Intelligent maintenance decision-making method and system based on knowledge graph

Abstract

The invention relates to an intelligent maintenance decision method and system based on a knowledge graph, and relates to the technical field of intelligent operation and maintenance and decision support and information, and the method comprises the steps: obtaining the operation state data and historical maintenance data of industrial equipment; constructing an industrial knowledge graph and generating an incidence relation model; an optimal maintenance path is obtained based on the analysis of the relationship between the fault type and the maintenance path; optimizing a maintenance task allocation scheme through the optimal maintenance path, the incidence relation model and the maintenance resource constraint condition; carrying out maintenance time window dynamic adjustment analysis to obtain an optimal maintenance strategy; and finally, performing multi-objective optimization iterative analysis to obtain a final maintenance decision. According to the scheme, various factors can be comprehensively considered, and the scientificity and the accuracy of maintenance decision making of the industrial equipment are improved.

Description

Technical Field

[0001] The present application relates to the fields of intelligent operation and maintenance, decision support, and information technology, and in particular to an intelligent maintenance decision-making method and system based on a knowledge graph. Background Art

[0002] Currently, in the field of industrial equipment maintenance, efficient handling of complex faults is critical to ensuring production continuity. As industrial systems evolve toward higher levels of integration and intelligence, equipment operating environments are becoming increasingly complex. Traditional maintenance models that rely solely on manual experience are no longer able to meet the demand for rapid response. However, factors such as the decentralized storage of maintenance data, difficulty in transferring expert experience, and inefficient use of equipment manuals make traditional maintenance methods inadequate for complex scenarios involving multi-source, heterogeneous information, resulting in extended fault handling times and increased downtime costs.

[0003] In one existing technology, equipment maintenance decisions mainly rely on manual review of equipment manuals combined with historical maintenance records. Maintenance plans are formulated by manually searching relevant cases and referring to expert advice, lacking in-depth correlation analysis of multi-source data. At the same time, existing methods mostly use rule-based static knowledge bases for fault matching, without fully considering the impact of dynamic changes in equipment status on maintenance strategies, resulting in insufficient adaptability and deviations in new types of faults or complex scenarios. It can be seen that the equipment maintenance decisions of the existing technology have the problem of low accuracy. Summary of the Invention

[0004] The main purpose of this application is to provide an intelligent maintenance decision-making method and system based on knowledge graph to improve the accuracy of maintenance decisions for industrial equipment.

[0005] To achieve the above objectives, an embodiment of the present invention provides an intelligent maintenance decision-making method based on a knowledge graph, the method comprising: Obtain equipment operating status data and historical maintenance data of industrial equipment; Constructing an industrial knowledge graph based on the equipment operation status data and the historical maintenance data, and generating a correlation relationship model between the equipment operation status and maintenance behavior based on the industrial knowledge graph; Analyzing the relationship between the fault type and the maintenance path based on the association relationship model and the equipment operating status data to obtain the optimal maintenance path; Optimizing maintenance task allocation according to the optimal maintenance path, the association relationship model, and preset maintenance resource constraints to obtain an optimal maintenance task allocation plan; According to the optimal maintenance path, the optimal maintenance task allocation plan and the association relationship model, a dynamic adjustment analysis of the maintenance time window is performed to obtain an optimal maintenance strategy; Based on the optimal maintenance strategy, a multi-objective optimization iterative analysis is performed to obtain a final maintenance decision for the industrial equipment.

[0006] Accordingly, the present application also provides an intelligent maintenance decision system based on a knowledge graph, including: An acquisition module is used to obtain equipment operating status data and historical maintenance data of industrial equipment; a graph construction module, configured to construct an industrial knowledge graph based on the equipment operation status data and the historical maintenance data, and generate a correlation relationship model between the equipment operation status and the maintenance behavior based on the industrial knowledge graph; A relationship analysis module, configured to analyze the relationship between the fault type and the maintenance path based on the association relationship model and the equipment operation status data, and obtain the optimal maintenance path; An optimization module, configured to optimize the maintenance task allocation according to the optimal maintenance path, the association relationship model, and preset maintenance resource constraints to obtain an optimal maintenance task allocation solution; A window adjustment module is used to perform dynamic adjustment analysis of the maintenance time window based on the optimal maintenance path, the optimal maintenance task allocation plan and the association relationship model to obtain an optimal maintenance strategy; The decision module is used to perform multi-objective optimization iterative analysis based on the optimal maintenance strategy to obtain a final maintenance decision for the industrial equipment.

[0007] To sum up, by adopting the technical solution of the present application, we first obtain the operating status data and historical maintenance data of industrial equipment; then we construct an industrial knowledge graph and generate an association model, which can deeply explore the intrinsic connection between equipment operation and maintenance behavior; based on this analysis, the relationship between fault type and maintenance path is obtained to obtain the optimal maintenance path, which can effectively improve maintenance efficiency and reduce fault investigation time; through the optimal maintenance path, association model and maintenance resource constraint conditions, the maintenance task allocation plan is optimized to reasonably allocate maintenance resources and avoid resource waste; then the maintenance time window is dynamically adjusted and analyzed to obtain the optimal maintenance strategy, making the maintenance plan more flexible and adaptable; finally, multi-objective optimization iterative analysis is performed to obtain the final maintenance decision; this solution can comprehensively consider multiple factors, improve the scientificity and accuracy of industrial equipment maintenance decisions, especially in the face of complex fault scenarios, further comprehensively improve the overall benefits of industrial equipment maintenance, ensure the stable and efficient operation of industrial equipment, reduce the impact of equipment failures on production, and reduce maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 This is a scenario diagram of the intelligent maintenance decision-making method based on the knowledge graph in the embodiment of the present application; Figure 2 A flowchart of an intelligent maintenance decision-making method based on a knowledge graph is provided for an embodiment of the present application; Figure 3 A schematic diagram of the process of constructing a knowledge graph provided in an embodiment of the present application; Figure 4 A schematic diagram of a process for determining the optimal maintenance path provided in an embodiment of the present application; Figure 5 A schematic diagram of the path optimization process provided in the embodiment of the present application; Figure 6 Another schematic diagram of a process for path optimization provided in an embodiment of the present application; Figure 7 A schematic diagram of a multi-objective optimization iterative analysis process is provided for the embodiment of the present application; Figure 8 A schematic diagram of the structure of the knowledge graph-based intelligent maintenance decision system provided in an embodiment of the present application; Figure 9 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0010] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0011] The embodiments of the present application provide an intelligent maintenance decision-making method and system based on a knowledge graph, which will be described in detail below.

[0012] In the embodiment of the present application, the intelligent maintenance decision-making method based on knowledge graph is an innovative method specifically for industrial equipment maintenance decision-making. It first collects the equipment operation status data and historical maintenance data of industrial equipment, which is the basic data source of the entire method. Then, the industrial knowledge graph is constructed using this data to generate a correlation model between the equipment operation status and maintenance behavior. This model can tap into the deep connection between equipment operation and maintenance. On this basis, the optimal maintenance path is determined by analyzing the relationship between the fault type and the maintenance path, thereby improving the pertinence and efficiency of maintenance. Then, the maintenance task allocation plan is optimized by combining the optimal maintenance path, the correlation model and the preset maintenance resource constraints to achieve rational resource utilization. Based on the above results, the maintenance time window is dynamically adjusted and analyzed to obtain the optimal maintenance strategy, thereby enhancing the ability to cope with actual production changes. Finally, a multi-objective optimization iterative analysis is performed to obtain the final maintenance decision, which improves the scientificity and rationality of industrial equipment maintenance decisions as a whole, ensures stable operation of equipment and reduces maintenance costs.

[0013] As shown in Figure 1, a scenario of an intelligent maintenance decision-making method based on a knowledge graph is provided. The intelligent maintenance decision-making scenario based on a knowledge graph mainly includes data acquisition equipment, knowledge graph construction equipment, path optimization equipment, task allocation equipment, intelligent decision-making equipment, etc. Data interaction is achieved between each device through the network.

[0014] Data collection equipment is used to obtain equipment operating status data and historical maintenance data. These data come from real-time sensor data of the equipment, equipment operation log data, and maintenance operation steps, maintenance consumables usage and maintenance working hours data in historical maintenance records.

[0015] Data acquisition devices use standardization to convert this data into a unified format for use by subsequent modules. Standardization includes data cleaning and normalization to ensure data consistency and usability. For example, sensor data from different devices is normalized using maximum and minimum values ​​to map the value range to between 0 and 1. Missing values ​​are interpolated to ensure data integrity.

[0016] The knowledge graph construction device is used to process standardized data based on domain knowledge extraction algorithms to generate an industrial knowledge graph. The knowledge graph construction device first extracts initial knowledge graph nodes and edges from equipment operating status data and historical maintenance data through entity recognition and relationship extraction methods to form a preliminary knowledge graph structure. Subsequently, the association strength distribution of the initial knowledge graph is obtained by calculating the relationship weights between nodes, and the knowledge graph is mapped to a low-dimensional space using a graph embedding algorithm to enhance the graph's expressive power. Finally, a hierarchical clustering method is used to further optimize node grouping and edge weights, ultimately forming a correlation model between equipment operating status and maintenance behavior. This model can clearly reflect the intrinsic connection between equipment operating status and maintenance behavior, providing a basis for subsequent analysis.

[0017] The path optimization device is used to construct a device-generated association model based on the knowledge graph, and combined with real-time device operating status data, analyze the relationship between fault type and maintenance path and optimize it. The path optimization device first analyzes the dynamic response relationship between the probability of fault type occurrence and the execution order of the maintenance path to derive the impact weight of the fault type on the maintenance path. Then, based on the impact weight and the association relationship model, the historical matching degree matrix between the fault type and the maintenance path is calculated to determine the path selection constraints. On this basis, a simulated annealing algorithm is used to comprehensively evaluate the set of candidate maintenance paths, calculate the execution efficiency index of each path, and sort the paths. If the highest-ranked path meets the preset efficiency threshold, it is determined as the optimal maintenance path; otherwise, a local search optimization is performed on the candidate path set until the optimal solution is found.

[0018] After obtaining the optimal maintenance path, the task allocation device optimizes maintenance task allocation by combining the association model with pre-set maintenance resource constraints. First, the nonlinear relationship between maintenance task allocation and maintenance resource utilization is analyzed to construct a nonlinear maintenance resource relationship model. Next, based on this model and the optimal maintenance path, the maintenance task allocation optimization function is decomposed and iteratively calculated using a particle swarm optimization algorithm to obtain a set of maintenance task allocation solutions that meet the maintenance resource constraints. Each solution in the set is analyzed for balance, and the solution with the best balance is ultimately selected as the optimal maintenance task allocation solution. For example, in a specific task allocation, the task allocation device prioritizes the match between the maintenance personnel's skill level and the task complexity, while also taking into account the inventory status of maintenance consumables to ensure that the allocation results are both efficient and reasonable.

[0019] The planning optimization device is responsible for dynamically adjusting and analyzing the maintenance time window to optimize the maintenance plan. First, the dynamic coupling relationship between the maintenance time window and the equipment's operating status is analyzed to extract the characteristics of the equipment's operating status changes. Then, a time series prediction algorithm is used to analyze the fluctuations in the maintenance time window and determine the adjustment range. On this basis, a priority ranking function for the maintenance time window adjustment strategy is constructed, and a dynamic programming algorithm is used to analyze the global optimization results. Finally, the planning optimization device combines the optimal maintenance path, the optimal maintenance task allocation plan, and the maintenance time window adjustment strategy to comprehensively optimize the maintenance plan and generate the optimal maintenance plan and optimal maintenance strategy. For example, in one scenario, the planning optimization device will flexibly adjust the maintenance time window based on changes in the equipment's operating load to avoid production interruptions due to maintenance.

[0020] The intelligent decision-making device is used to generate the optimal maintenance strategy based on the planned optimization equipment load, perform iterative analysis using a multi-objective optimization algorithm, and output the final maintenance decision. The intelligent decision-making device first constructs a maintenance decision optimization space that includes maintenance paths, maintenance task allocation, and maintenance time windows, and combines preset maintenance cost thresholds and maintenance efficiency thresholds for collaborative constraints to generate a first maintenance decision combination. Subsequently, the multi-objective optimization algorithm is initialized and iterative calculations are performed. After each iteration, a determination is made as to whether the second maintenance decision combination is superior to the first. If so, the first maintenance decision combination is updated; otherwise, the current combination is discarded. When the number of iterations reaches a preset value, the current first maintenance decision combination is output as the final maintenance decision. For example, during a certain iteration, the intelligent decision-making device discovers that a maintenance decision combination significantly improves maintenance efficiency while reducing maintenance costs, and therefore selects it as the final decision. The above devices work closely together to achieve the overall functionality of the intelligent maintenance decision system based on the knowledge graph.

[0021] refer to Figure 2 , Figure 2 This is a flow chart of a method for intelligent maintenance decision-making based on a knowledge graph provided in an embodiment of the present application. The execution subject of the method may be a computer device, which may be a single computer device or a cluster of multiple computer devices. The computer device may be a terminal device or a server. The method for intelligent maintenance decision-making based on a knowledge graph provided in an embodiment of the present application specifically includes: Step S10: Acquire equipment operating status data and historical maintenance data of industrial equipment.

[0022] Equipment operating status data refers to a collection of various types of information generated during the operation of industrial equipment that reflects its operating status. For example, for a large industrial compressor, equipment operating status data includes but is not limited to the compressor's intake pressure, exhaust pressure, operating temperature, speed, vibration frequency, and amplitude. Equipment operating status data can be obtained through various sensors installed in key areas of the equipment, such as pressure sensors, temperature sensors, speed sensors, and acceleration sensors.

[0023] Historical maintenance data is a detailed record of past maintenance activities for industrial equipment, including the date of maintenance, reason for maintenance (e.g., fault symptoms, fault diagnosis results), location of maintenance, maintenance measures (e.g., replaced parts, maintenance process), and maintenance personnel information. This data is typically stored in the enterprise's Equipment Management Information System (EMIS) database and can be retrieved from the EMIS.

