Mine equipment maintenance task priority intelligent control method and system

Abstract

The invention relates to the technical field of mining equipment management and maintenance scheduling, in particular to a mining equipment maintenance task priority intelligent control method and system. Comprising the steps of performing data preprocessing on acquired mining equipment operation parameters; generating graph structure data based on the preprocessed mining equipment operation parameters; performing feature representation on the graph structure data by using a multilayer graph convolutional neural network, and outputting a health score; calculating a resource constraint score according to the obtained resource data; calculating a maintenance task priority based on the health score and the resource constraint score; carrying out descending sorting on the maintenance tasks according to the priority; and carrying out dynamic planning optimization based on a multi-stage decision, and solving to obtain a global optimal scheduling scheme. According to the method, GNN is introduced into mining equipment health state evaluation for the first time, and the evaluation precision is remarkably improved through equipment associated information modeling and node embedding representation learning.

Description

Technical Field

[0001] The present invention relates to the technical field of mine equipment management and maintenance scheduling, and in particular, to an intelligent control method and system for the priority of mine equipment maintenance tasks. Background Art

[0002] The reliable operation of mine equipment is the core guarantee for the efficient production of mining enterprises. However, the traditional maintenance task scheduling mode has the following significant problems: Response lag: Traditional scheduling is based on static priorities or regular plans, lacking dynamic adaptability and unable to respond promptly to sudden failures, resulting in an increase in equipment downtime.

[0003] Ignoring resource constraints: The dynamic supply of maintenance resources (such as personnel, spare parts, tools, etc.) is not fully considered, resulting in the lack of practical feasibility of the task scheduling plan.

[0004] Single state assessment: Only relying on a single operating parameter of the equipment (such as vibration or temperature) for health assessment makes it difficult to comprehensively reflect the operating state and potential risks of the equipment.

[0005] Moreover, there are also many deficiencies in the existing technologies: Lack of correlation analysis: Existing methods usually do not consider the correlation between equipment. For example, the failure of some equipment may lead to a decline in the performance of its associated equipment.

[0006] Insufficient resource dynamics: The real-time supply and demand of resources are not effectively modeled, resulting in the disconnection between the scheduling results and the actual situation.

[0007] Limited optimization ability: Existing scheduling optimization methods mostly use simple sorting or dynamic adjustment based on a single parameter, and fail to achieve global optimization under complex multi-constraint conditions.

[0008] In view of the above problems, there is an urgent need to propose an intelligent priority adjustment method based on the comprehensive assessment of equipment health status and resource dynamic optimization. By introducing the Graph Neural Network (GNN) and dynamic programming algorithms, the limitations of traditional methods can be solved, and the intelligence and executability of maintenance task scheduling can be improved. Summary of the Invention

[0009] In order to solve the above-mentioned problems, the present invention provides an intelligent control method and system for the priority of mine equipment maintenance tasks, which adopts intelligent health status assessment and resource dynamic optimization, and combines multi-task scheduling optimization algorithms to solve problems such as insufficient dynamics and inadequate combination of multi-dimensional factors in traditional methods.

[0010] In the first aspect, an intelligent control method for the priority of mine equipment maintenance tasks provided by the present invention adopts the following technical solutions: An intelligent control method for the priority of mine equipment maintenance tasks, comprising: Obtain the operation parameters and resource data of mine equipment; Perform data preprocessing on the obtained operation parameters of mine equipment; Generate graph structure data based on the preprocessed operation parameters of mine equipment; Use a multi-layer graph convolutional neural network to perform feature representation on the graph structure data and output a health score; Calculate the resource constraint score according to the obtained resource data; Calculate the maintenance task priority based on the health score and the resource constraint score; Sort the maintenance tasks in descending order according to the priority; Perform dynamic programming optimization based on multi-stage decision-making to solve for the global optimal scheduling plan.

[0011] Update the model weights and resource constraint coefficients of the GNN model in real time based on the online learning algorithm.

