Intelligent dispatching method and system for hydraulic equipment maintenance based on semantic image and multiple constraints

Abstract

The application provides a method and system for intelligent dispatching of maintenance tasks for hydraulic equipment based on semantic portraits and multiple constraints. The method includes: standardizing multiple source maintenance data to obtain unified maintenance data; based on a hydraulic fault semantic atlas and a domain knowledge enhanced semantic model, fusing and analyzing repair semantics, operation abnormalities and historical maintenance information in the unified maintenance data to generate fault semantic understanding results; generating a maintenance task semantic portrait according to the fault semantic understanding results and the structure relationship of the hydraulic equipment; generating a maintenance resource capability portrait according to historical maintenance feedback, real-time operation status and available resource conditions of the maintenance resources, and determining a candidate maintenance resource set; using a multi-objective intelligent dispatching model to evaluate the candidate maintenance resource set for dispatching adaptation, generating a comprehensive dispatching score, and generating a maintenance task allocation result carrying dispatching basis. The application can improve the accuracy of maintenance task recognition, improve resource matching efficiency, and enhance the optimization capability of dispatching feedback.

Description

Technical Field

[0001] This application relates to the field of intelligent operation and maintenance technology for industrial equipment, and in particular to a method and system for intelligent dispatching of maintenance orders for hydraulic equipment based on semantic profiling and multi-constraints. Background Technology

[0002] Hydraulic equipment is widely used in engineering machinery, metallurgical production lines, port loading and unloading, mining equipment, and industrial automation production scenarios. Its operating status directly affects production continuity and equipment safety. With the development of equipment networking, work order management, and predictive maintenance technologies, some systems can collect operational alarms, sensor data, and maintenance records of hydraulic equipment, and dispatch maintenance tasks based on the reported repair area, personnel availability, or preset fault categories. Existing technologies also attempt to use fault diagnosis models or knowledge bases to identify hydraulic system anomalies to assist maintenance personnel in determining the cause of faults.

[0003] However, hydraulic equipment failures typically exhibit multi-source characteristics, component coupling, and causal transmission. The same failure phenomenon may correspond to different hydraulic components or different maintenance actions. Existing dispatch methods rely heavily on manual experience or simple rules, making it difficult to integrate and understand the semantics of repair requests, operational anomalies, and historical maintenance information. This leads to inaccurate identification of the fault object, complexity, and resource requirements of maintenance tasks. Furthermore, existing systems typically describe maintenance personnel capabilities at the job type, region, or level level, failing to dynamically characterize maintenance resource capabilities by incorporating historical maintenance results, real-time operational status, and available resource conditions. Moreover, the existing dispatch process lacks collaborative optimization of multiple objectives such as skill matching, response timeliness, resource accessibility, and repair reliability. It also struggles to continuously adjust fault perception and dispatch parameters based on maintenance execution feedback, easily resulting in inaccurate dispatch, low response efficiency, and high repeat visit rates. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method and system for intelligent dispatching of maintenance tasks for hydraulic equipment based on semantic profiling and multi-constraints, in order to solve the problems of inaccurate identification of maintenance tasks, suboptimal resource matching, and difficulty in feedback and optimization of dispatching in the prior art.

[0005] A first aspect of this application provides a method for intelligent dispatching of hydraulic equipment maintenance tasks based on semantic profiling and multi-constraints. The method includes: acquiring multi-source maintenance data associated with hydraulic equipment maintenance tasks to be dispatched, and standardizing the multi-source maintenance data to obtain unified maintenance data; fusing and analyzing the reported repair semantics, operational anomalies, and historical maintenance information in the unified maintenance data based on a hydraulic fault semantic graph and a domain knowledge-enhanced semantic model to generate a fault semantic understanding result representing the fault object, suspected fault mode, and maintenance requirements; and generating a result including task urgency, fault complexity, and skills based on the fault semantic understanding result and the structural relationship of the hydraulic equipment. The system generates a semantic profile of maintenance tasks based on demand and resource requirements. It also generates a capability profile of maintenance resources based on historical maintenance feedback, real-time operational status, and available resource conditions. Constraint matching between the semantic profile and the capability profile determines a set of candidate maintenance resources. A multi-objective intelligent dispatch model is used to evaluate the dispatch suitability of the candidate resource set, generating a comprehensive dispatch score. Based on this score, target maintenance resources are determined, and a maintenance task allocation result carrying the dispatch criteria is generated. Finally, maintenance execution feedback is obtained, and the hydraulic fault semantic graph, maintenance resource capability profile, and dispatch decision parameters are updated based on this feedback.