[0024] In one embodiment, IoT technology can be used to acquire status and historical data. IoT modules are installed on industrial equipment, and device operating status data is transmitted to a cloud platform in real time via wireless communication protocols. Simultaneously, data mining tools are deployed on the company's equipment maintenance database to extract historical maintenance data. This helps improve data collection efficiency and facilitates centralized management of data from multiple devices.

[0025] Step S20: constructing an industrial knowledge graph based on the equipment operation status data and the historical maintenance data, and generating an association relationship model between the equipment operation status and maintenance behavior based on the industrial knowledge graph.

[0026] The industrial knowledge graph is a graph-based data structure used to represent entities, concepts, and the relationships between them within the industrial equipment domain. In this application, entities can include equipment components (such as motors, bearings, and valves), equipment operating conditions (such as overheating and abnormal vibration), maintenance actions (such as replacing parts and adjusting parameters), and maintenance resources (such as maintenance tools and maintenance personnel skills). Relationships describe the connections between these entities. For example, "motor overheating" may have a causal relationship with the maintenance action "replacing the heat sink."

[0027] In this embodiment, when constructing an industrial knowledge graph, the equipment operating status data and historical maintenance data must first be preprocessed, including data cleaning (removing noise and duplicate data) and data annotation (adding semantic tags to the data). Knowledge extraction techniques are then used to identify entities and relationships within the data, thereby constructing the initial knowledge graph structure.

[0028] The association model is a form of representation that quantifies and models the relationships between entities based on the industrial knowledge graph. For example, by calculating indicators such as the frequency of occurrence and association strength of relationships between entities, a model can be constructed that reflects the quantitative relationship between equipment operating status and maintenance behavior.

[0029] In one embodiment, graph neural network (GNN) technology is used to construct an industrial knowledge graph and generate an association model. Preprocessed equipment operating status data and historical maintenance data are converted into a graph-structured data representation, where nodes represent entities and edges represent relationships between entities. This graph data is then input into the GNN. The GNN automatically learns the relationship features between entities by performing convolution operations on the graph structure. To generate the association model, during the GNN training process, an appropriate loss function is designed to enable the network to learn a quantitative representation of the relationships between entities. For example, the loss function can be designed based on the prediction accuracy of the relationships between entities. This GNN-based technology has strong learning capabilities and can automatically discover complex relationships in the data. Compared to traditional rule-based and handcrafted feature-based approaches, it is more suitable for processing large-scale and complex industrial equipment data.

[0030] In one embodiment, in order to improve the accuracy of industrial knowledge graph and maintenance decision making, Figure 3 As shown, the knowledge graph can be constructed in the following ways: Step S201: Based on the equipment operation status data and the historical maintenance data, construct the nodes and edges of the initial knowledge graph to obtain a preliminary knowledge graph structure.

[0031] In this step, the construction of nodes and edges forms the foundation for building the industrial knowledge graph. Nodes represent various entities related to industrial equipment maintenance. For example, in the maintenance scenario of an industrial robot, nodes could be the robot's joints (such as the shoulder and elbow joints), key components (such as the motor and reducer), potential faults (such as motor overheating and joint jamming), and maintenance operations (such as replacing the motor and lubricating the joints). Edges represent the relationships between these entities. For example, a causal relationship exists between the fault node "Motor overheating" and the maintenance operation node "Replace motor," and this relationship can be represented by an edge.

[0032] In the embodiment of the present application, the nodes and edges of the initial knowledge graph need to be deeply mined from the equipment operating status data and historical maintenance data. First, the data needs to be semantically analyzed to identify meaningful entity and relationship descriptions. For example, from the maintenance record description in the historical maintenance data, "due to bearing wear, the equipment vibration is abnormal, and the bearing replacement maintenance was performed", the entities "bearing wear", "abnormal equipment vibration", and "bearing replacement" can be identified, as well as the two relationships "bearing wear-cause-abnormal equipment vibration" and "abnormal equipment vibration-need-bearing replacement", thereby constructing the corresponding nodes and edges.

[0033] In one embodiment, semantic template matching technology can be used to achieve this. Specifically, a series of semantic templates related to industrial equipment maintenance are predefined, such as "[Component Name] - Fault Cause - [Fault Symptom]," "[Fault Symptom] - Maintenance Action - [Maintenance Measures]," and so on. The text information in the equipment operating status data and historical maintenance data is then matched against these semantic templates. Based on the matching results, entities and relationships are extracted, and the nodes and edges of the initial knowledge graph are constructed. This approach is simple and direct, and works well for data with relatively fixed structures.

[0034] Step S202: Based on the preliminary knowledge graph structure, calculate the relationship weights between nodes to obtain the association strength distribution of the initial knowledge graph.

[0035] The association strength distribution reflects the closeness of the relationships between nodes in a knowledge graph. For example, in a knowledge graph for chemical production equipment, the association between "reactor temperature too high" and "cooling system failure" might be very close, while the association with "mixer blade damage" might be relatively weak. The calculation of association strength is based on various factors, such as the frequency of simultaneous occurrence of the two in historical data and the certainty of the causal relationship.

[0036] In one embodiment, calculating the weight of the relationship between nodes requires comprehensive consideration of various information in the data. This can be done using statistical methods. For example, the ratio of the number of times "reactor temperature is too high" and "cooling system failure" appear together in all maintenance records to the total number of times "reactor temperature is too high" appears can be calculated as a weight indicator for the relationship between the two. Domain expert knowledge can also be incorporated to assign appropriate weights to pairs of nodes that are empirically determined to have close relationships.

[0037] In one embodiment, Bayesian network technology can also be used to calculate the weights of relationships between nodes. The initial knowledge graph structure is converted into a Bayesian network structure, with nodes serving as variables and edges representing conditional dependencies between variables. Based on sample data from equipment operating status data and historical maintenance data, the conditional probability distribution in the Bayesian network is estimated. This conditional probability distribution can then be used as the weights of relationships between nodes. The advantage of this approach is that it can effectively handle uncertain relationships and can easily incorporate prior knowledge.

[0038] Step S203: Enhance the preliminary knowledge graph structure according to the association strength distribution to obtain an industrial knowledge graph.

[0039] Enhancement involves optimizing the initial knowledge graph structure, enabling it to more accurately reflect the entity relationships within industrial equipment maintenance. Based on the distribution of association strength, relationships with high association strength can be more prominently represented in the knowledge graph, such as by bolding edges and increasing the density of node connections. (This is merely a conceptual representation; actual data structures will incorporate corresponding approaches.)

[0040] In one embodiment, the data structure of the knowledge graph can also be adjusted. For example, if a node has a strong correlation with multiple other nodes, a cluster analysis can be performed on the node, and the node and the strongly correlated nodes can be grouped to better reflect the relationship structure between them.

[0041] In one embodiment, a graph convolutional neural network (GCN) can be used to enhance the initial knowledge graph structure. The knowledge graph is fed into the GCN as graph data. The GCN learns feature representations for the nodes based on the distribution of inter-node association strengths. During training, the GCN automatically adjusts node representations, placing nodes with high association strengths closer together in the feature space, thereby enhancing the knowledge graph structure. This approach leverages the GCN's powerful graph data processing capabilities and effectively improves the knowledge graph's ability to represent industrial equipment maintenance relationships.

[0042] In one embodiment, the association strength distribution and the preliminary knowledge graph structure may be mapped to a low-dimensional space using a graph embedding algorithm to obtain an enhanced knowledge graph representation.

[0043] In the embodiments of this application, the graph embedding algorithm aims to convert high-dimensional graph structure data (here, the preliminary knowledge graph structure and its association strength distribution) into a vector representation in a low-dimensional space. The advantage of this is that it can preserve the graph structure information while facilitating subsequent calculations and analysis operations.

[0044] The initial knowledge graph structure for industrial equipment maintenance consists of numerous nodes and complex edge relationships, situated in a high-dimensional space. This makes data processing complex. For example, in a knowledge graph for a large industrial production system containing multiple entities such as equipment components, fault types, and maintenance operations, each entity has its own attributes and multiple relationships with other entities, forming a high-dimensional data structure. Graph embedding algorithms can map these complex structures into a low-dimensional space.

[0045] During the mapping process, graph embedding algorithms adjust the mapping results based on the distribution of association strength. Node relationships with strong association strengths will have closer vector representations in low-dimensional space, better reflecting the inherent connections within the knowledge graph. For example, if the association strength between "equipment component failure" and "specific maintenance operation" is high, then the corresponding node vectors will be closer in distance in the low-dimensional vector representation.

[0046] In one embodiment, the Node2Vec graph embedding algorithm can be used. Node2Vec is a graph embedding algorithm based on random walks. First, a random walk is performed on the preliminary knowledge graph structure, and the probability of the random walk is determined based on the distribution of association strengths. For example, edges with high association strengths are more likely to be selected for the walk. Then, by learning a large number of random walk paths, each node is mapped to a vector representation in a low-dimensional space. The advantage of this algorithm is that it can effectively capture local and global relationships in the graph structure and can perform biased random walks based on different association strengths, thereby better reflecting the relational structure in the knowledge graph.

[0047] In one embodiment, generating a correlation model between the equipment operating status and maintenance behavior based on the industrial knowledge graph in step S20 may include: optimizing node grouping and edge weights using a hierarchical clustering method based on the enhanced knowledge graph representation to obtain a correlation model between the equipment operating status and maintenance behavior.

[0048] Hierarchical clustering is a clustering method that analyzes data at different levels based on inter-cluster similarities, forming a tree-like clustering structure. Based on an enhanced knowledge graph representation, hierarchical clustering can be used to group nodes based on the similarity of their vector representations (derived from a graph embedding algorithm). For example, in a knowledge graph for industrial equipment, nodes related to the equipment's operating status (such as nodes representing different operating parameters) might be grouped together based on their similarity. Nodes related to maintenance activities (such as nodes representing different maintenance operations) could also be grouped together. This grouping helps to more clearly visualize the relationship between equipment operating status and maintenance activities.

[0049] Hierarchical clustering also optimizes edge weights. During the clustering process, edge weights between nodes within the same group may be adjusted based on the clustering results. If two nodes are determined to have a closer relationship after clustering, the edge weight between them may increase; conversely, if the relationship between the nodes is deemed relatively weak after clustering analysis, the edge weight may decrease. In this way, the structure of the knowledge graph is further optimized, ensuring that the resulting relationship model between equipment operating status and maintenance behavior more accurately reflects the actual situation.

[0050] In one embodiment, specifically, each node can be regarded as a separate class. Then, the distance between each pair of classes (nodes) is calculated. This distance can be based on a metric such as the Euclidean distance or cosine similarity represented by the node vector. Based on the calculated distance, the two classes with the closest distance are merged into a new class. This process is repeated until a predetermined stopping condition is reached, such as reaching a specified number of clusters or the distance between classes is greater than a certain threshold. In each process of merging classes, the edge weight is adjusted according to the new clustering structure. For example, if the relationship between nodes in the class becomes closer after two classes are merged, the corresponding edge weight increases. This method can naturally form a hierarchical structure and is suitable for optimizing complex knowledge graph structures.

[0051] Step S30: Analyze the relationship between the fault type and the maintenance path based on the association relationship model and the equipment operation status data to obtain the optimal maintenance path.

[0052] Fault types refer to the classification of industrial equipment based on factors such as fault symptoms and causes. For example, equipment faults on automated production lines can be categorized as mechanical (such as worn transmission components and loose connectors), electrical (such as short circuits and motor burnouts), and control system (such as program errors and sensor failures).

[0053] Among them, the maintenance path refers to an ordered combination of a series of maintenance operations taken for a specific fault type. When analyzing the relationship between the fault type and the maintenance path, it can be done based on the association model and the equipment operation status data. Specifically, in one embodiment, the knowledge in the association model is first used to infer the possible fault type based on the current equipment operation status data. For example, if the equipment operation status data shows that the current of a certain electrical equipment suddenly increases and the voltage fluctuates abnormally, combined with the characteristic relationship of electrical faults in the association model, it can be determined that the possible fault type is a circuit short circuit. Then, according to the fault type, the relevant maintenance operations and their sequential relationship are searched in the association model to form a candidate set of maintenance paths. Finally, the optimal maintenance path is determined by evaluating factors such as the cost, maintenance time, and maintenance success rate of each maintenance path.

[0054] In one embodiment, reinforcement learning technology can be used to analyze the relationship between fault type and maintenance path and determine the optimal maintenance path. The fault diagnosis and maintenance path selection process is structured as a reinforcement learning environment, where the state represents the equipment operating status data and the inferred fault type, the action represents the selected maintenance operation, and the reward function is designed based on factors such as maintenance cost, maintenance time, and repair success rate. For example, a higher reward is given when a maintenance operation successfully fixes the fault with low cost and short maintenance time. In this environment, the agent continuously interacts with the environment (selecting maintenance operations and observing the results) to learn the optimal maintenance path strategy for different fault types. This reinforcement learning implementation can continuously optimize the maintenance path selection strategy based on actual maintenance feedback, without the need for complex pre-set rules, and can adapt to dynamic changes in equipment operating status and fault type.

[0055] Step S40: optimizing maintenance task allocation according to the optimal maintenance path, the association relationship model and preset maintenance resource constraints to obtain an optimal maintenance task allocation solution.

[0056] Maintenance resource constraints refer to the various resource constraints faced during equipment maintenance, including but not limited to the number and skill levels of maintenance personnel, the types and quantities of maintenance tools, and the inventory status of maintenance parts. When optimizing maintenance task allocation, it is necessary to comprehensively consider the maintenance resource requirements of each maintenance operation in the optimal maintenance path, as well as the sequence and dependencies between maintenance operations in the association model. For example, some maintenance operations require maintenance personnel with specific skills and specialized maintenance tools and can only be performed after other maintenance operations have been completed.

[0057] In one embodiment, an integer programming algorithm can be used for optimization. Maintenance tasks are treated as variables, maintenance resource constraints serve as constraint equations, and the relationships between maintenance tasks (based on an association model) serve as part of the objective function. For example, if maintenance task A must be completed before maintenance task B, this sequential relationship is reflected in the objective function. Using the integer programming algorithm, a maintenance task allocation plan is obtained that satisfies the maintenance resource constraints and optimizes the objective function (e.g., minimizing maintenance time or cost). This approach utilizes mathematical optimization methods to determine the optimal maintenance task allocation plan while satisfying various constraints, thereby improving the utilization efficiency of maintenance resources.