[0012] Further, the data preprocessing of the obtained operation parameters of mine equipment includes noise filtering, normalization, and feature enhancement processing. Among them, the noise filtering includes using the interquartile range method for outlier detection and then smoothing the detected outliers. The interquartile range method is expressed as: = 25th percentile, = 75th percentile Outlier: or Parameter description: : First quartile (25% of the data points are less than this value); : Third quartile (75% of the data points are less than this value); IQR: Interquartile range, measuring the dispersion of data; X: Data point, if it exceeds the range, it is determined as an outlier.

[0013] Further, the generation of graph structure data based on the preprocessed operation parameters of mine equipment includes constructing an association network between devices, quantifying the correlation of nodes using mutual information and load transfer relationships, and generating graph structure data.

[0014] Further, the use of a multi-layer graph convolutional neural network to perform feature representation on the graph structure data and output a health score includes using a multi-layer graph convolutional neural network GCN to capture device features and their correlations, and outputting a comprehensive health score through node embedding representation learning , and the forward propagation process of GCN is as follows: Wherein: represents the node feature matrix of the th layer; represents the activation function, represents the adjacency matrix of the graph; represents the degree matrix of the nodes; represents the symmetrically normalized adjacency matrix, which is used to smooth the node features; represents the node feature matrix of the th layer; represents the weight matrix of the th layer, which is used for linear transformation.

[0015] Furthermore, calculating the resource constraint score according to the obtained resource data includes calculating the resource constraint coefficient according to the real-time collected resource data , integrating multiple resource constraint coefficients to generate a comprehensive resource constraint score , which reflects the degree of resource supply limitation, and the calculation formula is: Wherein, is the weight coefficient of the th type of resource, is the constraint coefficient of the th type of resource.

[0016] Furthermore, calculating the maintenance task priority based on the health score and the resource constraint score includes calculating the health score and the resource constraint score respectively, and calculating the priority in a weighted manner according to the health score and the resource constraint score. The priority formula is expressed as:

[0017] Wherein, and are dynamically adjustable weight parameters, which balance the influence of the device health status and the resource limitation, and The initial values of are obtained through training with historical data and are adaptively adjusted according to the feedback data during the scheduling process.

[0018] Furthermore, sorting the maintenance tasks in descending order according to the priority includes sorting the maintenance tasks according to the priority. Among them, if the priorities of multiple tasks are the same, then a secondary sorting is performed according to the severity of the failure and the influence range factors; if they still cannot be distinguished, then they are sorted according to the task creation time, that is, first come, first served.

[0019] Further, the dynamic programming optimization based on multi-stage decision-making is used to solve for the global optimal scheduling scheme, including modeling the multi-stage decision-making process, defining the state transition equation and the optimization objective function, and solving for the global optimal scheduling scheme. Among them, the state transition equation is expressed as: ; The optimization objective function is expressed as: ; Among them: represents the minimum cost when the th task is completed and completed at time ; represents the task executed in the time period ; represents the task executed in the time period ; represents the th task's weight coefficient; represents the th task's execution time; represents the th type of resource's weight coefficient; is the usage amount of the th type of resource.

[0020] Further, the model weights and resource constraint coefficients of the GNN model are updated in real time based on the online learning algorithm, including using the online learning algorithm to dynamically adjust the model parameters according to the new task execution data, and using the reinforcement learning algorithm to dynamically optimize the and values according to the feedback rewards of task completion quality and resource utilization efficiency.

[0021] Second, a smart control system for the priority of mine equipment maintenance tasks includes: A data acquisition module configured to acquire the operating parameters and resource data of mine equipment; A preprocessing module configured to perform data preprocessing on the acquired operating parameters of mine equipment; A graph structure module configured to generate graph structure data based on the preprocessed operating parameters of mine equipment; A feature representation module configured to perform feature representation on the graph structure data using a multi-layer graph convolutional neural network and output a health score; A constraint module configured to calculate a resource constraint score based on the acquired resource data; A priority module, configured to calculate the maintenance task priority based on the health score and the resource constraint score; A sorting module, configured to sort the maintenance tasks in descending order according to the priority; An optimization module, configured to perform dynamic programming optimization based on multi-stage decision-making, and solve to obtain a globally optimal scheduling scheme.