[0006] A second aspect of this application provides an intelligent dispatch system for the maintenance of hydraulic equipment based on semantic profiling and multi-constraints, comprising: an acquisition module for acquiring multi-source maintenance data associated with hydraulic equipment maintenance tasks to be dispatched, and standardizing the multi-source maintenance data to obtain unified maintenance data; a parsing module for fusing and parsing the repair reporting semantics, operational anomalies, and historical maintenance information in the unified maintenance data based on a hydraulic fault semantic graph and a domain knowledge-enhanced semantic model, and generating a fault semantic understanding result representing the fault object, suspected fault mode, and maintenance requirements; and a generation module for generating a system that includes task urgency, fault complexity, and skills based on the fault semantic understanding result and the structural relationship of the hydraulic equipment. The system comprises the following modules: a semantic profile of maintenance tasks based on demand and resource requirements; a determination module, which generates a maintenance resource capability profile based on historical maintenance feedback, real-time operation status, and available resource conditions, and performs constraint matching between the maintenance task semantic profile and the maintenance resource capability profile to determine a set of candidate maintenance resources; an evaluation module, which uses a multi-objective intelligent dispatch model to evaluate the dispatch adaptation of the candidate maintenance resource set, generates a comprehensive dispatch score, determines the target maintenance resource based on the comprehensive dispatch score, and generates a maintenance task allocation result carrying the dispatch basis; and an update module, which obtains maintenance execution feedback and updates the hydraulic fault semantic graph, maintenance resource capability profile, and dispatch decision parameters based on the maintenance execution feedback.

[0007] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: By acquiring multi-source maintenance data associated with hydraulic equipment maintenance tasks awaiting dispatch and standardizing this data, unified maintenance data is obtained. Based on a hydraulic fault semantic graph and a domain knowledge-enhanced semantic model, the reported repair semantics, operational anomalies, and historical maintenance information in the unified maintenance data are fused and analyzed to generate fault semantic understanding results that characterize fault objects, suspected fault modes, and maintenance needs. Based on the relationship between the fault semantic understanding results and the hydraulic equipment structure, a maintenance task semantic profile including task urgency, fault complexity, skill requirements, and resource requirements is generated. Based on historical maintenance feedback, real-time operating status, and available resource conditions, a maintenance resource capability profile is generated, and constraint matching is performed between the maintenance task semantic profile and the maintenance resource capability profile to determine a set of candidate maintenance resources. A multi-objective intelligent dispatch model is used to evaluate the dispatch adaptation of the candidate maintenance resource set, generating a comprehensive dispatch score. The target maintenance resource is determined based on the comprehensive dispatch score, and a maintenance task allocation result carrying the dispatch basis is generated. Maintenance execution feedback is obtained, and the hydraulic fault semantic graph, maintenance resource capability profile, and dispatch decision parameters are updated based on the maintenance execution feedback. This application can improve the accuracy of maintenance task identification, enhance resource matching efficiency, and strengthen the ability to optimize dispatch feedback. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating the intelligent dispatching method for maintenance of hydraulic equipment based on semantic profiling and multi-constraints provided in the embodiments of this application. Figure 2 This is a schematic diagram of the structure of the intelligent dispatch system for the maintenance of hydraulic equipment based on semantic profiling and multi-constraints provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0010] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0011] In existing technologies, hydraulic equipment maintenance tasks typically rely on work order systems, equipment monitoring systems, and manual dispatching experience. Some systems can collect alarm information, operating status data, and historical maintenance records of hydraulic equipment, and dispatch maintenance orders according to the reported repair area, the availability of maintenance personnel, preset fault categories, or fixed shift relationships. With the development of intelligent operation and maintenance technologies, existing technologies have also introduced fault diagnosis models, maintenance knowledge bases, or rule engines to assist in the judgment of hydraulic system anomalies, so as to provide maintenance personnel with certain fault references.