[0058] Step S50: performing dynamic adjustment analysis of the maintenance time window according to the optimal maintenance path, the optimal maintenance task allocation plan and the association relationship model to obtain an optimal maintenance strategy.

[0059] The maintenance time window refers to the time period during the operation of industrial equipment that is suitable for maintenance operations. The determination of this time period requires consideration of multiple factors, including the equipment's operation plan, production task schedule, and equipment operating status. For example, for chemical equipment in continuous production, when production tasks are tight, the maintenance time window may only be selected during the equipment's scheduled maintenance downtime or the brief interval between production task switching; when production tasks are relatively relaxed, maintenance operations can be performed during the equipment's low-load operation phase. When dynamically adjusting the maintenance time window based on the optimal maintenance path, the optimal maintenance task allocation plan, and the association relationship model, the equipment's operating status over a period of time in the future should be predicted based on the equipment's operating status data. The optimal time window for each maintenance operation should be determined based on the characteristics of the maintenance operation (e.g., some maintenance operations require equipment shutdown, while others can be performed while the equipment is operating) and the maintenance sequence in the maintenance task allocation plan, thereby obtaining the optimal maintenance strategy.

[0060] In one embodiment, a long short-term memory network (LSTM) or a gated recurrent unit (GRU) can be used to dynamically adjust and analyze maintenance time windows. Equipment operating status data is input into the network in a time series format. The network then learns the changing patterns of equipment operating status based on historical data. Simultaneously, the relationship between maintenance operations and equipment operating status is modeled by combining optimal maintenance paths, optimal maintenance task allocation plans, and relevant information from the association model. Leveraging the network's predictive capabilities, the equipment's operating status at different time periods in the future can be predicted in advance, thereby determining the optimal maintenance time window for each maintenance operation and deriving the optimal maintenance strategy. This deep learning-based time series prediction method can better capture the dynamic changes in equipment operating status and improve the accuracy of maintenance time window predictions.

[0061] Step S60: performing a multi-objective optimization iterative analysis based on the optimal maintenance strategy to obtain a final maintenance decision for the industrial equipment.

[0062] Multi-objective optimization refers to the simultaneous consideration of multiple objective functions within a single optimization problem. In industrial equipment maintenance decisions, these multiple objectives typically include maintenance costs, maintenance time, equipment reliability, and equipment lifespan. These objectives often constrain each other. For example, reducing maintenance costs may result in longer maintenance times or reduced equipment reliability. When performing iterative multi-objective optimization analysis based on optimal maintenance strategies, the goal is to find an optimal solution that balances all objectives while satisfying their constraints.

[0063] In one embodiment, an AI-based evolutionary multi-objective optimization algorithm, such as the non-dominated sorting genetic algorithm (NSGA-II), can be used for iterative multi-objective optimization analysis. Each solution in the optimal maintenance strategy is encoded as a chromosome, with each chromosome representing a possible maintenance decision. Multiple objectives, such as maintenance cost, maintenance time, and equipment reliability, are used as components of the fitness function. Through selection, crossover, and mutation operations within the NSGA-II algorithm, the population is continuously evolved. In each generation, superior individuals (maintenance decision solutions) are selected based on non-dominated sorting and crowding distance calculations. After multiple generations of iteration, a set of Pareto optimal solutions is obtained—solutions in which no objective can be improved without degrading other objectives. Finally, a final maintenance decision is selected from these Pareto optimal solutions based on the enterprise's actual needs (e.g., emphasis on maintenance cost, minimum equipment reliability requirements, etc.). The NSGA-II algorithm can effectively handle the complex relationships between multiple objectives, providing multiple representative optimal solutions for enterprises to choose from, which better meets the multi-objective optimization requirements of actual industrial equipment maintenance decisions.

[0064] In one embodiment, reference Figure 4 To further improve the accuracy of decision-making, step S30 of this application may specifically include: Step S301: Analyze the dynamic response relationship between the probability of occurrence of a fault type and the execution order of a maintenance path based on the association relationship model and the equipment operation status data, and obtain the impact weight of the fault type on the maintenance path.

[0065] The association model is a data model built based on knowledge about industrial equipment. It describes the complex relationships between entities such as equipment operating status, fault type, and maintenance actions. These relationships, represented through quantitative and logical connections, serve as a crucial basis for analyzing equipment maintenance decisions. For example, it might indicate the likelihood of a specific fault type occurring under a given equipment operating state and the degree of correlation between it and the corresponding maintenance action.

[0066] The probability of a fault type refers to the likelihood of a particular fault type occurring under specific equipment operating conditions. This probability is calculated based on a variety of factors, including the equipment's historical fault data, current operating parameters, and environmental factors. For example, for a motor that frequently operates in high-temperature environments, the probability of a winding short-circuit fault may increase with increasing operating time and temperature.

[0067] The maintenance path execution sequence is the order in which a series of repair actions are planned for different fault types. This sequence is determined based on factors such as the device structure, the fault mechanism, and the effectiveness and efficiency of the repair. For example, if a complex device with multiple components fails, it may be necessary to first check external connectors (such as wiring connections and interfaces) before further inspecting core internal components (such as circuit boards and sensors). This is a typical maintenance path execution sequence.

[0068] In the embodiment of the present application, the association model provides basic relationship information between fault types and maintenance paths, and the equipment operation status data reflects the actual current operation status of the equipment. There is a dynamic response relationship between the probability of occurrence of a fault type and the execution order of the maintenance path. For example, when the equipment operation status data shows that the temperature of a certain component of the equipment suddenly rises, the probability of a certain temperature-related fault type (such as an overheating fault) will increase. This increase in probability may prompt a change in the execution order of the maintenance path, and the inspection or processing operations for overheating faults that were originally performed later in the maintenance path may be brought forward.

[0069] By analyzing this dynamic response relationship, the embodiments of the present application can determine the impact weight of the fault type on the maintenance path. This weight is a quantitative value that reflects the importance of the fault type in influencing the maintenance path selection. For example, if a fault type, upon occurrence, causes severe damage to the equipment and develops rapidly, its impact weight on the maintenance path will be high, meaning that maintenance operations targeting this fault type should be prioritized when determining the maintenance path.

[0070] In one embodiment, a Markov chain model is constructed based on an association model and equipment operating status data. In this model, the state space represents different fault types and maintenance path execution stages. For example, different fault type states (such as initial, intermediate, and severe) and each operation step during the maintenance path execution process are considered different states. Then, using the MCMC (Markov Chain Monte Carlo) algorithm, simulations are performed to estimate the transition of fault type probability and the corresponding adjustment of the maintenance path execution sequence based on dynamic information in the equipment operating status data (such as the changing trends of equipment operating parameters). During the simulation, the frequency and magnitude of changes in the maintenance path execution sequence when the fault type probability changes are statistically analyzed to calculate the impact weight of the fault type on the maintenance path. This method can effectively handle complex dynamic relationships and account for uncertainty in the data.

[0071] Step S302: Analyze the historical matching matrix between fault types and maintenance paths based on the impact weights and the association relationship model to obtain path selection constraints.

[0072] The impact weight is a quantitative value that reflects the importance of the fault type in influencing the maintenance path selection. This value is obtained by analyzing the dynamic response relationship between the probability of the fault type and the maintenance path execution sequence, and serves as an important reference factor in subsequent analysis.

[0073] The historical matching matrix is ​​a matrix constructed based on historical data. It records the success rate of matching different fault types with various repair paths during past equipment maintenance. The rows of the matrix represent different fault types, and the columns represent different repair paths. The elements in the matrix represent the number of successful matches, success rates, or other relevant matching metrics between a specific fault type and a repair path. For example, if repair path B has historically successfully repaired equipment for fault type A more often than not, then the corresponding element in row A, column B in the historical matching matrix will have a higher value.

[0074] In this embodiment, the impact weight reflects the importance of the fault type to the maintenance path, while the association model contains rich information about the relationships between device-related entities. The historical matching matrix between fault types and maintenance paths provides historical empirical data. By analyzing the impact weight and association model, we can gain a deeper understanding of the inherent connection between fault types and maintenance paths.

[0075] Then, combined with the historical matching matrix, path selection constraints can be determined. These constraints include requirements for the maintenance path's historical success rate and restrictions on the repair operation sequence for specific fault types. For example, if a fault type has a high impact on the maintenance path, and the historical matching matrix shows that only a specific maintenance path has a high success rate for that fault type, then this maintenance path becomes a key constraint, meaning that it will be prioritized or other irrelevant maintenance paths will be excluded.

[0076] In one embodiment, the impact weights, relevant information from the association model, and data from the historical matching matrix are integrated to form a data set suitable for mining. Then, an association rule mining algorithm, such as the Apriori algorithm, is applied to identify association rules between fault types, impact weights, historical matching, and maintenance paths. These association rules serve as path selection constraints. For example, mining can yield a rule such as "If the impact weight of fault type A is greater than a certain threshold and its matching degree with maintenance path B in the historical matching matrix is ​​the highest, then maintenance path B is the preferred path." This technical implementation can automatically discover hidden relationships within large amounts of historical data, providing a basis for determining path selection constraints.

[0077] Step S303: Obtain candidate maintenance paths according to the path selection constraint conditions to obtain a candidate maintenance path set.

[0078] Path selection constraints are a set of conditions that restrict and filter maintenance paths. These constraints are derived by comprehensively considering factors such as the impact of fault types on maintenance paths, the historical matching matrix, and the association model. They include various requirements that maintenance paths must meet, such as restrictions on the maintenance operation sequence and historical success rate.

[0079] In this application, path selection constraints serve as the basis for determining candidate maintenance paths. These constraints limit the range of possible maintenance paths. For example, if the constraints obtained in the previous step specify that certain detection operations must be performed for a certain fault type, then maintenance paths that do not include these detection operations will not be included in the candidate range.

[0080] In real-world industrial equipment maintenance scenarios, there may be multiple repair paths that satisfy path selection constraints. For example, for a large piece of equipment with multiple subsystems, when a subsystem fails, different repair paths may exist, all of which satisfy constraints determined by the fault type, impact weight, and historical matching. These satisfying repair paths constitute the candidate repair path set.

[0081] In one embodiment, path selection constraints can be input into the inference engine in the form of rules. The inference engine traverses a predefined maintenance path library and, based on the rules, selects maintenance paths that meet the path selection constraints, forming a set of candidate maintenance paths. For example, if a rule states, "For fault type X, the maintenance path must include operation Y and cannot include operation Z," the inference engine will search and filter the maintenance path library based on this rule. This simple and direct approach allows for the rapid and accurate acquisition of candidate maintenance paths based on the rules.

[0082] In one embodiment, a simulated annealing algorithm can be used to calculate a comprehensive evaluation function of the maintenance path cost and maintenance effect according to the path selection constraint conditions to obtain a candidate maintenance path set. The simulated annealing algorithm is a randomized search algorithm based on the principles of the physical annealing process. In this scenario, it can be used to optimize a comprehensive evaluation function of repair path cost and repair effectiveness while satisfying path selection constraints. By accepting less favorable solutions with a certain probability, the simulated annealing algorithm escapes the local optimum, thereby increasing the chance of finding the global optimum. Much like the annealing process of metals, where atoms at high temperatures have sufficient energy to escape the local minimum energy state and find a more stable global minimum energy state, the "temperature" parameter in the simulated annealing algorithm controls the probability of accepting less favorable solutions. As the temperature decreases, the algorithm gradually converges to the optimal solution.

[0083] Maintenance path costs encompass all expenses associated with the repair process. These can include direct costs, such as the purchase of parts required for repair, the cost of repair tools, and the labor costs of repair personnel; as well as indirect costs, such as production losses during equipment downtime and energy consumption during the repair process. For example, for a large piece of industrial equipment, if a maintenance path requires expensive imported parts and takes a long time to repair, resulting in prolonged equipment downtime, then the cost of this maintenance path will be relatively high.

[0084] Maintenance effectiveness measures the effectiveness of a maintenance path in repairing equipment failures and the degree to which equipment performance is restored after the repair. This can be assessed from multiple perspectives, such as the success rate of repairs, the stability of equipment operation after repairs, and the time between failures. For example, a maintenance path is considered effective if it completely resolves equipment failures and maintains stable operation for an extended period of time.

[0085] A comprehensive evaluation function is a mathematical function that comprehensively considers the cost and effectiveness of a repair path. Its purpose is to compare and select between different repair paths. By assigning appropriate weights to these two factors, it unifies these two factors into a single quantitative metric. For example, the repair effectiveness can be quantified in some way and then weighted together with the repair cost to produce a comprehensive score. This score is the output of the comprehensive evaluation function.

[0086] Specifically, the search space is determined based on the path selection constraints. This means that among all possible maintenance paths, only those that meet constraints such as the repair operation sequence and required resources are selected as the initial search scope. For example, if a constraint stipulates that a fault repair must precede a detection operation, then maintenance paths that do not include this detection operation are excluded from the search scope.

[0087] Next, a comprehensive evaluation function for repair path costs and repair effectiveness is constructed. This function requires determining the quantification method for repair costs and repair effectiveness, as well as their weighting in the comprehensive evaluation, based on the actual situation. For example, the repair cost can be set as C and the repair effectiveness as E. The comprehensive evaluation function F = w1 × C + w2 × E, where w1 and w2 are the weights of the repair cost and repair effectiveness, respectively, and w1 + w2 = 1. The choice of weights depends on the company's emphasis on cost and effectiveness. If a company prioritizes cost control, it may assign a larger value to w1; if it prioritizes rapid and effective equipment repair, it may assign a larger value to w2.