[0022] In a third aspect, the present invention provides a computer-readable storage medium, in which multiple instructions are stored, and the instructions are adapted to be loaded and executed by a processor of a terminal device for the intelligent control method for the maintenance task priority of mining equipment.

[0023] In a fourth aspect, the present invention provides a terminal device, including a processor and a computer-readable storage medium. The processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are adapted to be loaded and executed by the processor for the intelligent control method for the maintenance task priority of mining equipment.

[0024] In summary, the present invention has the following beneficial technical effects: The present invention first introduces GNN into the health state assessment of mining equipment. Through device association information modeling and node embedding representation learning, the assessment accuracy is significantly improved. Compared with the traditional single-index assessment method, the GNN model of the present invention has increased the accuracy rate in the health state assessment of equipment by more than 20%.

[0025] An innovative joint framework of resource dynamic constraint modeling and multi-task optimization is proposed, with resource dynamic constraints as the core, and combined with the ant colony optimization algorithm to solve resource conflict problems in real time. Compared with the traditional static resource allocation method, the dynamic resource optimization method of the present invention has increased the executable rate of task scheduling by more than 30%.

[0026] An adaptive feedback mechanism is introduced, and the dynamic optimization of model parameters is realized through online learning and reinforcement learning algorithms, significantly improving the real-time performance and adaptability of the scheduling system. Compared with the traditional fixed-parameter model, the adaptive feedback mechanism of the present invention has improved the scheduling effect by more than 15%.

[0027] An innovative application of the dynamic programming algorithm for multi-task scheduling optimization, through the definition of the state transition equation and the optimization objective function, solves the globally optimal scheduling scheme under multiple constraint conditions. Compared with the traditional heuristic scheduling algorithm, the dynamic programming optimization method of the present invention has shortened the task completion time by more than 10% and increased the resource utilization rate by more than 5%. Description of the Drawings

[0028] Figure 1 It is a schematic diagram of the method of Embodiment 1 of the present invention. Detailed Embodiments

[0029] The present invention will be further described in detail below with reference to the accompanying drawings.

[0030] Embodiment 1 Referring to Figure 1 , an intelligent control method for the priority of mine equipment maintenance tasks in this embodiment includes: Obtaining the operating parameters and resource data of mine equipment; Performing data preprocessing on the obtained operating parameters of mine equipment; Generating graph structure data based on the preprocessed operating parameters of mine equipment; Using a multi-layer graph convolutional neural network GCN model to perform feature representation on the graph structure data and output a health score; Calculating a resource constraint score according to the obtained resource data; Calculating the priority of maintenance tasks based on the health score and the resource constraint score; Sorting the maintenance tasks in descending order according to the priority; Performing dynamic programming optimization based on multi-stage decision-making to solve and obtain a global optimal scheduling plan.

[0031] Updating the model weights and resource constraint coefficients of the GCN model in real time based on an online learning algorithm.

[0032] Specifically, it includes the following steps: Step 1: Comprehensive evaluation model for equipment health status 1.1 Feature extraction: Collecting equipment operating parameters (such as vibration, temperature, current, etc.) and performing noise filtering, normalization, and feature enhancement processing.

[0033] Among them, 1.1.1 Noise filtering: Outlier detection, using the IQR (interquartile range) method: = 25th percentile, = 75th percentile Outlier: or Parameter description: : First quartile (25% of the data points are less than this value); : Third quartile (75% of the data points are less than this value); IQR: Interquartile range, measuring the dispersion of data; X: Data point, if it exceeds the range, it is determined as an outlier.

[0034] Smoothing processing, using the moving average method (applicable to time series data): Parameter Description: : Smoothed data points; : Data points at the current and the previous N time steps; : Window size, determining the degree of smoothing.

[0035] 1.1.2 Normalization: Min - Max Scaling: Parameter Description: : Original data points; : Minimum value in the dataset; : Maximum value in the dataset; : Normalized data, in the range [0, 1].

[0036] Z - score Standardization (Mean - Standard Deviation Normalization): Parameter Description: : Original data points; : Mean of the data; : Standard deviation of the data; : Standardized data, with mean 0 and standard deviation 1.

