not experience faults during the operation of the continuous bucket wheel system.

Claims

1. A collaborative monitoring system for equipment in a continuous bucket wheel system, characterized in that, include: The working status data acquisition unit is used to acquire the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period. The working state feature extraction unit is used to arrange the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into power input vectors according to the time dimension, and then pass them through the multi-scale neighborhood feature extraction module to obtain the first to fourth power feature vectors. A globalization unit is used to arrange the first to fourth power feature vectors in a two-dimensional arrangement to obtain a global power feature matrix. The transfer unit is used to calculate the transfer matrix between every two power feature vectors in the first to fourth power feature vectors to obtain multiple transfer matrices; A transfer topology construction unit is used to calculate the global mean of each of the multiple transfer matrices to obtain multiple transfer feature values, and to arrange the multiple transfer feature values ​​in a two-dimensional manner to obtain a transfer topology matrix; A transfer topology feature extraction unit is used to pass the transfer topology matrix through a convolutional neural network model as a feature extractor to obtain a transfer topology feature matrix. A graph neural network encoding unit is used to pass the global power feature matrix and the transfer topology feature matrix through a graph neural network model to obtain a transfer topology global power feature matrix; as well as The equipment collaborative monitoring result generation unit is used to pass the global power feature matrix of the transfer topology through a classifier to obtain a classification result, which is used to indicate whether there is a fault in the conveying collaboration between the first to the fourth belt conveyors.

2. The equipment collaborative monitoring system in the continuous bucket wheel system according to claim 1, characterized in that, The working state feature extraction unit includes: The first scale feature extraction subunit is used to input the power input vector of the first belt conveyor into the first convolutional layer of the multi-scale neighborhood feature extraction module to obtain the first scale power feature vector, wherein the first convolutional layer has a first one-dimensional convolutional kernel of a first length; The second-scale feature extraction subunit is used to input the power input vector of the first belt conveyor into the second convolutional layer of the multi-scale neighborhood feature extraction module to obtain a second-scale power feature vector, wherein the second convolutional layer has a second one-dimensional convolutional kernel of a second length, and the first length is different from the second length; and A multi-scale cascaded subunit is used to cascade the first-scale power feature vector and the second-scale power feature vector to obtain a first power feature vector.

3. The equipment collaborative monitoring system in the continuous bucket wheel system according to claim 2, characterized in that, The first scale feature extraction subunit is further used for: The first convolutional layer of the multi-scale neighborhood feature extraction module performs one-dimensional convolutional encoding on the power input vector of the first belt conveyor using the following formula to obtain the first-scale power feature vector. The formula is as follows: Where a is the width of the first convolution kernel in the x direction, F(a) is the parameter vector of the first convolution kernel, G(xa) is the local vector matrix operated with the convolution kernel function, w is the size of the first convolution kernel, and X represents the power input vector of the first belt conveyor.

4. The equipment collaborative monitoring system in the continuous bucket wheel system according to claim 3, characterized in that, The second scale feature extraction subunit is further used for: The second convolutional layer of the multi-scale neighborhood feature extraction module performs one-dimensional convolutional encoding on the power input vector of the first belt conveyor using the following formula to obtain the second-scale power feature vector; The formula is as follows: Where b is the width of the second convolution kernel in the x direction, F(b) is the parameter vector of the second convolution kernel, G(xb) is the local vector matrix of the operation with the convolution kernel function, m is the size of the second convolution kernel, and X represents the power input vector of the first belt conveyor.

5. The equipment collaborative monitoring system in the continuous bucket wheel system according to claim 4, characterized in that, The transfer unit is further configured to: The transition matrix between every two power eigenvectors in the first to fourth power eigenvectors is calculated using the following formula to obtain multiple transition matrices; The formula is as follows: Where V a V represents the first of every two power eigenvectors in the first to fourth power eigenvectors. b This represents the second power eigenvector among every two power eigenvectors in the first to fourth power eigenvectors, and M represents one of the multiple transition matrices. This represents matrix multiplication.

6. The equipment collaborative monitoring system in the bucket wheel continuous system according to claim 5, characterized in that, The transfer topology feature extraction unit includes: Each layer of the convolutional neural network model is processed during the forward propagation of the layer as follows: The input data is processed by convolution to obtain a convolutional feature map; The convolutional feature map is subjected to mean pooling based on the local feature matrix to obtain a pooled feature map; and The pooled feature map is nonlinearly activated to obtain an activated feature map; The output of the last layer of the convolutional neural network model is the transition topology feature matrix, and the input of the first layer of the convolutional neural network model is the transition topology matrix.