[0088] The simulated annealing algorithm searches within this search space and based on a comprehensive evaluation function. At the beginning of the algorithm, a high "temperature" value is set. At this point, the algorithm is more likely to accept poorer solutions, allowing it to explore a larger search space. As the "temperature" gradually decreases, the probability of accepting poorer solutions also decreases, and the algorithm gradually converges to a better solution. During the search process, a new maintenance path is generated each time (by randomly adjusting the current path, such as changing the order of maintenance operations or replacing maintenance tools), and the comprehensive evaluation function is calculated for that path. If the evaluation function value of the new path is better than the current optimal solution or meets a certain acceptance probability (according to the simulated annealing algorithm's probabilistic acceptance criterion), the new path is accepted as the current optimal solution.

[0089] By continuously iterating this process, the algorithm will eventually obtain a series of maintenance paths that meet the path selection constraints and have better comprehensive evaluation function values. These maintenance paths constitute the candidate maintenance path set.

[0090] Step S304: determining an optimal maintenance path according to the candidate maintenance path set.

[0091] The candidate maintenance path set is a collection of maintenance paths that satisfy the path selection constraints. Each element (maintenance path) in this set is selected after considering factors such as the impact weight of the fault type on the maintenance path, the historical matching matrix, and the association model. It is a potential maintenance path that may be used to repair the equipment fault.

[0092] The candidate maintenance path set contains multiple possible maintenance paths, from which the optimal one must be determined. This requires considering multiple factors, such as maintenance cost, maintenance time, and repair success rate. For example, one candidate maintenance path might have a lower cost but a longer maintenance time, while another might have a shorter maintenance time but a lower repair success rate.

[0093] To determine the optimal maintenance path, each candidate path can be quantitatively evaluated. For example, factors such as maintenance cost, maintenance time, and repair success rate can be weighted differently, and then a comprehensive score can be calculated for each candidate path. The candidate path with the highest comprehensive score is the optimal path. Furthermore, factors such as the equipment's current production schedule and the equipment's remaining service life can also be considered to influence the maintenance path.

[0094] In one embodiment, a hierarchical model can be constructed, with factors such as maintenance cost, maintenance time, and maintenance success rate serving as the criteria layer and candidate maintenance paths serving as the solution layer. A judgment matrix is ​​then constructed through pairwise comparisons to determine the relative importance, or weights, of each factor. Next, a weight vector is calculated for each candidate maintenance path relative to each criterion, and finally, a weighted summation is performed to determine the overall score for each candidate maintenance path. The candidate maintenance path with the highest overall score is the optimal maintenance path. This approach systematically considers the relationships between multiple factors, thereby scientifically determining the optimal maintenance path.

[0095] In one embodiment, step S304 may specifically include: Based on the set of candidate maintenance paths, the execution efficiency index of each path is calculated to obtain a path execution efficiency ranking. When the path execution efficiency ranking meets a preset efficiency threshold, the candidate maintenance path with the highest ranking is determined to be the optimal maintenance path. When the path execution efficiency ranking does not meet the preset efficiency threshold, a local search optimization is performed on the set of candidate maintenance paths to obtain the optimal maintenance path.

[0096] The execution efficiency index is a comprehensive quantitative indicator that measures the efficiency of a maintenance path during its actual execution. It can include multiple factors, such as maintenance time, resource utilization, and operational complexity. Maintenance time refers to the time it takes from the start of a maintenance operation to the restoration of normal equipment operation, including the duration of the maintenance operation and possible waiting time (such as waiting for parts to arrive or for specific maintenance personnel to arrive). Resource utilization refers to the extent to which various resources (such as maintenance tools, maintenance personnel, and parts) are effectively utilized during the maintenance process. For example, if a maintenance path can fully utilize existing maintenance tools and personnel skills without causing idle or wasted resources, then it will score higher in terms of resource utilization. Operational complexity reflects the difficulty of the maintenance operation. Simpler operations are generally easier to perform and have a lower probability of error, which also affects the execution efficiency index.

[0097] In the embodiment of the present application, for each maintenance path in the candidate maintenance path set, calculating its execution efficiency index is a comprehensive evaluation process. First, consider the maintenance time factor. Suppose that in a maintenance scenario of an automated production equipment, there is a candidate maintenance path that requires troubleshooting and repairing the control system of the equipment. If this path involves multiple complex detection steps, and there is a long waiting time between each step (for example, waiting for a certain test result to be fed back before proceeding to the next step), then its maintenance time is relatively long. On the contrary, another maintenance path may use a more direct detection and repair method, which can quickly locate and solve the problem, and its maintenance time is shorter.

[0098] Regarding resource utilization, consider a maintenance scenario involving multiple specialized tools and personnel with varying skill sets. If one maintenance path requires a specific, limited-availability specialized tool and the full involvement of a skilled maintenance personnel, this can leave other personnel and tools idle, resulting in low resource utilization. However, another maintenance path, which allows for flexible deployment of existing personnel and tools to maximize their effectiveness, offers greater resource utilization advantages.

[0099] For complex industrial equipment, such as large chemical reactors, a maintenance path with numerous steps requiring high precision, high skill levels, and high error rates is considered high in operational complexity. In contrast, a maintenance path with relatively simple steps requiring moderate skill levels is rated higher in operational complexity.

[0100] By comprehensively calculating factors such as maintenance time, resource utilization, and operational complexity according to specific weights (for example, the weights of different factors can be determined based on the company's actual situation, such as 40% for maintenance time, 30% for resource utilization, and 30% for operational complexity), the execution efficiency index of each candidate maintenance path is obtained. The candidate maintenance paths are then ranked based on these indicators to obtain a path execution efficiency ranking.

[0101] The preset efficiency threshold is a standard value set by the enterprise based on various factors, including production requirements, equipment importance, and cost control. It measures whether the execution efficiency of a candidate maintenance path meets acceptable standards. For example, if the enterprise requires equipment maintenance to be completed within a short timeframe to minimize production impact, the efficiency threshold might be set high, requiring the execution efficiency of the maintenance path to reach a high value. If the equipment is not critical and the enterprise has a high tolerance for maintenance time, the efficiency threshold might be lower.

[0102] In the embodiments of the present application, local search optimization is an optimization method that searches and improves within a localized set of candidate maintenance paths. Based on the current set of candidate maintenance paths, it attempts to improve their execution efficiency by adjusting portions of the path. These adjustments typically optimize local operations, resource allocation, and so on, without changing the overall framework of the maintenance path. For example, if a maintenance operation within a maintenance path is found to utilize more efficient tools or be performed in parallel with other operations, such local adjustments can be used to improve the execution efficiency of the entire maintenance path.

[0103] After obtaining the path execution efficiency ranking, this application first compares this ranking with a preset efficiency threshold. If the path execution efficiency ranking meets the preset efficiency threshold, this means that some maintenance paths in the current set of candidate maintenance paths have achieved an acceptable execution efficiency level for the enterprise. In this case, it is reasonable to directly determine the candidate maintenance path with the highest ranking (i.e., the highest execution efficiency index) as the optimal maintenance path. This is because this maintenance path not only meets the various constraints previously mentioned (such as path selection constraints, comprehensive cost and effectiveness assessment, etc.), but also has the highest execution efficiency and is the most suitable maintenance solution currently available.

[0104] However, if the path execution efficiency ranking does not meet the preset efficiency threshold, it means that the overall execution efficiency of the current candidate maintenance path set is not ideal and needs further optimization. At this time, the local search optimization method is used to process the candidate maintenance path set. For example, for an electronic equipment maintenance scenario, some maintenance paths in the candidate maintenance path set have low execution efficiency. This may be because the order of certain maintenance operations is not reasonable or the utilization of certain resources is not sufficient. Through local search optimization, the order of maintenance operations in these maintenance paths can be readjusted to see if the maintenance time can be shortened; or resource allocation can be optimized to improve resource utilization, thereby improving the execution efficiency index.

[0105] A variety of techniques can be employed during local search optimization. For example, a detailed operational flow analysis can be conducted on each candidate maintenance route to identify potential bottlenecks (e.g., where an operation takes too long or a resource is idle), and then improvements can be proposed to address these bottlenecks. Alternatively, heuristic algorithms, such as simulated annealing (applied locally) or genetic algorithms, can be used to make local adjustments to the maintenance route without changing the overall framework. New solutions can be continuously tested until an optimal maintenance route that meets a preset efficiency threshold is found.

[0106] In one embodiment, the local search optimization based on the greedy algorithm can be For candidate maintenance paths that do not meet the preset efficiency threshold, the analysis begins with the first maintenance operation on each path. For each maintenance operation, the analysis examines whether there are alternative operations that could improve execution efficiency or whether the order of operations can be adjusted to improve efficiency. For example, if a maintenance operation requires a long wait before the next operation, the analysis examines whether other operations can be inserted into this waiting time. If an adjustment is found that could improve efficiency, the maintenance path is partially modified. The execution efficiency index for this maintenance path is then recalculated.

[0107] Repeat the above steps until a round of local search optimization is performed on each maintenance path in the set.

[0108] If after one round of optimization, a maintenance path meets the preset efficiency threshold, then the maintenance path with the highest execution efficiency index is selected as the optimal maintenance path; if there is still no maintenance path that meets the threshold, the next round of local search optimization will continue until the optimal maintenance path that meets the requirements is found or the set maximum number of search rounds is reached.

[0109] In one embodiment, reference Figure 5 Step S40 may specifically include the following steps: Step S401: performing a nonlinear relationship analysis between maintenance task allocation and maintenance resource utilization based on the association relationship model to obtain a maintenance resource nonlinear relationship model.

[0110] Maintenance task allocation is the process of breaking down the entire maintenance work into multiple specific tasks and determining which maintenance resources (such as personnel, tools, and parts) are responsible for performing each task. For example, the maintenance of a large piece of equipment may include fault detection, parts replacement, and equipment commissioning, each of which requires specific maintenance resources.

[0111] Maintenance resource utilization involves the use of various resources during the maintenance process, including resource input, utilization efficiency, idle time, etc. For example, maintenance personnel's work schedule, maintenance tool usage frequency, and parts inventory consumption are all manifestations of maintenance resource utilization.

[0112] Nonlinear relationships mean that the relationship between maintenance task allocation and maintenance resource utilization is not a simple linear proportional relationship. For example, increasing the number of maintenance personnel does not necessarily linearly reduce maintenance time. There may be an optimal staffing range, and exceeding or falling below this range actually reduces maintenance efficiency. This is due to complex factors such as task coordination, resource competition, and operational space constraints in the maintenance process.

[0113] In embodiments of the present application, information related to maintenance tasks and maintenance resources can be extracted from the association model. For example, the model may include the types of skills and resources required for maintenance tasks corresponding to different fault types. For example, repairing an electrical fault may require personnel with electrician skills and electrical testing tools; repairing a mechanical fault may require personnel with mechanical maintenance skills and the corresponding mechanical tools.

[0114] Next, consider how the sequencing and dependencies between maintenance tasks affect maintenance resource utilization. Suppose a piece of equipment maintenance involves multiple tasks, some of which must be completed before others. This sequencing relationship affects the scheduling and utilization efficiency of maintenance resources. For example, in the maintenance of equipment on a complex industrial production line, replacing a critical component (Task A) may require first disassembling the equipment (Task B), which requires specific tools and personnel. Improper allocation of maintenance resources between Tasks A and B can result in idle tools or waiting personnel, thus demonstrating nonlinear resource utilization.

[0115] By analyzing a large amount of maintenance task allocation and corresponding maintenance resource utilization data (this data can be obtained from historical maintenance records, monitoring of actual maintenance scenarios, and other sources), we can identify the complex relationships between various factors. For example, the relationship between the number of maintenance personnel and maintenance time may not be a simple inverse relationship; there may be a critical number of personnel around which the trend of maintenance time changes. Based on these analysis results, a nonlinear relationship model for maintenance resources is constructed. This model can be a mathematical equation, a set of logical rules, or a predictive model based on a machine learning algorithm to describe the nonlinear relationship between maintenance task allocation and maintenance resource utilization.

[0116] In one embodiment, a neural network-based nonlinear relationship modeling approach can be employed. Specifically, a large amount of maintenance task allocation and maintenance resource utilization data is collected as training samples. This data should include maintenance task type, task allocation (e.g., assigned personnel, tools, etc.), and corresponding maintenance resource utilization metrics (e.g., maintenance time, resource idle rate, etc.). A neural network model, such as a multilayer perceptron (MLP), is constructed. Input layer nodes correspond to variables related to maintenance task allocation (e.g., task type, number of personnel, tool type, etc.), while output layer nodes correspond to maintenance resource utilization metrics (e.g., maintenance time, resource idle rate, etc.). The neural network is trained using the collected training samples. During training, the neural network adjusts its internal weights and biases to learn the nonlinear mapping relationship between maintenance task allocation and maintenance resource utilization. After sufficient training, the resulting neural network model is a maintenance resource nonlinear relationship model. This model can predict corresponding maintenance resource utilization outcomes based on the input maintenance task allocation. Neural network-based technology can automatically learn complex nonlinear relationships without prior assumptions about the specific form of the relationships. This makes it suitable for processing large amounts of data and complex relationships in maintenance task allocation and maintenance resource utilization scenarios.

[0117] Step S402: performing function decomposition based on the maintenance resource nonlinear relationship model and the optimal maintenance path in combination with preset maintenance resource constraints to obtain a maintenance task allocation optimization function.

[0118] The optimal repair path is the sequence and process for determining the best repair actions for a device failure. It can include a series of ordered steps from fault detection and repair operations to restoring normal equipment operation. This optimal solution is determined after considering multiple factors, including the type of failure, repair cost, and repair effectiveness.

[0119] Pre-set maintenance resource constraints are restrictions set by the enterprise based on its maintenance resource situation (e.g., number of personnel, availability of tools and equipment, parts inventory, etc.). For example, the enterprise may stipulate that when repairing a certain type of equipment, no more than a certain number of maintenance personnel with specific skills can be used, or that a certain expensive maintenance tool can only be used a limited number of times.

[0120] Function decomposition is a method that transforms complex relationships or models into functional forms that are easier to analyze and optimize. In this step, the nonlinear relationship model of maintenance resources, the optimal maintenance path, and the maintenance resource constraints are combined to decompose a function specifically designed for optimizing maintenance task allocation.