[0037] Logarithmic Transformation (Applicable to Data with Long - Tail Distribution): Parameter Description: : Original data points; : Transformed data, reducing the gap between large values.

[0038] 1.1.3 Feature Enhancement: Time - Series Feature Extraction: Rate of Trend Change:

[0039] Parameter Description: : Data at the current time point; : Data at the previous time point; : Rate of change, measuring the data trend.

[0040] Interactive Feature Construction: Generate combined features between variables, such as ratios, products, etc.

[0041] Product Feature: Parameter Description: , : Original feature; : Newly constructed interactive feature.

[0042] Dimensionality reduction processing: Use PCA or Autoencoder to extract low-dimensional features and improve computational efficiency.

[0043] PCA (Principal Component Analysis): Parameter description: : The original data matrix; : The feature matrix, only retaining the principal components with the highest contribution rate; : The data after dimensionality reduction, improving computational efficiency.

[0044] 1.2 Association modeling: Construct an association network between devices, quantify the correlation between nodes using mutual information or load transfer relationships, and generate graph-structured data.

[0045] Among them, 1.2.1 Generate graph-structured data, Define nodes: Device nodes (device ID, type, operating status).

[0046] Sensor nodes (temperature, vibration, pressure, etc.).

[0047] Task nodes (repair tasks, historical records).

[0048] Define edges (relationships): Device - Sensor (data acquisition relationship).

[0049] Device - Task (repair impact relationship).

[0050] Device - Device (dependency or impact relationship, such as devices in the same system).

[0051] Construct attributes: Each node carries features (health score, historical failure rate, service life, etc.).

[0052] Each edge is attached with weights (fault correlation, signal strength, etc.).

[0053] 1.2.2 Construct the graph structure, Initialize the graph G=(V,E), where V is the set of nodes and E is the set of edges.

[0054] Traverse the device data and add device, sensor, and task nodes to V.

[0055] According to the device association rules, add edges to E: Connect devices and sensors (based on data flow).

[0056] Connect devices and repair tasks (based on historical repair records).

[0057] Inter-device connection (based on topological structure or influence propagation relationship).

[0058] Calculate the adjacency matrix A or use an edge list to store the graph structure.

[0059] 1.2.3 Generate examples, Suppose there are three devices (E1, E2, E3), two sensors (S1, S2), and one maintenance task (T1): Device E1 is connected to sensor S1 and maintenance task T1.

[0060] Device E2 affects device E3 (dependency between devices).

[0061] Sensor S2 is connected to devices E2 and E3.

[0062] The generated graph structure is as follows: E1—S1 E1—T1 E2—E3 E2—S2—E3 Finally, a graph structure of device-task-sensor association is formed for subsequent analysis and optimization.

[0063] 1.3 Graph neural network modeling: Use a multi-layer graph convolutional neural network (Graph Convolutional Network, GCN) to capture device features and their correlations, and through node embedding representation learning, output a comprehensive health score . The forward propagation process of GCN is as follows: Where: represents the node feature matrix of the layer; represents the activation function, such as ReLU or Sigmoid; represents the adjacency matrix of the graph; represents the degree matrix of the nodes; represents the symmetrically normalized adjacency matrix, used to smooth node features; represents the node feature matrix of the layer; represents the layer weight matrix, used for linear transformation.

[0064] 1.4 Output health score: Use the softmax function to convert the node embedding representation of the last layer into a normalized health score. A value closer to 1 indicates a better health status of the device.

[0065] Step 2: Model dynamic resource constraints; 2.1 Collect resource data in real time (such as personnel skills, spare part inventory, tool status), and calculate the resource constraint coefficient , and the calculation steps are as follows: Define resource constraint parameters: : Human resource availability (number of maintenance personnel / required number).

[0066] : Spare part supply situation (number of available spare parts / required number).

[0067] : Time constraint (available maintenance time / estimated task time).

[0068] : Budget constraint (available budget / task budget requirement).

[0069] Calculate the resource constraint coefficient (combining various resource constraints): where , , , are the weights of each resource, satisfying + + + = 1.

[0070] Normalization processing: If > 1, then set it to 1 (abundant resources).

[0071] If < 0, then set it to 0 (unavailable resources).