7. The equipment collaborative monitoring system in the continuous bucket wheel system according to claim 6, characterized in that, The device collaborative monitoring result generation unit is further used for: The classifier is used to process the global power feature matrix of the transfer topology using the following formula to obtain the classification result; The formula is: softmax{(W n B n ):…:(W1,B1)|Project(M)}, where Project represents the projection of the global power feature matrix of the transfer topology into a vector, W1 to W n Here are the weight matrices for each fully connected layer, B1 to b. n This represents the bias vector of each fully connected layer.

8. The equipment collaborative monitoring system in the bucket wheel continuous system according to claim 7, characterized in that, It also includes a training module for training the multi-scale neighborhood feature extraction module, the convolutional neural network model that serves as the feature extractor, and the classifier; The training module includes: The training data acquisition unit is used to acquire the training power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period, as well as the actual values ​​of whether there is a fault in the conveying coordination between the first to fourth belt conveyors. The training working state feature extraction unit is used to arrange the training power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period into training power input vectors according to the time dimension, and then pass them through the multi-scale neighborhood feature extraction module to obtain the first to fourth training power feature vectors. A globalization unit is trained to arrange the first to fourth training power feature vectors in two dimensions to obtain a training global power feature matrix. The training transition unit is used to calculate the training transition matrix between every two training power feature vectors in the first to fourth training power feature vectors to obtain multiple training transition matrices. The training transition topology construction unit is used to calculate the global mean of each of the plurality of training transition matrices to obtain a plurality of training transition feature values, and to arrange the plurality of training transition feature values ​​in two dimensions to obtain a training transition topology matrix. A training transfer topology feature extraction unit is used to pass the training transfer topology matrix through the convolutional neural network model, which acts as a feature extractor, to obtain a training transfer topology feature matrix. A training graph neural network encoding unit is used to pass the training global power feature matrix and the training transition topology feature matrix through the graph neural network model to obtain the training transition topology global power feature matrix; A classification loss unit is used to pass the global power feature matrix of the transition topology through the classifier to obtain a classification loss function value; A multi-distribution binary classification quality loss unit is used to calculate the multi-distribution binary classification quality loss function value of the row vectors of the global power feature matrix of the transfer topology; and The training unit is used to train the multi-scale neighborhood feature extraction module, the convolutional neural network model that serves as the feature extractor, and the classifier using a weighted sum of the multi-distribution binary classification quality loss function value and the classification loss function value as the loss function value.

9. The equipment collaborative monitoring system in the continuous bucket wheel system according to claim 8, characterized in that, The multi-distribution binary classification quality loss unit is further used for: The multi-distribution binary classification quality loss function value of the row vectors of the global power feature matrix of the transfer topology is calculated using the following formula; The formula is as follows: Among them, V1 to V n V represents the row vectors of the global power characteristic matrix of the transfer topology. r It is a reference vector, and This represents the classification result of the feature vector, ||·||1 represents the 1-norm of the vector, and log represents the logarithmic function operation with base 2.

10. A method for coordinated monitoring of equipment in a continuous bucket wheel system, characterized in that, include: Obtain the power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period; The power values ​​of the first to fourth belt conveyors at multiple predetermined time points within a predetermined time period are arranged according to the time dimension to form a power input vector, and then the first to fourth power feature vectors are obtained by the multi-scale neighborhood feature extraction module. The first to fourth power feature vectors are arranged in a two-dimensional arrangement to obtain the global power feature matrix; Calculate the transition matrix between every two power eigenvectors in the first to fourth power eigenvectors to obtain multiple transition matrices; Calculate the global mean of each of the multiple transition matrices to obtain multiple transition feature values, and arrange the multiple transition feature values ​​in a two-dimensional arrangement to obtain a transition topology matrix; The transition topology matrix is ​​passed through a convolutional neural network model as a feature extractor to obtain a transition topology feature matrix; The global power feature matrix and the transfer topology feature matrix are used through a graph neural network model to obtain the transfer topology global power feature matrix; as well as The global power feature matrix of the transfer topology is passed through a classifier to obtain a classification result, which is used to indicate whether there is a fault in the conveying coordination between the first to fourth belt conveyors.

Images