[0012] However, hydraulic equipment failures are characterized by multi-source manifestations, component coupling, and causal transmission. The same failure phenomenon may be caused by different hydraulic components, different operating conditions, or different failure mechanisms. Existing dispatch methods struggle to integrate and analyze repair request semantics, operational anomalies, and historical maintenance information, leading to inaccurate identification of fault objects, failure modes, complexity, and resource requirements in maintenance tasks. Furthermore, existing systems often characterize maintenance resources at the level of personnel location, job level, or shift status, failing to dynamically generate maintenance resource capability profiles by combining historical maintenance feedback, real-time operational status, and available resource conditions. This results in insufficient matching between maintenance tasks and maintenance resources.

[0013] Furthermore, existing dispatching processes typically lack a collaborative evaluation of multiple objectives, such as skill suitability, response time, resource availability, and repair reliability. They also lack a mechanism for continuously optimizing fault identification, personnel capabilities, and dispatching parameters based on maintenance execution feedback. Therefore, existing technologies suffer from inaccurate maintenance task identification, suboptimal resource matching, and difficulty in optimizing dispatching feedback, easily leading to inaccurate dispatching, slow maintenance response, insufficient spare parts preparation, and high rates of repeat visits.

[0014] To address the aforementioned issues, this application provides a method for intelligent dispatching of hydraulic equipment maintenance tasks based on semantic profiling and multi-constraints. This method acquires multi-source maintenance data associated with hydraulic equipment maintenance tasks awaiting dispatch, and standardizes this data to obtain unified maintenance data. Based on a hydraulic fault semantic graph and a domain knowledge-enhanced semantic model, it fuses and analyzes the reporting semantics, operational anomalies, and historical maintenance information within the unified maintenance data to generate a fault semantic understanding result. Based on the fault semantic understanding result and the structural relationship of the hydraulic equipment, it generates a semantic profile of the maintenance task, including task urgency, fault complexity, skill requirements, and resource requirements.

[0015] Based on this, this application generates a maintenance resource capability profile according to the historical maintenance feedback, real-time operation status, and available resource conditions of maintenance resources. It then performs constraint matching between the maintenance task semantic profile and the maintenance resource capability profile to determine a set of candidate maintenance resources. A multi-objective intelligent dispatching model is used to evaluate the dispatching suitability of the candidate maintenance resource set and generate a comprehensive dispatching score. The target maintenance resource is determined based on the comprehensive dispatching score, and a maintenance task allocation result carrying the dispatching basis is generated. After the maintenance is completed, the hydraulic fault semantic graph, maintenance resource capability profile, and dispatching decision parameters are updated based on the maintenance execution feedback.

[0016] Through the above technical solution, this application can transform the hydraulic equipment maintenance dispatching process from simple rule matching to a closed-loop decision-making process based on fault semantic understanding, task profiling, resource profiling, and multi-constraint intelligent evaluation. This improves the accuracy of maintenance task identification, enhances the efficiency of maintenance resource matching, strengthens the interpretability and feedback optimization capabilities of dispatching results, and reduces the probability of dispatching errors, response delays, and duplicate maintenance.

[0017] The technical solution of this application will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0018] Figure 1 This is a flowchart illustrating the intelligent dispatching method for maintenance of hydraulic equipment based on semantic profiling and multi-constraints provided in this application embodiment. Figure 1 As shown, the method may specifically include: S101: Obtain multi-source maintenance data associated with the hydraulic equipment maintenance tasks to be dispatched, and standardize the multi-source maintenance data to obtain unified maintenance data. S102, based on the hydraulic fault semantic map and the domain knowledge-enhanced semantic model, integrates and analyzes the repair reporting semantics, operational anomalies and historical maintenance information in the unified maintenance data to generate fault semantic understanding results that represent fault objects, suspected fault modes and maintenance needs; S103. Based on the semantic understanding of the fault and the relationship between the structure of the hydraulic equipment, generate a semantic profile of the maintenance task that includes the urgency of the task, the complexity of the fault, the skill requirements, and the resource requirements. S104. Based on the historical maintenance feedback, real-time operation status and available resource conditions of maintenance resources, generate a maintenance resource capability profile, and perform constraint matching based on the maintenance task semantic profile and the maintenance resource capability profile to determine the candidate maintenance resource set. S105: Use a multi-objective intelligent dispatch model to evaluate the dispatch adaptation of the candidate maintenance resource set, generate a comprehensive dispatch score, determine the target maintenance resource based on the comprehensive dispatch score, and generate a maintenance task allocation result carrying the dispatch basis. S106, obtain maintenance execution feedback, and update the hydraulic fault semantic graph, maintenance resource capability profile, and dispatch decision parameters based on the maintenance execution feedback.