[0072] Finally, as an important factor in calculating the priority of maintenance tasks, affects task sorting and scheduling decisions.

[0073] 2.2 Combine multiple resource constraint coefficients to generate a comprehensive resource constraint score , reflecting the degree of resource supply limitation. The calculation formula of is: is the The weight coefficient of the resource, is the constraint coefficient of the

[0074] Step 3: Calculate the priority of the maintenance task; The calculation of the maintenance task priority is based on the health score and the resource constraint score. The core principle is to comprehensively consider the influence of both through the weighted summation method and dynamically adjust the weights according to real-time feedback. The specific steps are as follows: Calculate the health score (HS). Use a multi-layer graph convolutional neural network (GCN) to extract features from the operation data of the device and evaluate the health status, generating a normalized health score (the higher the value, the better the device status).

[0075] Calculate the resource constraint score (RS). Calculate the resource constraint coefficient based on the currently available resources (personnel, spare parts, tools, etc.), and comprehensively consider multiple resource factors to calculate the resource constraint score (the higher the value, the less the resource constraint and the easier the task to execute).

[0076] Calculate the priority of the maintenance task (Priority). Use the weighted formula to calculate the priority: Where: and are the dynamically adjusted weight coefficients, balancing the influence of the device health status and resource limitations; reflects the urgency of the device (the worse the health status, the higher the priority); reflects the resource availability (the less the resource constraint, the higher the priority).

[0077] Sort the tasks. According to the calculated priority sort the maintenance tasks in descending order, and give priority to processing the tasks with the highest priority.

[0078] Dynamically optimize the weights. Combined with historical task data and scheduling feedback, use reinforcement learning or adaptive optimization methods to dynamically adjust and to ensure the optimal scheduling strategy.

[0079] 3.1 Priority formula: 3.2 and : Dynamically adjustable weight parameters, balancing the influence of the device health status and resource limitations. and The initial values can be obtained through training with historical data and adjusted adaptively according to the feedback data during the scheduling process.

[0080] Step 4: Task scheduling and optimization; 4.1 Task sorting: According to the priority Sort the tasks in descending order.

[0081] The descending order of the maintenance tasks is based on the calculated priority , and the specific steps are as follows: Task list acquisition, Obtain all the tasks to be maintained, including task numbers, health scores , resource constraint scores and the calculated priority .

[0082] Task sorting, Arrange in descending order according to the priority , that is, the tasks with higher priority are arranged in the front: Processing tasks with the same priority, If multiple tasks have the same priority, perform a secondary sort according to factors such as the severity of the failure and the affected range; If still unable to distinguish, sort according to the task creation time (first come, first served).

[0083] Task queue generation, Generate a list of the execution order of the maintenance tasks and allocate them to the maintenance team or the scheduling system.

[0084] Real-time adjustment, Combined with the task completion status, newly added tasks or resource changes, dynamically adjust the task sorting to ensure optimal scheduling.

[0085] This method ensures that the most urgent and easiest-to-execute tasks are processed first, improving the maintenance efficiency and resource utilization rate.

[0086] 4.2 Dynamic programming optimization: Adopt a multi-stage decision-making process modeling, define the state transition equation and the optimization objective function, and solve the global optimal scheduling plan.

[0087] State transition equation: Optimization objective function: Where: represents the minimum cost when the th task is completed and completed at time ; Represents the cost of a task executed over a time period ; Represents the time of a task executed over a time period ; Represents the weight coefficient of the th task; Represents the execution time of the th task; Represents the weight coefficient of the th class of resources; is the usage amount of the th class of resources.

[0088] Step 5: Feedback optimization; 5.1 Update the GNN model weights and resource constraint coefficients in real time to improve the accuracy of health status assessment and scheduling. Use an online learning algorithm (such as stochastic gradient descent) to dynamically adjust the model parameters based on new task execution data.

[0089] Example: Dynamically adjust the model parameters using Stochastic Gradient Descent (SGD), Goal: Train a fault prediction model to accurately predict the device health score HS.

[0090] Example steps: Set the loss function( such as Mean Squared Error (MSE).

[0091] Randomly sample a training data point (x, y), where x is the device feature and y is the actual health score.