[0121] In the embodiment of the present application, the nonlinear relationship model of maintenance resources provides a basic relationship framework between maintenance task allocation and maintenance resource utilization. The optimal maintenance path determines the operation sequence of the maintenance work, and the preset maintenance resource constraints set the boundaries of resource utilization.

[0122] First, each maintenance task in the optimal maintenance path is matched by sequence and type to the relevant components of the nonlinear maintenance resource model. For example, if the first maintenance task in the optimal maintenance path is electrical fault detection, then the maintenance resource utilization components related to electrical fault detection are found in the nonlinear maintenance resource model, such as the number of electricians required and the usage of electrical inspection tools. Then, combined with the preset maintenance resource constraints, this information is integrated into a function. For example, let the number of maintenance personnel be x, the number of tool usage times be y, and the maintenance task execution time be z. The maintenance resource constraints may contain constraints such as x ≤ X (X is the preset upper limit on the number of personnel) and y ≤ Y (Y is the preset upper limit on the number of tool usage times). Based on the nonlinear maintenance resource model, the relationship z = f (x, y) (f is a nonlinear function) may exist. Furthermore, by considering factors such as the temporal order of the maintenance tasks in the optimal maintenance path, a maintenance task allocation optimization function is constructed that incorporates all this information. This function aims to optimize the maintenance task allocation while satisfying the maintenance resource constraints to achieve a specific optimal outcome (such as the shortest maintenance time or lowest maintenance cost).

[0123] In one embodiment, a mathematical expression can be determined between maintenance task allocation and maintenance resource utilization variables based on a nonlinear maintenance resource relationship model. For example, if a nonlinear relationship exists between maintenance time, the number of maintenance personnel, and the number of tool usages, the relationship can be expressed as t = g (p, t), where t is the maintenance time, p is the number of maintenance personnel, t is the number of tool usages, and g is a nonlinear function.

[0124] Translate the order and time relationships of the maintenance tasks in the optimal maintenance path into mathematical constraints. For example, if maintenance task A must be completed before maintenance task B, and the execution time of maintenance task A is t1 and the execution time of maintenance task B is t2, then the constraint t1 + Δt ≤ t2 can be obtained (Δt is the minimum time interval between tasks A and B).

[0125] Combined with pre-set maintenance resource constraints, such as the maintenance personnel limit p ≤ P (where P is the pre-set number of personnel) and the tool usage limit t ≤ T (where T is the pre-set number of tool usage), a complete maintenance task allocation optimization function is constructed. This function can be an objective function (such as minimizing maintenance time or cost) plus a series of constraints, such as min (t) st p ≤ P, t ≤ T, t1 + Δt ≤ t2.

[0126] The above-mentioned technical implementation method based on mathematical programming transforms various relationships and constraints into clear mathematical forms, which is convenient for solving using optimization algorithms.

[0127] Step S403: According to the maintenance task allocation optimization function, a particle swarm optimization algorithm is used to perform iterative calculations to obtain a maintenance task allocation solution set that meets the maintenance resource constraints.

[0128] The particle swarm optimization algorithm (PSO) is an optimization algorithm based on swarm intelligence that simulates the behavior of flocks of birds or schools of fish. In the algorithm, each particle represents a possible solution (in this step, the maintenance task allocation scheme). The particle moves through the search space, continuously updating its position to find the optimal solution. The particle's position represents the specific maintenance task allocation scheme (e.g., personnel assigned to tasks, tool allocation, etc.), while the particle's velocity determines the direction and speed of its movement within the search space.

[0129] The maintenance task allocation solution set is a set of maintenance task allocation schemes that meet the maintenance resource constraints. Through the iterative calculation of the particle swarm optimization algorithm, multiple possible maintenance task allocation schemes are obtained, which are all within the allowable range of the maintenance resource constraints.

[0130] In one embodiment, based on the maintenance task allocation optimization function, the particle swarm optimization algorithm starts iterative calculation, specifically: First, a swarm of particles is initialized, each representing an initial maintenance task allocation plan. These initial plans can be randomly generated, but must satisfy maintenance resource constraints. For example, for each particle, maintenance personnel are randomly assigned to various maintenance tasks, ensuring that the number of assigned personnel does not exceed the preset personnel number constraint and that the allocation of tools also meets the corresponding constraints.

[0131] Then, the fitness value of each particle is calculated. The fitness value is calculated based on the maintenance task allocation optimization function and reflects the quality of each maintenance task allocation solution. For example, if the maintenance task allocation optimization function aims to minimize maintenance time, the fitness value can be the inverse of the maintenance time, with solutions with shorter maintenance time having higher fitness values.

[0132] In each iteration, particles update their speed and position based on their own experience (individual optimal position) and the experience of the swarm (global optimal position). The individual optimal position refers to the best position (i.e., maintenance task allocation solution) found by each particle during the search process, while the global optimal position is the best position found by the entire swarm. Particles adjust their speed to move toward more optimal positions, thereby exploring different maintenance task allocation solutions.

[0133] As the iterations progress, the particle swarm gradually converges to an area that satisfies the maintenance resource constraints, ultimately obtaining a set of maintenance task allocation solutions, known as the maintenance task allocation solution set. These solutions are all within the maintenance resource constraints and are optimal solutions found through the particle swarm optimization algorithm.

[0134] In one embodiment, the size of the particle group may be determined, for example, the number of particles may be set to N.

[0135] For each particle i (i = 1, 2, …, N), its position vector Xi and velocity vector Vi are randomly initialized. The position vector Xi represents the maintenance task allocation scheme, for example, Xi = [x1, x2, …, xn], where xj represents the resources (such as personnel, tools, etc.) assigned to the jth maintenance task; the velocity vector Vi represents the particle's movement speed, for example, Vi = [v1, v2, …, vn].

[0136] According to the maintenance task allocation optimization function, the fitness value Fi of each particle is calculated.

[0137] Initialize the individual optimal position Pbest_i of each particle to its initial position Xi, and the global optimal position Gbest to the position of the particle with the highest fitness value.

[0138] In each iteration k, for each particle i, its velocity and position are updated according to the following formula: Vi(k + 1)= w * Vi(k)+ c1 * r1 * (Pbest_i - Xi(k))+ c2 * r2 * (Gbest -Xi(k)) Xi(k + 1) = Xi(k) + Vi(k + 1) Among them, w is the inertia weight, which is used to balance the global search and local search capabilities of the particle; c1 and c2 are learning factors, which control the degree to which the particle approaches the individual optimal position and the global optimal position respectively; r1 and r2 are random numbers between [0, 1].

[0139] According to the updated position Xi (k + 1), the fitness value Fi (k + 1) of the particle is recalculated.

[0140] If Fi (k + 1) is better than the fitness value of Pbest_i, update Pbest_i = Xi (k + 1); if Fi (k + 1) is better than the fitness value of Gbest, update Gbest = Xi (k + 1).

[0141] Repeat the above steps until the preset number of iterations is reached or the convergence condition is met (such as the global optimal position does not change in multiple iterations).

[0142] Finally, all obtained Pbest_i constitute the maintenance task allocation solution set that satisfies the maintenance resource constraints.

[0143] This standard particle swarm optimization algorithm implementation is an effective solution method that can improve efficiency and accuracy by continuously adjusting the speed and position of particles to search for the optimal solution.

[0144] Step S404: Analyze the balance of maintenance task allocation based on the maintenance task allocation solution set and the maintenance resource nonlinear relationship model, and determine the solution with the best balance in the maintenance task allocation solution set as the optimal maintenance task allocation solution. In the context of maintenance task allocation, balance refers to the degree to which maintenance tasks are evenly distributed across different maintenance resources (such as personnel and tools). For example, in personnel allocation, if some maintenance personnel are overburdened with tasks while others are underburdened, this indicates an imbalance in task distribution. Similarly, in tool usage, if some tools are overused while others remain idle, this also indicates an imbalance. Balanced task distribution helps improve maintenance efficiency and reduce resource waste and personnel fatigue.

[0145] In the embodiment of the present application, the maintenance task allocation solution set includes multiple maintenance task allocation schemes that meet the maintenance resource constraints, and the optimal scheme needs to be selected from these schemes. Specifically: First, the balance of each maintenance task allocation scheme is analyzed based on the nonlinear relationship model of maintenance resources. To determine the balance of personnel allocation, we can calculate the degree of variation in the workload of each maintenance personnel. For example, by calculating the standard deviation of the workload of each personnel's assigned tasks, a small standard deviation indicates a relatively balanced personnel-task allocation. To determine the balance of tool usage, we can analyze differences in tool usage frequency. If a tool is used very frequently in a maintenance task allocation scheme, while other tools are rarely used, this indicates an imbalance in tool usage.

[0146] In addition to balancing personnel and tools, it's also important to consider the balance between maintenance tasks. For example, some maintenance tasks may take too long to complete due to inappropriate resource allocation, impacting the efficiency of the entire maintenance process. A maintenance task allocation plan in which the execution times of individual tasks are relatively close is considered a good balance in terms of maintenance task balance.

[0147] Each solution in the maintenance task allocation set is evaluated by comprehensively considering the balance of maintenance personnel, tools, and maintenance tasks. The solution that achieves the best balance among these factors is identified and is considered the optimal maintenance task allocation solution. For example, if a solution has minimal variance in personnel task allocation, tool usage, and maintenance task execution time, and achieves more efficient resource utilization and maintenance process efficiency than other solutions, then it is considered the optimal maintenance task allocation solution.

[0148] In one embodiment, the evaluation may be based on the balance index, specifically: For each solution in the maintenance task allocation solution set, define a balance indicator for the personnel task allocation. For example, assume the set of maintenance personnel is {P1, P2, …, Pm}, and each person Pi is assigned a workload of Wi. Calculate the balance indicator for the personnel task allocation: E1 = √(∑(Wi - W_avg)^2 / m), where W_avg is the average workload. The smaller the value of E1, the more balanced the personnel task allocation.

[0149] Define a tool usage balance indicator. Let the tool set be {T1, T2, …, Tn}, and the usage frequency of each tool Tj be Fj. Calculate the tool usage balance indicator E2 = √(∑(Fj - F_avg)^2 / n), where F_avg is the average usage frequency. The smaller the E2 value, the more balanced the tool usage.

[0150] Define a maintenance task execution time balance indicator. Assume the set of maintenance tasks is {M1, M2, …, Mk}, and the execution time of each task Mj is Tj. Calculate the maintenance task execution time balance indicator E3 = √(∑(Tj - T_avg)^2 / k), where T_avg is the average execution time. The smaller the E3 value, the more balanced the maintenance task execution time.

[0151] For each maintenance task allocation scheme, calculate the comprehensive balance index E = w1 × E1 + w2 × E2 + w3 × E3, where w1, w2, and w3 are weights that can be set based on the company's emphasis on balance among personnel, tools, and tasks. The maintenance task allocation scheme with the smallest comprehensive balance index E is selected as the optimal maintenance task allocation scheme.

[0152] The above-mentioned evaluation technology implementation method based on balance indicators evaluates the balance of maintenance task allocation plans through quantitative indicators, and can objectively select the optimal plan.

[0153] In one embodiment, reference Figure 6 , step S50 may specifically include the following steps: Step S501: performing a dynamic coupling analysis between the maintenance time window and the equipment operation status according to the association relationship model to obtain the equipment operation status change characteristics.

[0154] The maintenance window is the period of time during which maintenance operations can be performed during equipment operation. The determination of this time period requires consideration of various factors, such as the equipment's production schedule, operational stability, and the availability of maintenance resources. For example, for some continuously operating equipment, the maintenance window may only be scheduled during scheduled maintenance downtime or during brief breaks between production tasks.

[0155] Dynamic coupling analysis is a method for analyzing the dynamic interrelationships between two or more variables. In this application, the analysis involves analyzing the interplay between maintenance windows and equipment operating status over time. For example, changes in equipment operating status may affect the selection of maintenance windows, and the selection of maintenance windows may in turn affect the equipment's operating status.

[0156] Equipment operating state change characteristics describe the patterns of equipment state changes during operation. These characteristics can include trends in equipment operating parameters (such as temperature, pressure, and speed), the frequency and patterns of equipment failures, and the degradation of equipment performance.

[0157] In one embodiment, information related to the equipment's operating status and maintenance window can be extracted from the association model. For example, the model may include information about suitable maintenance times under different equipment operating conditions, as well as information about the impact of different maintenance windows on the equipment's operating status. If the equipment is operating at high temperatures, the association model may indicate that performing maintenance in this state may increase the risk of equipment failure, and therefore the maintenance window should be avoided as much as possible. Alternatively, if the equipment operates relatively stably during a certain period, the model may suggest that this period is a more appropriate maintenance window. Next, the model analyzes how dynamic changes in the equipment's operating status affect the selection of the maintenance window. For example, as equipment operates over time, certain components may gradually wear out, and the operating temperature may increase. These changes can influence the selection of the maintenance window. If critical components of the equipment are worn to a certain extent, maintenance may need to be scheduled as soon as possible, which may shorten the originally scheduled maintenance window. On the other hand, if the equipment's operating status has been relatively stable for a period of time, the maintenance window may be appropriately extended.

[0158] By analyzing large amounts of equipment operating data and historical maintenance window data (this data can be obtained from equipment monitoring systems, maintenance records, etc.), we can identify characteristics of equipment operating status changes. For example, by analyzing equipment temperature data, we can find that the equipment temperature fluctuates significantly during specific time periods each day, which is a characteristic of equipment operating status changes. Or by analyzing equipment failure records, we can find that the probability of equipment failure increases significantly after a certain period of continuous operation, which is also a manifestation of equipment operating status changes.