[0019] In some embodiments, multi-source maintenance data associated with hydraulic equipment maintenance tasks to be dispatched is obtained, and the multi-source maintenance data is standardized to obtain unified maintenance data, including: Analyze the task identifier and equipment identifier in the hydraulic equipment maintenance task to be dispatched, and obtain the task-side data, equipment-side data and resource-side data associated with the maintenance task based on the preset data mapping relationship; Based on a unified time base and equipment structure index, multi-source maintenance data are correlated and aligned, and unstructured data is semantically normalized based on semantic rules in the hydraulic field. After the multi-source maintenance data is aligned and associated, quality verification, source labeling, and credibility encoding are performed to generate unified maintenance data that carries task association, equipment association, and data credibility representation.

[0020] Specifically, after receiving a hydraulic equipment repair request, the intelligent dispatch platform first parses the hydraulic equipment repair task to be dispatched, extracting the task identifier, equipment identifier, repair time, repair source, and site location identifier. The task identifier is used to associate the repair record, processing status, and dispatch priority in the work order system, while the equipment identifier is used to associate the equipment ledger, hydraulic system configuration, and historical operating data. For example, if the main boom lifting hydraulic system of a port loading and unloading equipment experiences slow lifting and lifting movements accompanied by rising oil temperature, the work order system generates a task number, and the equipment ledger records the target equipment number, hydraulic pump station number, main boom cylinder number, and proportional valve group number. Based on the task number and equipment number, the platform calls the preset data mapping relationship to obtain the corresponding data from the work order system, equipment monitoring system, inspection system, spare parts system, and personnel resource system.

[0021] During data acquisition, the platform aggregates multi-source maintenance data according to task-side, equipment-side, and resource-side data. Task-side data includes repair report text, voice-transcribed content, repair report image tags, alarm summaries, and manual supplementary explanations; equipment-side data includes pressure, flow rate, oil temperature, valve position feedback, motor current, operating cycle, historical alarms, and equipment structural relationships; resource-side data includes maintenance personnel availability, maintenance team coverage area, tool configuration, spare parts inventory, and warehouse location. To ensure unified processing of data across different systems, the platform configures a data source identifier, collection time identifier, and object association identifier for each data record, and writes multi-source data under the same task into a data cache area to be standardized.

[0022] Subsequently, the platform performs time correlation and alignment on multi-source maintenance data according to a unified time benchmark. For sensor data, a preset time interval before and after the repair report time is used as the analysis window, and pressure curves, oil temperature curves, action cycle changes, and alarm trigger times are converted to the same time axis. For work order data and inspection records, an event sequence is established according to the task occurrence time, inspection time, and alarm time. For equipment structure data, hydraulic pump stations, valve groups, cylinders, pipelines, and sensor points are mapped to the corresponding hydraulic functional units according to the equipment structure index. Taking the aforementioned main boom lifting hydraulic system as an example, the platform uniformly correlates abnormal oil temperature curves, main pump outlet pressure drops, lifting action cycle extensions, and proportional valve response delays to the main boom lifting circuit, enabling subsequent semantic understanding to revolve around the same hydraulic structural object, rather than treating various anomalies as isolated data.

[0023] For unstructured data, the platform performs semantic normalization based on semantic rules in the hydraulic domain. In the repair report text, phrases like "unable to lift," "weak movement," and "slow lifting" are normalized to the semantic tag "slow actuator movement"; "oil hot" and "high temperature" are normalized to the semantic tag "abnormal oil temperature"; and "loud pump noise" and "obvious vibration" are normalized to the semantic tag "abnormal hydraulic source operation." When the speech-to-text contains colloquial expressions, the platform combines a hydraulic terminology dictionary and a contextual semantic model to identify key fault descriptions and maps synonymous expressions to standard fault phenomena. For alarm codes, the platform converts them into corresponding abnormal semantics according to an alarm mapping table. For example, high oil temperature alarms, pressure establishment timeout alarms, and valve control response abnormal alarms are mapped to standard abnormal features that can be identified by the hydraulic fault semantic map.