[0092] Calculate the predicted value .

[0093] Calculate the loss .

[0094] Calculate the gradient , and adjust the parameters: , where is the learning rate.

[0095] Repeat the above steps until the loss converges or reaches the set number of iterations.

[0096] 5.2 Dynamically adjust the weight parameters and , making the model more in line with actual requirements. Using reinforcement learning algorithms (such as Q-learning), dynamically optimize according to the feedback rewards of task completion quality and resource utilization efficiency and values.

[0097] Among them, use Q-learning to dynamically optimize and values Objective: Optimize the learning rate in the maintenance task scheduling (exploration intensity) and discount factor (long-term reward weight).

[0098] Including the following steps: State definition: Set the state s as the current equipment health status and maintenance task queue.

[0099] Action space: Set different maintenance decisions (such as preferentially repairing high-risk equipment or low-cost maintenance).

[0100] Q-table initialization: Initialize the Q-value table Q(s,a) with random values.

[0101] Q-value update (Bellman equation): , where: r is the current reward (such as the improvement of equipment availability after maintenance).

[0102] represents the optimal future reward after executing action .

[0103] Dynamically adjust and : Adopt a larger (strengthen exploration) at the beginning and decrease over time.

[0104] Adjust according to the task urgency. If the future impact is large, increase .

[0105] Policy optimization: When the Q-value converges, adopt a greedy policy to select the optimal maintenance task scheduling plan.

[0106] Example 2 This example provides an intelligent control system for the priority of mine equipment maintenance tasks, including: A data acquisition module, configured to acquire the operating parameters and resource data of mine equipment; A preprocessing module, configured to perform data preprocessing on the acquired operating parameters of mine equipment; A graph structure module, configured to generate graph structure data based on the preprocessed operation parameters of mining equipment; A feature representation module, configured to perform feature representation on the graph structure data by using a multi-layer graph convolutional neural network and output a health score; A constraint module, configured to calculate a resource constraint score according to the obtained resource data; A priority module, configured to calculate the maintenance task priority based on the health score and the resource constraint score; A sorting module, configured to perform a descending order sorting of the maintenance tasks according to the priority; An optimization module, configured to perform dynamic programming optimization based on multi-stage decision-making and solve to obtain a globally optimal scheduling plan.

[0107] A computer-readable storage medium, in which multiple instructions are stored, and the instructions are adapted to be loaded and executed by a processor of a terminal device for the intelligent control method for the priority of mining equipment maintenance tasks described above.

[0108] A terminal device, including a processor and a computer-readable storage medium, where the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are adapted to be loaded and executed by the processor for the intelligent control method for the priority of mining equipment maintenance tasks described above.

[0109] The above are all preferred embodiments of the present invention, and the protection scope of the present invention is not limited thereby. Therefore, all equivalent changes made according to the structure, shape, and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for intelligent control of the priority of mining equipment maintenance tasks, characterized in that: include: Obtain mining equipment operating parameters and resource data; Carry out data preprocessing on the acquired mining equipment operating parameters; Generate graph structure data based on preprocessed mining equipment operating parameters; Use a multi-layer graph convolutional neural network to represent the features of graph structure data and output a health score; Calculating resource constraint scores based on the acquired resource data; Calculate maintenance task priority based on health score and resource constraint score; Sort maintenance tasks in descending order according to priority; Based on multi-stage decision making, dynamic programming optimization is performed to obtain the global optimal scheduling solution.

2. The method for intelligent control of the priority of mining equipment maintenance tasks according to claim 1 is characterized in that: The data preprocessing of the acquired mining equipment operating parameters includes noise filtering, normalization and feature enhancement processing, wherein the noise filtering includes outlier detection using the interquartile range method, and then smoothing the detected outliers, wherein the interquartile range method is expressed as: = 25th percentile, = 75th percentile Outliers: or Parameter Description: : The first quartile (25% of the data points are less than this value); : The third quartile (75% of the data points are less than this value); IQR: Interquartile range, which measures the degree of dispersion of the data; X: Data points, if they are out of range, are judged as outliers.