[0159] In one embodiment, features can be mined based on dynamic coupling analysis using data mining. Specifically, equipment operating status data (such as historical data on parameters like temperature, pressure, and speed) and historical data on maintenance time windows are collected as analysis samples. Data mining techniques, such as association rule mining algorithms, are used to mine this data. The goal is to identify association rules between equipment operating status parameters and maintenance time windows. For example, a rule such as "When the equipment temperature exceeds threshold A, the maintenance time window should be shortened to [specific time period]" can be mined. Based on the mined association rules, the changing trends of the equipment operating status and the impact of these changes on the maintenance time window are analyzed. For example, if the equipment temperature shows a recent upward trend, the association rules indicate that the maintenance time window may need to be adjusted accordingly. By analyzing the equipment operating status data in multiple dimensions (e.g., considering changes in parameters like temperature, pressure, and speed), characteristics of equipment operating status changes are summarized. For example, if it is found that the equipment temperature and pressure increase simultaneously while the speed decreases, the risk of equipment failure increases significantly. This is a comprehensive characteristic of equipment operating status changes. This data mining-based technical implementation method can automatically discover the hidden relationship between equipment operating status and maintenance time window from a large amount of data, which helps to accurately analyze the changing characteristics of equipment operating status.

[0160] Step S502: Based on the characteristics of the equipment operation status change, a time series prediction algorithm is used to perform maintenance time window fluctuation analysis to obtain a maintenance time window adjustment interval.

[0161] A time series forecasting algorithm uses the chronological order of historical data to predict future data trends. In this step, it uses historical data on equipment operating status changes to predict fluctuations in maintenance windows. Common time series forecasting algorithms include the Autoregressive Moving Average (ARMA) model and the Autoregressive Inverse Moving Average (ARIMA) model.

[0162] Maintenance window fluctuation analysis analyzes the fluctuations in the maintenance window as the equipment's operating status changes. Because equipment operating status is not fixed, the maintenance window must be adjusted accordingly. This analysis aims to determine the possible adjustment range of the maintenance window to accommodate changes in equipment operating status.

[0163] The maintenance window adjustment interval is a range that indicates the possible adjustment range of the maintenance window. For example, if the original maintenance window is [9:00 AM - 5:00 PM], after fluctuation analysis, the maintenance window adjustment interval may be [8:00 AM - 6:00 PM], indicating that the maintenance window may be adjusted within this range.

[0164] In one embodiment, the characteristic data of equipment operating state changes is organized chronologically to form time series data. For example, if the characteristics of equipment operating state changes include changes in parameters such as equipment temperature and pressure, the values ​​of these parameters at different time points are arranged chronologically to form temperature time series, pressure time series, and so on. Then, an appropriate time series prediction algorithm is selected to analyze this time series data. Taking the ARIMA model as an example, the data must first be tested for stationarity. If the data is not stationary, differencing is performed to stabilize it. The model order (p, d, q) is then determined, where p is the autoregressive order, d is the differencing order, and q is the moving average order. The parameters of the ARIMA model are obtained by fitting the historical data.

[0165] Using the resulting time series prediction model, the equipment's operating status over the next period of time can be predicted. For example, the changes in equipment parameters such as temperature and pressure can be predicted over the next week. Based on these predictions, the maintenance window can be analyzed for potential impacts. If the temperature of the equipment is predicted to rise to a level that could affect normal operation within a certain period of time, the maintenance window may need to be advanced or shortened. If the equipment is predicted to operate relatively stably within a certain period of time, the maintenance window may be appropriately extended.

[0166] By comprehensively considering the prediction results of various parameters of the equipment's operating status, the maintenance window adjustment range is determined. This range must take into account both changes in the equipment's operating status and practical factors such as the company's production plan and the availability of maintenance resources. ARIMA-based technical implementation can provide relatively accurate predictions of changes in equipment's operating status, providing a basis for determining the maintenance window adjustment range.

[0167] Step S503: constructing a priority ranking function of the maintenance plan according to the maintenance time window adjustment interval to obtain a maintenance time window adjustment strategy.

[0168] The maintenance plan prioritization function is a mathematical function used to determine the priority of maintenance tasks. It prioritizes different maintenance tasks based on the maintenance window adjustment interval and other relevant factors, such as the severity of the equipment failure and the availability of maintenance resources. For example, if a maintenance task has a short maintenance window adjustment interval and a serious equipment failure, it will be given a higher priority in the maintenance plan.

[0169] A maintenance time window adjustment strategy is a strategy for adjusting maintenance plans based on the maintenance time window adjustment interval. It includes determining which maintenance tasks should be prioritized, how to adjust the order of maintenance tasks, and whether to adjust the allocation of maintenance resources.

[0170] In one embodiment, first, the length of the maintenance time window adjustment interval is considered. If the maintenance time window adjustment interval of a maintenance task is short, it means that the maintenance task needs to be arranged as soon as possible, otherwise the best maintenance opportunity may be missed. For example, for a key component failure of a device, its maintenance time window adjustment interval may be only one day, so this maintenance task should be in a higher position in the priority sorting. Secondly, the severity of the equipment failure is taken into consideration. Even if the maintenance time window adjustment interval of two maintenance tasks is the same, if one of the failures will cause the equipment to shut down immediately, while the other failure will only affect part of the performance of the equipment, then the maintenance task that causes the equipment to shut down has a higher priority. This can be reflected by assigning different weights to the severity of the equipment failure. For example, the weight of the equipment shutdown failure is 0.8, and the weight of the partial performance impact failure is 0.2.

[0171] The availability of maintenance resources must also be considered. If the resources required for a maintenance task are unavailable within a certain time period, the priority of the task may need to be adjusted, even if the task has a short window adjustment interval and the fault severity is high. For example, if a maintenance task requires a specific tool that will not be available for the next three days, the priority of the task may need to be lowered, or the allocation of maintenance resources may need to be adjusted to meet the needs of the task.

[0172] Based on these factors, a maintenance plan priority ranking function is constructed. For example, let the length of the maintenance time window adjustment interval for maintenance task i be Ti, the equipment failure severity be Si, and the maintenance resource availability be Ri. Then, the maintenance plan priority ranking function can be expressed as Pi = w1 * (1 / Ti) + w2 * Si + w3 * Ri, where w1, w2, and w3 are weights that can be set based on the company's actual situation. By calculating the Pi value for each maintenance task, the maintenance tasks are prioritized, thereby deriving a maintenance time window adjustment strategy.

[0173] In one embodiment, a priority ranking function based on weighted multiple factors may be constructed, for example: Identify factors that influence maintenance plan priorities, such as maintenance window adjustment intervals, equipment failure severity, and maintenance resource availability.

[0174] Quantify each factor. For the maintenance window adjustment interval, the reciprocal of the interval length can be used to indicate its urgency. For the equipment failure severity, a score can be assigned based on the impact of the failure on equipment operation (e.g., 0-10). For maintenance resource availability, the availability of the resource can be quantified based on its availability (e.g., the closer the availability time, the higher the value).

[0175] The weights w1, w2, and w3 are determined based on the importance the company places on each factor. For example, if the company prioritizes the urgency of the repair window, w1 can be set higher; if the company prioritizes the severity of equipment failures, w2 can be set higher.

[0176] Construct the maintenance plan priority ranking function Pi = w1 * (1 / Ti) + w2 * Si + w3 * Ri, where Ti is the length of the maintenance time window adjustment interval, Si is the quantitative value of the equipment failure severity, and Ri is the quantitative value of the maintenance resource availability.

[0177] Prioritize maintenance tasks based on the calculated Pi value and determine a maintenance time window adjustment strategy. For example, schedule maintenance tasks from highest to lowest Pi value, prioritizing those with higher Pi values. For maintenance tasks with lower Pi values, adjust their maintenance time windows or resource allocation based on actual conditions.

[0178] The above-mentioned technical implementation method based on multi-factor weighting can comprehensively consider multiple factors and reasonably construct a priority ranking function of the maintenance plan, thereby obtaining an effective maintenance time window adjustment strategy.

[0179] Step S504: performing overall optimization of the maintenance plan according to the maintenance time window adjustment strategy, the optimal maintenance path, and the optimal maintenance task allocation plan to obtain an optimal maintenance strategy.

[0180] Maintenance plan optimization is a comprehensive process that considers maintenance window adjustment strategies, optimal maintenance paths, and optimal maintenance task allocation to comprehensively optimize the entire maintenance plan. The goal is to develop an optimal maintenance plan that meets equipment maintenance requirements and resource constraints, thereby improving maintenance efficiency, reducing maintenance costs, and minimizing equipment downtime.

[0181] The optimal maintenance strategy is a comprehensive equipment maintenance plan derived from the coordinated optimization of maintenance plans. It includes the optimal maintenance schedule, maintenance task execution sequence, and maintenance resource allocation. It is the most beneficial strategy for equipment maintenance, derived after comprehensive consideration of various factors.

[0182] In an embodiment of the present application, the execution order of maintenance tasks can be determined based on the maintenance time window adjustment strategy. Maintenance tasks are arranged in descending order of priority to ensure that urgent and important maintenance tasks are handled first within the maintenance time window. For example, if the maintenance task for a core component failure of a device has the highest priority, then this task will be scheduled first in the maintenance plan.

[0183] Then, the specific operations for each maintenance task are optimized based on the optimal maintenance path. The optimal maintenance path specifies the order of the various operations within a maintenance task. During the overall optimization of the maintenance plan, it is important to ensure that each maintenance task follows the optimal maintenance path. For example, if the optimal maintenance path indicates that repairing a fault requires inspection followed by repair operations, then the maintenance operations should be arranged in this order in the maintenance plan.

[0184] At the same time, maintenance resources are allocated according to the optimal maintenance task allocation plan. This ensures that each maintenance task has sufficient maintenance resources to support it and that maintenance resources are used appropriately. For example, if the optimal maintenance task allocation plan specifies that a maintenance task requires maintenance personnel with specific skills and specific maintenance tools, then the maintenance plan should arrange for the corresponding personnel and tools to participate in the maintenance task at the appropriate time.

[0185] By comprehensively optimizing the execution sequence of maintenance tasks, the maintenance operation sequence, and the allocation of maintenance resources, we can obtain the optimal maintenance strategy. This strategy can maximize maintenance efficiency, reduce maintenance costs, and minimize equipment downtime while meeting equipment maintenance requirements and maintenance resource constraints.

[0186] In one embodiment, the maintenance plan can be comprehensively optimized based on integer programming, for example: Assume that the maintenance task set is {T1, T2, …, Tn} and define the decision variable xij. If the maintenance task Ti is executed in the jth time unit, then xij = 1, otherwise xij = 0.

[0187] According to the maintenance time window adjustment strategy, the earliest start time Ei and the latest end time Li of each maintenance task Ti are determined. This can be obtained by analyzing factors such as the priority of the maintenance task and the maintenance time window adjustment interval.

[0188] According to the optimal maintenance path, the order constraints of each operation in each maintenance task Ti are determined. For example, if operation O1 in maintenance task T1 must be performed before operation O2, then the corresponding constraint conditions can be obtained.

[0189] Based on the optimal maintenance task allocation plan, the number of maintenance resources required for each maintenance task Ti is determined. Combined with the availability of maintenance resources, the maintenance resource constraints are obtained. For example, if maintenance task T1 requires two maintenance personnel with specific skills, but the company only has three such personnel available, then the personnel resource constraints can be obtained.

[0190] Construct an objective function, such as minimizing the total maintenance time, minimizing the maintenance cost, or maximizing equipment availability. For example, if the goal is to minimize the total maintenance time, you can construct the objective function Z = ∑∑xij * tij, where tij is the maintenance time when the maintenance task Ti is executed in the jth time unit.

[0191] According to the above decision variables, constraints and objective functions, an integer programming model is established.

[0192] This integer programming model is solved using an integer programming algorithm (such as branch and bound or cutting plane methods). The optimal solution is the optimal maintenance strategy. This optimal solution determines the execution time of each maintenance task, the maintenance operation sequence, and the allocation of maintenance resources. Integer programming-based technical implementations can achieve the optimal maintenance strategy through optimization algorithms while satisfying various constraints, making it suitable for complex maintenance planning optimization problems.

[0193] In one embodiment, step S504 may be implemented by the following steps: Based on the maintenance time window adjustment strategy, a dynamic programming algorithm is used to perform global optimization analysis to obtain a global optimization result. Based on the global optimization result, the maintenance time window adjustment strategy, the optimal maintenance path, and the optimal maintenance task allocation plan, a maintenance plan is comprehensively optimized to obtain an optimal maintenance plan and an optimal maintenance strategy.

[0194] The maintenance window adjustment strategy is based on factors such as the maintenance window adjustment interval, equipment fault severity, and maintenance resource availability. It determines the priority of maintenance tasks and how to adjust the maintenance window. It provides basic guidance for optimizing maintenance plans, such as which maintenance tasks should be prioritized and within what timeframe.

[0195] Dynamic programming is an algorithm used to solve optimization problems in multi-stage decision-making processes. It breaks down a complex problem into a series of interconnected subproblems and solves these subproblems to obtain the global optimal solution. In this step, it uses the maintenance time window adjustment strategy as a foundation, considers each stage of the maintenance process (such as the sequencing of different maintenance tasks and the changes in resource allocation at different stages), and achieves global optimization through step-by-step analysis and calculation.

[0196] The global optimization result is obtained by comprehensively analyzing and optimizing the entire maintenance plan using a dynamic programming algorithm. This result comprehensively considers various maintenance task constraints and objectives (such as minimizing maintenance time, minimizing costs, and maximizing equipment availability). It contains optimal decision-making information regarding maintenance task execution sequence, scheduling, resource allocation, and other aspects.

[0197] In one embodiment, the problem is first divided into stages based on the maintenance time window adjustment strategy. For example, maintenance tasks are sorted from high to low priority, and each maintenance task can be considered as a stage. Within each stage, the execution conditions of the maintenance task need to be considered, such as whether the required maintenance resources are available and whether the task is within the maintenance time window.

[0198] Then, determine the status of each phase. This status can include the equipment's current operating status, completed maintenance tasks, and remaining maintenance resources. For example, at a certain phase, the equipment may be in an operational test state after partial fault repair, with maintenance tasks for some key components completed and a specific number of maintenance personnel and tools available.

[0199] Next, define the state transition equation. This equation describes how a state transitions from one phase to another. For example, if a maintenance task is completed in the current phase, the equipment's operating status may improve, and the remaining maintenance resources will decrease accordingly. This is a state transition. The state transition equation needs to consider various rules in the maintenance time window adjustment strategy, such as executing high-priority maintenance tasks first and allocating maintenance resources appropriately.