[0024] After completing association alignment and semantic normalization, the platform performs quality verification on multi-source maintenance data. Quality verification includes data integrity verification, temporal continuity verification, equipment identifier consistency verification, and outlier credibility verification. If a pressure measurement point experiences short-term packet loss within the analysis window, the platform marks the missing status based on adjacent sampling points and related measurement points in the same circuit, instead of directly deleting the data. If the reported equipment number does not match the alarm equipment number, the platform identifies the conflict based on the site location, equipment hierarchy, and task source, and writes the conflict result into the data credibility representation. Data with reliable sources and temporal continuity is configured with a high credibility level, while data with incomplete manual descriptions, abnormal sensor fluctuations, or conflicting sources is configured with a credibility level requiring verification.

[0025] Ultimately, the platform generates unified maintenance data. This unified maintenance data carries task relationships, equipment relationships, and data reliability representations. Task relationships represent the correspondence between various data types and the hydraulic equipment maintenance tasks to be dispatched; equipment relationships represent the correspondence between various abnormal data types and the structural objects of the hydraulic equipment; and data reliability representations represent the data source, completeness, temporal consistency, and semantic normalization reliability.

[0026] Through the above processing, scattered repair reports, operational anomalies, historical records, and resource information are transformed into unified maintenance data with clear structure, well-defined objects, and distinguishable credibility. This provides a stable data foundation for subsequent fault semantic understanding, maintenance task semantic profile generation, and multi-constraint intelligent dispatching, thereby improving the accuracy of maintenance task identification and the reliability of dispatching decisions.

[0027] In some embodiments, based on a hydraulic fault semantic map and a domain knowledge-enhanced semantic model, the reported repair semantics, operational anomalies, and historical maintenance information in unified maintenance data are fused and analyzed to generate fault semantic understanding results that characterize fault objects, suspected fault modes, and maintenance needs, including: Semantic features of maintenance tasks and abnormal features of equipment operation are extracted from unified maintenance data, and the extraction results are mapped to equipment structure nodes and fault mechanism nodes in the hydraulic fault semantic map. By leveraging domain knowledge to enhance the semantic model, contextual semantic encoding is performed on the mapped features to generate fault semantic representations corresponding to the hydraulic equipment maintenance tasks to be dispatched. Based on the associated paths in the hydraulic fault semantic graph, graph-constrained reasoning and historical similar maintenance evidence fusion are performed on the fault semantic representation to determine the matching confidence between fault objects, suspected fault modes and maintenance needs. Generate fault semantic understanding results based on matching confidence.

[0028] Specifically, after obtaining unified maintenance data, the intelligent dispatch platform invokes the hydraulic fault semantic analysis service to integrate the reported repair semantics, operational anomalies, and historical maintenance information in the hydraulic equipment maintenance tasks to be dispatched. Taking a repair task for the coiler pressing mechanism of a hydraulic station in a metallurgical production line as an example, the fault description submitted by the repair personnel is slow pressing action, pressure not rising, and high oil temperature. The equipment monitoring system simultaneously records that the pump outlet pressure build-up time is prolonged, the return oil temperature is continuously rising, and there is a lag between the proportional valve opening command and the actual feedback. Historical work orders show maintenance records of the same model of equipment with slow action due to filter blockage and pump internal leakage. The platform first extracts the semantic features of the maintenance task from the unified maintenance data to identify standard fault phenomena such as slow action, abnormal pressure build-up, and abnormal oil temperature. At the same time, it extracts operational anomaly features such as pressure rise slope, oil temperature change trend, valve position response deviation, and the proportion of extended action cycle from the equipment operation data.

[0029] The platform maps the extracted semantic features of maintenance tasks and abnormal equipment operation features to a hydraulic fault semantic graph. The hydraulic fault semantic graph pre-constructs nodes for equipment structure, hydraulic functional units, fault phenomena, fault modes, abnormal measurement points, and maintenance actions. For the aforementioned repair task of the pressing mechanism, the platform maps the task to the coiler pressing hydraulic circuit based on the equipment structure index, maps abnormal pump outlet pressure to nodes related to the hydraulic pump and pressure measurement points, maps valve position response lag to proportional valve control nodes, and maps abnormal oil temperature to nodes related to oil state and system heating. Through this mapping process, the abnormal information originally scattered in the repair text and operating curves is converted into node features in the graph with structural location and fault mechanism indications.