3. The method for intelligently controlling the priority of mining equipment maintenance tasks according to claim 2, characterized in that: The method of generating graph structure data based on the preprocessed mining equipment operating parameters includes constructing an inter-equipment association network, quantifying the correlation of nodes using mutual information and load transfer relationships, and generating graph structure data.

4. The method for intelligently controlling the priority of mining equipment maintenance tasks according to claim 3 is characterized in that: The method uses a multi-layer graph convolutional neural network to represent the features of the graph structure data and output a health score, including using a multi-layer graph convolutional neural network GCN to capture device features and their relevance, and outputs a comprehensive health score through node embedding representation learning. , the forward propagation process of GCN is as follows: in: Indicates The node feature matrix of the layer; represents the activation function, Represents the adjacency matrix of a graph; represents the degree matrix of the node; represents the symmetric normalized adjacency matrix used to smooth node features; Indicates The node feature matrix of the layer; Indicates The weight matrix of the layer, used for linear transformation.

5. The method for intelligently controlling the priority of mining equipment maintenance tasks according to claim 4, characterized in that: The resource constraint score is calculated based on the acquired resource data, including calculating the resource constraint coefficient based on the real-time collected resource data. , combining multiple resource constraint coefficients to generate a comprehensive resource constraint score , reflecting the resource supply constraints, The calculation formula is: in, For the The weight coefficient of the resource, For the The constraint coefficient of a resource.

6. A method for intelligently controlling the priority of mining equipment maintenance tasks according to claim 5, characterized in that: The maintenance task priority is calculated based on the health score and the resource constraint score, including calculating the health score and the resource constraint score respectively, and calculating the priority in a weighted manner according to the health score and the resource constraint score. The priority formula is expressed as: in, and Dynamically adjustable weight parameters to balance the impact of device health status and resource constraints. and The initial value of is obtained through historical data training and is adaptively adjusted according to feedback data during the scheduling process.

7. The method for intelligently controlling the priority of mining equipment maintenance tasks according to claim 6, characterized in that: The descending sorting of maintenance tasks according to priority includes sorting maintenance tasks according to priority, wherein, if multiple tasks have the same priority, secondary sorting is performed based on fault severity and impact scope factors; if they still cannot be distinguished, they are sorted according to task creation time, that is, first come first served.

8. The method for intelligently controlling the priority of mining equipment maintenance tasks according to claim 7, characterized in that: The dynamic programming optimization based on multi-stage decision-making is performed to solve the global optimal scheduling solution, including adopting multi-stage decision-making process modeling, defining state transfer equations and optimization objective functions, and solving the global optimal scheduling solution, wherein the state transfer equation is expressed as: ; The optimization objective function is expressed as: ; in: Indicated in tasks completed and in time Minimum cost to complete; Indicates the task In time period The cost of execution; Indicates the task In time period The time of execution; Indicates The weight coefficient of each task; Indicates The execution time of each task; Indicates The weight coefficient of the class resource; It is The usage of class resources.

9. The method for intelligently controlling the priority of mining equipment maintenance tasks according to claim 8, characterized in that: The method of updating the model weights and resource constraint coefficients of the GNN model in real time based on the online learning algorithm includes dynamically adjusting the model parameters according to the new task execution data using the online learning algorithm, and dynamically optimizing the feedback rewards based on the task completion quality and resource utilization efficiency using the reinforcement learning algorithm. and The value of .

10. An intelligent control system for priority of maintenance tasks of mining equipment, comprising: The data acquisition module is configured to acquire mining equipment operating parameters and resource data; The preprocessing module is configured to perform data preprocessing on the acquired mining equipment operating parameters; A graph structure module is configured to generate graph structure data based on the preprocessed mining equipment operating parameters; The feature representation module is configured to perform feature representation on the graph structure data using a multi-layer graph convolutional neural network and output a health score; A constraint module is configured to calculate a resource constraint score based on the acquired resource data; A priority module configured to calculate a maintenance task priority based on a health score and a resource constraint score; The sorting module is configured to sort the maintenance tasks in descending order according to the priority; The optimization module is configured to perform dynamic programming optimization based on multi-stage decisions to obtain the global optimal scheduling solution.

Images