[0200] Using a dynamic programming algorithm, starting from the initial state, the optimal decision is calculated for each state in a sequential order. For example, at each stage, the cost and time required to perform different maintenance tasks in the current state are calculated, and the optimal sequence for executing these tasks is selected, ensuring that the entire maintenance process reaches a global optimum while satisfying various constraints. The final result is a global optimization result, which includes the optimal decisions at each stage, such as the optimal execution time for each maintenance task and the optimal allocation of maintenance resources.

[0201] Since the global optimization results provide preliminary optimization information on maintenance task execution sequence, time arrangement and resource allocation, further overall optimization is carried out based on this information by combining the maintenance time window adjustment strategy, optimal maintenance path and optimal maintenance task allocation plan. Specifically: Fine-tune the maintenance task execution time in the global optimization results based on the maintenance time window adjustment strategy. For example, if the global optimization results determine the execution time of a maintenance task, the maintenance time window adjustment strategy may require moving this execution time forward or backward to account for changes in the equipment's operating status or temporary adjustments to maintenance resources.

[0202] Optimize the order of maintenance tasks based on the optimal maintenance path. While the global optimization results determine the approximate order of maintenance tasks, the optimal maintenance path provides a more detailed order within each maintenance task. For example, for a maintenance task on a piece of equipment, the global optimization results determine that it should be performed within a certain time period, while the optimal maintenance path indicates that inspection operations within this maintenance task should be performed before repair operations. Therefore, this order of operations must be clearly defined in the maintenance plan.

[0203] At the same time, maintenance resources are allocated according to the optimal maintenance task allocation plan. This ensures that each maintenance task has sufficient and appropriate maintenance resources to support its execution. For example, if the optimal maintenance task allocation plan stipulates that a maintenance task requires maintenance personnel with specific skills and specific maintenance tools, then the maintenance plan should clearly specify the allocation of these resources at the appropriate time and location.

[0204] Through comprehensive adjustment and optimization of these aspects, the optimal maintenance plan is obtained. This plan achieves the optimal combination of maintenance task execution time, operation sequence and resource allocation based on the satisfaction of various constraints.

[0205] Furthermore, the optimal maintenance plan can be combined with other equipment maintenance-related strategies (such as equipment operating status monitoring strategies and preventive maintenance strategies) to form an optimal maintenance strategy. For example, the optimal maintenance strategy can stipulate specific equipment operating status monitoring before and after maintenance tasks, or formulate preventive maintenance plans based on the equipment's operating history and maintenance status, such as regularly replacing wearing parts, to ensure the long-term stable operation of the equipment.

[0206] In one embodiment, reference Figure 7 , step S60 may specifically include the following steps: Step S601: Based on the optimal maintenance strategy, multi-objective parameter optimization is performed to construct a maintenance decision optimization space including maintenance paths, maintenance task allocation, and maintenance time windows.

[0207] Multi-objective parameter optimization involves optimizing parameters related to multiple objectives. In the context of industrial equipment maintenance, factors such as maintenance routes, maintenance task allocation, and maintenance time windows all involve multiple parameters. For example, each operation step in a maintenance route may involve different parameters such as operation time and resource requirements; maintenance task allocation involves parameters such as the skill level of different maintenance personnel and task priority; and maintenance time windows involve parameters such as start time and duration. Multi-objective parameter optimization aims to comprehensively consider these parameters to optimize overall maintenance decisions.

[0208] The maintenance decision optimization space is an abstract conceptual space encompassing all possible combinations of maintenance paths, maintenance task assignments, and maintenance time windows, along with their associated parameters. For example, for a complex piece of automated production equipment, there may be multiple maintenance path options, each with a different operation sequence and method, corresponding to different maintenance task assignment schemes (e.g., assigning different tasks to different maintenance personnel) and maintenance time windows (maintenance during different time periods has different impacts on production). All these possible combinations constitute the maintenance decision optimization space.

[0209] In one embodiment, a detailed parameter analysis is first performed on the maintenance path, maintenance task allocation, and maintenance time window in the optimal maintenance strategy. Taking the maintenance path as an example, it is broken down into multiple operational steps, and parameters such as the operation time, required tools, and required personnel skills are determined for each step. For maintenance task allocation, parameters such as the number of personnel, skill level, and working hours required for each task are determined. For the maintenance time window, parameters such as its adjustable range and the coefficient of impact on production are determined. Next, a maintenance decision optimization space is constructed using mathematical modeling tools (such as linear programming or nonlinear programming, depending on the actual situation). These parameters are used as variables, and the objective function is set to comprehensively optimize these parameters (e.g., minimizing total cost, maximizing equipment availability, etc.), with constraints including actual equipment operating requirements and personnel resource limitations. This constructs a maintenance decision optimization space encompassing all possible decision scenarios. This approach comprehensively considers various factors related to maintenance decisions, providing a complete framework for subsequent optimization constraints, enabling optimization within a reasonable range and avoiding overlooking important factors.

[0210] Step S602: According to the maintenance decision optimization space, a preset maintenance cost threshold and a preset maintenance efficiency threshold are combined to perform collaborative constraints to obtain a first maintenance decision combination.

[0211] A maintenance cost threshold is a pre-set numerical limit that limits the range of maintenance costs. For example, for a specific piece of industrial equipment, if a company, based on past maintenance experience and budget planning, determines that the cost of each repair cannot exceed a certain amount, this amount is the maintenance cost threshold. This amount can be determined based on factors such as the value of the equipment, the company's financial situation, and market prices for maintenance.

[0212] A maintenance efficiency threshold is a pre-set minimum standard for measuring maintenance efficiency. Maintenance efficiency can be measured in various ways, such as repair time and the ratio of equipment downtime to normal operation time. For example, if an enterprise requires that equipment repair time after a failure must not exceed a certain time limit, the maintenance efficiency indicator corresponding to this time limit is the maintenance efficiency threshold.

[0213] Collaborative constraints are the process of simultaneously applying maintenance cost and efficiency thresholds to constrain the various possible decision options within the maintenance decision optimization space. This means that while maintenance costs must not exceed a set threshold, maintenance efficiency must also meet or exceed a pre-set threshold. This constraint approach doesn't consider either cost or efficiency in isolation; instead, it combines the two to ensure that the resulting maintenance decision is both economical and efficient. For example, when evaluating a maintenance option, one shouldn't simply choose it because it has a low cost but a long maintenance time (below the maintenance efficiency threshold), nor should one simply accept it because it is fast but costly (exceeding the maintenance cost threshold). A decision option must be found that satisfies both cost and efficiency requirements.

[0214] In one embodiment, within the maintenance decision optimization space, the maintenance cost and efficiency of each possible decision solution (i.e., a combination of maintenance path, maintenance task assignment, and maintenance time window) can be calculated. The maintenance cost can be calculated by accumulating factors such as the material cost, labor cost, and indirect costs associated with equipment downtime for each operation step. The maintenance efficiency can be comprehensively evaluated based on factors such as equipment downtime, maintenance time, and performance improvement after the equipment resumes normal operation. The calculated maintenance cost and efficiency are then compared with preset maintenance cost and efficiency thresholds. Decision solutions whose maintenance costs do not exceed the cost thresholds and whose maintenance efficiencies meet or exceed the efficiency thresholds are selected. These solutions constitute the first maintenance decision combination. This technical implementation method, by setting clear cost and efficiency boundaries, can quickly select maintenance decision combinations that initially meet the enterprise's requirements, reducing the workload of subsequent optimization calculations while ensuring that maintenance decisions meet the enterprise's basic requirements in terms of cost and efficiency.

[0215] Step S603: Initialize a multi-objective optimization algorithm based on the first maintenance decision combination; perform maintenance decision optimization iterations based on the multi-objective optimization algorithm to obtain a second maintenance decision combination, and record the number of iterations.

[0216] Multi-objective optimization algorithms are specifically designed to handle simultaneous optimization problems involving multiple objective functions. In industrial equipment maintenance decisions, multiple objectives, such as maintenance cost, maintenance efficiency, and equipment reliability, must be considered simultaneously. Multi-objective optimization algorithms can balance and optimize these objectives. For example, common multi-objective evolutionary algorithms (such as NSGA-II) mimic biological evolution to continuously search for optimal decision solutions within the constraints of multiple objective functions.

[0217] Take the NSGA-II algorithm as an example to illustrate.

[0218] First, the algorithm is initialized based on the first maintenance decision combination. Each maintenance decision solution in the first maintenance decision combination is considered an individual, and these individuals together constitute the initial population of the NSGA-II algorithm. For example, if there are five different maintenance decision solutions in the first maintenance decision combination, then these five solutions will serve as the five individuals in the initial population.

[0219] Next, define the fitness function. The fitness function is a mathematical expression used to evaluate the quality of each individual (i.e., maintenance decision plan). Fitness functions are defined for the three objectives of maintenance cost, equipment reliability, and production efficiency. For the maintenance cost objective, the fitness function can be defined as the inverse of the maintenance cost, that is, the lower the maintenance cost, the higher the fitness. For the equipment reliability objective, the fitness function can be defined based on the inverse of the equipment failure probability or the equipment's trouble-free operating time. The higher the equipment reliability, the higher the fitness. For the production efficiency objective, the fitness function can be defined based on the ratio of the equipment's normal operating time to the total time or the production output per unit time. The higher the production efficiency, the higher the fitness.

[0220] Then, iterative calculations are performed according to the basic process of the NSGA-II algorithm. During each iteration, a new maintenance decision plan (new individual) is generated through selection (selecting the best individual based on its fitness), crossover (exchanging some information between the selected individuals to generate new individuals), and mutation (randomly changing certain parameters of the individuals to generate new individuals). The fitness function value of the new individual is calculated, and the population is updated based on the non-dominated sorting and crowding calculations. The number of iterations is incremented by 1 after each iteration.

[0221] After multiple iterations, a new set of maintenance decision schemes are obtained, which constitute the second maintenance decision combination.

[0222] This approach leverages the mature NSGA-II algorithm, a multi-objective optimization algorithm, to effectively balance and search across multiple objectives. By properly defining a fitness function that accurately reflects the optimization direction of each objective, iteratively generates and selects optimal maintenance decision solutions, thereby improving the overall performance of decision-making and gradually achieving a better balance among multiple objectives such as maintenance costs, equipment reliability, and production efficiency.

[0223] Step S604: If the second maintenance decision combination is better than the first maintenance decision combination, the second maintenance decision combination is used as the first maintenance decision combination; if not, the second maintenance decision combination is discarded.

[0224] The second maintenance decision combination is superior to the first maintenance decision combination if, under a comprehensive multi-objective evaluation, the second maintenance decision combination performs better overall in terms of maintenance cost, maintenance efficiency, equipment reliability, and other objectives. For example, if the second maintenance decision combination improves maintenance efficiency and equipment reliability without increasing maintenance cost, or significantly improves maintenance efficiency and equipment reliability while slightly increasing maintenance cost, then the second maintenance decision combination is considered superior to the first maintenance decision combination.

[0225] In one embodiment, the values ​​of maintenance cost, maintenance efficiency, equipment reliability, and other objectives can be recalculated for each decision solution in the second maintenance decision combination and the first maintenance decision combination. A comprehensive evaluation is then performed using a multi-attribute decision-making method (such as the TOPSIS method). The TOPSIS method calculates the distance between each decision solution and the ideal solution (the optimal value of each objective) and the negative ideal solution (the worst value of each objective) to obtain a comprehensive evaluation index. The comprehensive evaluation index of each decision solution in the second maintenance decision combination is then compared with that of the first maintenance decision combination. If the comprehensive evaluation index of most decision solutions in the second maintenance decision combination is better than that of the corresponding decision solutions in the first maintenance decision combination, the second maintenance decision combination is considered superior to the first maintenance decision combination and the second maintenance decision combination replaces the first maintenance decision combination. Otherwise, the second maintenance decision combination is discarded. This approach, through scientific evaluation methods, can accurately determine the pros and cons of two decision combinations, ensuring that maintenance decisions are optimized.

[0226] Step S605: Continuing to the next iteration, when the number of iterations is greater than or equal to the preset number of iterations, outputting the current first maintenance decision combination as the final maintenance decision.

[0227] The preset number of iterations is a value set before the algorithm begins and is used to determine the stopping condition for the multi-objective optimization algorithm. This value can be determined based on factors such as the complexity of the problem, computing resource limitations, and the required decision accuracy. For example, for a complex industrial equipment maintenance decision problem, if a more precise answer is desired, a higher preset number of iterations might be used; for a relatively simple problem or with limited computing resources, a lower preset number of iterations might be used.

[0228] In one embodiment, a counter can be set within a multi-objective optimization algorithm (such as the NSGA-II algorithm) to record the number of iterations. Each time an iteration is completed, the counter is incremented by 1. After each iteration, the counter value is checked to see if it is greater than or equal to a preset number of iterations. If the preset number of iterations has not been reached, the next iteration is continued, and the maintenance decision combination is updated according to the methods in steps S603 and S604. If the number of iterations has reached or exceeded the preset number of iterations, the iteration is terminated, and the current first maintenance decision combination is output as the final maintenance decision. This technical implementation achieves the following technical benefits: by setting a reasonable preset number of iterations, computational costs can be controlled while ensuring decision quality, avoiding overcomputation and ultimately obtaining a final maintenance decision that strikes a good balance between multiple objectives.

[0229] Accordingly, in order to better implement the above method, the embodiment of the present application also provides an intelligent maintenance decision system based on knowledge graph. Figure 8 As shown, the knowledge graph-based intelligent maintenance decision system 80 includes an acquisition module 801, a graph construction module 802, a relationship analysis module 803, an optimization module 804, a window adjustment module 805, and a decision module 806, as follows: Acquisition module 801, used to acquire equipment operating status data and historical maintenance data of industrial equipment; A graph construction module 802 is used to construct an industrial knowledge graph based on the equipment operating status data and the historical maintenance data, and to generate an association relationship model between the equipment operating status and the maintenance behavior based on the industrial knowledge graph; a relationship analysis module 803 is used to analyze the relationship between the fault type and the maintenance path based on the association relationship model and the equipment operating status data to obtain the optimal maintenance path; an optimization module 804 is used to optimize the maintenance task allocation based on the optimal maintenance path, the association relationship model and the preset maintenance resource constraints to obtain the optimal maintenance task allocation plan; a window adjustment module 805 is used to perform dynamic adjustment analysis of the maintenance time window based on the optimal maintenance path, the optimal maintenance task allocation plan and the association relationship model to obtain the optimal maintenance strategy; a decision module 806 is used to perform multi-objective optimization iterative analysis based on the optimal maintenance strategy to obtain the final maintenance decision for the industrial equipment.

[0230] In one embodiment, the graph construction module 802 is configured to: Constructing nodes and edges of an initial knowledge graph based on the equipment operating status data and the historical maintenance data to obtain a preliminary knowledge graph structure; Based on the preliminary knowledge graph structure, calculate the relationship weights between nodes to obtain the association strength distribution of the initial knowledge graph; The preliminary knowledge graph structure is enhanced according to the association strength distribution to obtain an industrial knowledge graph.

[0231] In one embodiment, the graph construction module 802 is used to: map the association strength distribution and the preliminary knowledge graph structure to a low-dimensional space using a graph embedding algorithm to obtain an enhanced knowledge graph representation; and optimize the node grouping and edge weights using a hierarchical clustering method based on the enhanced knowledge graph representation to obtain a correlation relationship model between the equipment operating status and the maintenance behavior.

[0232] In one embodiment, the relationship analysis module 803 is configured to: Analyzing the dynamic response relationship between the probability of occurrence of a fault type and the execution order of a maintenance path based on the association relationship model and the equipment operating status data to obtain an impact weight of the fault type on the maintenance path; Analyzing the historical matching degree matrix between the fault type and the maintenance path based on the impact weight and the association relationship model to obtain path selection constraints; Obtain candidate maintenance paths according to the path selection constraint conditions to obtain a set of candidate maintenance paths; An optimal maintenance path is determined according to the candidate maintenance path set.

[0233] In one embodiment, the relationship analysis module 803 is used to: calculate the execution efficiency index of each path based on the candidate maintenance path set to obtain a path execution efficiency ranking; when the path execution efficiency ranking meets a preset efficiency threshold, determine the candidate maintenance path with the highest ranking as the optimal maintenance path; when the path execution efficiency ranking does not meet the preset efficiency threshold, perform local search optimization on the candidate maintenance path set to obtain the optimal maintenance path.

[0234] In one embodiment, the optimization module 804 is configured to: Based on the association relationship model, a nonlinear relationship analysis is performed between maintenance task allocation and maintenance resource utilization to obtain a nonlinear relationship model of maintenance resources; based on the nonlinear relationship model of maintenance resources and the optimal maintenance path, function decomposition is performed in combination with preset maintenance resource constraints to obtain a maintenance task allocation optimization function; based on the maintenance task allocation optimization function, a particle swarm optimization algorithm is used to perform iterative calculations to obtain a maintenance task allocation solution set that satisfies the maintenance resource constraints; based on the maintenance task allocation solution set and the nonlinear relationship model of maintenance resources, the balance of maintenance task allocation is analyzed, and the solution with the best balance in the maintenance task allocation solution set is determined to be the optimal maintenance task allocation solution.

[0235] In one embodiment, the window adjustment module 805 is configured to: Based on the association model, a dynamic coupling analysis of the maintenance time window and the equipment operating status is performed to obtain the characteristics of equipment operating status changes. Based on the characteristics of equipment operating status changes, a time series prediction algorithm is used to perform a maintenance time window fluctuation analysis to obtain a maintenance time window adjustment interval. Based on the maintenance time window adjustment interval, a priority ranking function for the maintenance plan is constructed to obtain a maintenance time window adjustment strategy. The maintenance plan is comprehensively optimized according to the maintenance time window adjustment strategy, the optimal maintenance path and the optimal maintenance task allocation plan to obtain the optimal maintenance strategy.

[0236] In one embodiment, the window adjustment module 805 is configured to: Based on the maintenance time window adjustment strategy, a dynamic programming algorithm is used to perform global optimization analysis to obtain a global optimization result. Based on the global optimization result, the maintenance time window adjustment strategy, the optimal maintenance path, and the optimal maintenance task allocation plan, a maintenance plan is comprehensively optimized to obtain an optimal maintenance plan and an optimal maintenance strategy.

[0237] In one embodiment, the decision module 806 is used to According to the optimal maintenance strategy, multi-objective parameter optimization is performed to construct a maintenance decision optimization space including maintenance paths, maintenance task allocation and maintenance time windows; according to the maintenance decision optimization space, a preset maintenance cost threshold and a preset maintenance efficiency threshold are combined for collaborative constraints to obtain a first maintenance decision combination; according to the first maintenance decision combination, a multi-objective optimization algorithm is initialized; according to the multi-objective optimization algorithm, maintenance decision optimization iterations are performed to obtain a second maintenance decision combination, and the number of iterations is recorded; if the second maintenance decision combination is better than the first maintenance decision combination, the second maintenance decision combination is used as the first maintenance decision combination; if it is not better than the first maintenance decision combination, the second maintenance decision combination is discarded; continue to the next iteration, and when the number of iterations is greater than or equal to the preset number of iterations, the current first maintenance decision combination is output as the final maintenance decision.

[0238] The implementation of each of the above modules can be specifically referred to the above method embodiments, which will not be described in detail here. The technical effects achieved by each module and device can be referred to the description of the above method embodiments.

[0239] It should be noted that, in specific implementations, the above modules can be arbitrarily combined, integrated into one or more modules, or implemented as independent entities. Furthermore, the above modules can be implemented in the form of hardware or software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. The aforementioned storage medium can be a read-only memory, a magnetic disk, or an optical disk, etc.

[0240] like Figure 9 As shown, an embodiment of the present application further provides a computer device 90, characterized in that it includes a processor 901 and a memory 902, wherein the memory 902 stores a computer program, and when the computer program is executed by the processor 901, the processor 901 performs the steps of any of the methods described above.

[0241] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0242] The above disclosure is only a preferred embodiment of the present application, and certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. An intelligent maintenance decision-making method based on knowledge graph, characterized in that: include: Obtain equipment operating status data and historical maintenance data of industrial equipment; Constructing an industrial knowledge graph based on the equipment operation status data and the historical maintenance data, and generating a correlation relationship model between the equipment operation status and maintenance behavior based on the industrial knowledge graph; Analyzing the relationship between the fault type and the maintenance path based on the association relationship model and the equipment operating status data to obtain the optimal maintenance path; Optimizing maintenance task allocation according to the optimal maintenance path, the association relationship model, and preset maintenance resource constraints to obtain an optimal maintenance task allocation plan; According to the optimal maintenance path, the optimal maintenance task allocation plan and the association relationship model, a dynamic adjustment analysis of the maintenance time window is performed to obtain an optimal maintenance strategy; Based on the optimal maintenance strategy, a multi-objective optimization iterative analysis is performed to obtain a final maintenance decision for the industrial equipment.

2. The intelligent maintenance decision-making method based on knowledge graph according to claim 1 is characterized in that: The constructing of an industrial knowledge graph based on the equipment operating status data and the historical maintenance data includes: Constructing nodes and edges of an initial knowledge graph based on the equipment operating status data and the historical maintenance data to obtain a preliminary knowledge graph structure; Based on the preliminary knowledge graph structure, calculate the relationship weights between nodes to obtain the association strength distribution of the initial knowledge graph; The preliminary knowledge graph structure is enhanced according to the association strength distribution to obtain an industrial knowledge graph.

3. The intelligent maintenance decision-making method based on knowledge graph according to claim 2 is characterized in that: The preliminary knowledge graph structure is enhanced according to the association strength distribution to obtain an industrial knowledge graph, including: According to the association strength distribution and the preliminary knowledge graph structure, a graph embedding algorithm is used to map them into a low-dimensional space to obtain an enhanced knowledge graph representation; Generate an association relationship model between equipment operating status and maintenance behavior based on the industrial knowledge graph, including: according to the enhanced knowledge graph representation, use a hierarchical clustering method to optimize node grouping and edge weights to obtain an association relationship model between equipment operating status and maintenance behavior.

4. The intelligent maintenance decision-making method based on knowledge graph according to any one of claims 1 to 3, characterized in that: Analyzing the relationship between the fault type and the maintenance path based on the association relationship model and the equipment operating status data to obtain the optimal maintenance path includes: Analyzing the dynamic response relationship between the probability of occurrence of a fault type and the execution order of a maintenance path based on the association relationship model and the equipment operating status data, and obtaining the impact weight of the fault type on the maintenance path; Analyzing the historical matching matrix between the fault type and the maintenance path according to the influence weight and the association relationship model to obtain the path selection constraint condition; Obtain candidate maintenance paths according to the path selection constraint conditions to obtain a set of candidate maintenance paths; An optimal maintenance path is determined according to the candidate maintenance path set.

5. The intelligent maintenance decision-making method based on knowledge graph according to claim 4 is characterized in that: Determining an optimal maintenance path according to the candidate maintenance path set includes: Calculating the execution efficiency index of each path based on the candidate maintenance path set to obtain a path execution efficiency ranking; When the path execution efficiency ranking satisfies a preset efficiency threshold, the candidate maintenance path with the highest ranking is determined as the optimal maintenance path; when the path execution efficiency ranking does not meet the preset efficiency threshold, a local search optimization is performed on the candidate maintenance path set to obtain the optimal maintenance path.

6. The intelligent maintenance decision-making method based on knowledge graph according to claim 1 is characterized in that: The optimizing maintenance task allocation according to the optimal maintenance path, the association relationship model and the preset maintenance resource constraint condition to obtain the optimal maintenance task allocation scheme includes: According to the association relationship model, a nonlinear relationship analysis is performed between maintenance task allocation and maintenance resource utilization to obtain a maintenance resource nonlinear relationship model; According to the nonlinear relationship model of maintenance resources and the optimal maintenance path, combined with preset maintenance resource constraints, function decomposition is performed to obtain a maintenance task allocation optimization function; According to the maintenance task allocation optimization function, a particle swarm optimization algorithm is used to perform iterative calculations to obtain a maintenance task allocation solution set that meets the maintenance resource constraints; The balance of maintenance task allocation is analyzed according to the maintenance task allocation solution set and the maintenance resource nonlinear relationship model, and the solution with the best balance in the maintenance task allocation solution set is determined as the optimal maintenance task allocation solution.

7. The intelligent maintenance decision-making method based on knowledge graph according to claim 1 is characterized in that: According to the optimal maintenance path, the optimal maintenance task allocation plan and the association relationship model, a dynamic adjustment analysis of the maintenance time window is performed to obtain an optimal maintenance strategy, including: According to the association relationship model, a dynamic coupling analysis of the maintenance time window and the equipment operation status is performed to obtain the equipment operation status change characteristics; According to the characteristics of the equipment operation status change, a time series prediction algorithm is used to perform maintenance time window fluctuation analysis to obtain the maintenance time window adjustment interval; According to the maintenance time window adjustment interval, a priority ranking function of the maintenance plan is constructed to obtain a maintenance time window adjustment strategy; The maintenance plan is comprehensively optimized according to the maintenance time window adjustment strategy, the optimal maintenance path and the optimal maintenance task allocation plan to obtain the optimal maintenance strategy.

8. The intelligent maintenance decision-making method based on knowledge graph according to claim 7 is characterized in that: Performing overall optimization of the maintenance plan based on the maintenance time window adjustment strategy, the optimal maintenance path, and the optimal maintenance task allocation plan to obtain an optimal maintenance strategy includes: According to the maintenance time window adjustment strategy, a dynamic programming algorithm is used to perform global optimization analysis to obtain a global optimization result; According to the global optimization result, the maintenance time window adjustment strategy, the optimal maintenance path and the optimal maintenance task allocation plan, the maintenance plan is comprehensively optimized to obtain the optimal maintenance plan and the optimal maintenance strategy.

9. The intelligent maintenance decision-making method based on knowledge graph according to claim 1 is characterized in that: Based on the optimal maintenance strategy, a multi-objective optimization iterative analysis is performed to obtain the final maintenance decision for the industrial equipment, including: Based on the optimal maintenance strategy, multi-objective parameter optimization is performed to construct a maintenance decision optimization space including maintenance path, maintenance task allocation and maintenance time window; According to the maintenance decision optimization space, a preset maintenance cost threshold and a preset maintenance efficiency threshold are combined to perform collaborative constraints to obtain a first maintenance decision combination; Initializing a multi-objective optimization algorithm according to the first maintenance decision combination; Performing maintenance decision optimization iterations according to the multi-objective optimization algorithm to obtain a second maintenance decision combination, and recording the number of iterations; If the second maintenance decision combination is better than the first maintenance decision combination, the second maintenance decision combination is used as the first maintenance decision combination; if it is not better than the first maintenance decision combination, the second maintenance decision combination is discarded; Continuing to the next iteration, when the number of iterations is greater than or equal to the preset number of iterations, the current first maintenance decision combination is output as the final maintenance decision.

10. An intelligent maintenance decision system based on knowledge graph, characterized in that: include: An acquisition module is used to obtain equipment operating status data and historical maintenance data of industrial equipment; a graph construction module, configured to construct an industrial knowledge graph based on the equipment operation status data and the historical maintenance data, and generate a correlation relationship model between the equipment operation status and the maintenance behavior based on the industrial knowledge graph; A relationship analysis module, configured to analyze the relationship between the fault type and the maintenance path based on the association relationship model and the equipment operation status data, and obtain the optimal maintenance path; An optimization module, configured to optimize the maintenance task allocation according to the optimal maintenance path, the association relationship model, and preset maintenance resource constraints to obtain an optimal maintenance task allocation solution; A window adjustment module is used to perform dynamic adjustment analysis of the maintenance time window based on the optimal maintenance path, the optimal maintenance task allocation plan and the association relationship model to obtain an optimal maintenance strategy; The decision module is used to perform multi-objective optimization iterative analysis based on the optimal maintenance strategy to obtain a final maintenance decision for the industrial equipment.

Images