[0030] Furthermore, after completing the graph mapping, the platform utilizes domain knowledge to enhance the semantic model and perform contextual semantic encoding on the mapped features. The domain knowledge-enhanced semantic model includes a hydraulic terminology encoding unit, a fault context encoding unit, and a graph neighborhood enhancement unit. The hydraulic terminology encoding unit converts on-site descriptions such as weak action, insufficient pressure, and significant overheating into standard hydraulic fault semantics; the fault context encoding unit combines the repair time, equipment operating condition, and abnormal change trends to determine the actual direction of the fault semantics; and the graph neighborhood enhancement unit introduces the mechanistic relationships between adjacent nodes in the hydraulic fault semantic graph. For example, when sluggish action and abnormal pressure build-up occur simultaneously, and the oil temperature rises rapidly, the model will enhance the contextual response of fault mechanisms such as pump internal leakage, filter blockage, and relief valve leakage, rather than directly determining the repair target based solely on a single textual description.

[0031] Subsequently, the platform performs graph-constrained reasoning on the semantic representation of faults based on the associated paths in the hydraulic fault semantic graph. Graph-constrained reasoning starts with the equipment structure node corresponding to the current task and propagates confidence along the associated paths between fault phenomena, abnormal measurement points, suspected fault modes, and maintenance actions. It then calculates the basic confidence of candidate fault modes based on path length, node weight, and consistency of abnormal features. For example, the path between a slowed pressing action and a slow pump outlet pressure build-up points to a decrease in hydraulic pump efficiency; the path between an increased oil temperature and an abnormal return oil temperature points to internal leakage or throttling overheating in the system; and a lag in proportional valve feedback points to valve core sticking or abnormal control response. The platform integrates these path reasoning results to form initial matching results for multiple suspected fault modes.

[0032] To avoid misjudgments due to relying solely on graph-based reasoning, the platform further integrates historical similar maintenance evidence. The platform retrieves maintenance samples from the historical maintenance database that are similar to the current task in terms of equipment model, hydraulic circuit, fault symptoms, abnormal measurement point trends, and operating conditions. It then extracts the actual fault cause, actual maintenance actions, spare parts replacement results, and maintenance success status. If similar samples repeatedly show filter blockage leading to slow pressure build-up, which is resolved after filter replacement, the platform increases the matching confidence level for filter blockage. If similar samples show pump internal leakage typically accompanied by abnormal pump current and a sustained pressure drop, but the current data does not show a significant sustained pressure drop, the platform decreases the matching confidence level for pump internal leakage. Through the fusion of historical similar maintenance evidence, the fault reasoning results can balance hydraulic mechanism constraints and on-site maintenance experience.

[0033] When determining the matching confidence level, the platform calculates the confidence level of the fault object, the confidence level of the suspected fault mode, and the confidence level of the maintenance requirement. The confidence level of the fault object indicates the probability that the fault is concentrated in objects such as hydraulic pumps, proportional valves, filters, or actuator cylinders; the confidence level of the suspected fault mode indicates the degree of matching of fault mechanisms such as blockage, leakage, jamming, response lag, and abnormal oil; and the confidence level of the maintenance requirement indicates the necessity of maintenance actions such as inspection, disassembly, replacement, cleaning, pressure adjustment, or joint debugging. The platform generates fault semantic understanding results according to the confidence level and retains supporting paths, abnormal features, and similar maintenance evidence in the results for subsequent semantic profiling of maintenance tasks and dispatch adaptation evaluation.

[0034] Through the above implementation methods, the platform can uniformly convert repair requests, operational anomalies, and historical maintenance information into fault semantic understanding results with structural orientation, mechanistic constraints, and evidence support. This reduces the bias caused by judging fault objects based solely on keywords or human experience, improves the accuracy of hydraulic equipment maintenance task identification, and provides a reliable basis for subsequently determining skill requirements, resource requirements, and candidate maintenance resources.

[0035] In some embodiments, based on the relationship between the fault semantic understanding results and the hydraulic equipment structure, a maintenance task semantic profile is generated, including task urgency, fault complexity, skill requirements, and resource requirements, including: The results of fault semantic understanding are mapped to the corresponding equipment units and hydraulic functional units in the structural relationship of hydraulic equipment to determine the structural association range of the object to be repaired. Based on the structural association range, fault propagation relationship and operational impact status, calculate the task urgency and fault complexity of the hydraulic equipment maintenance task to be dispatched; Based on the semantic association between suspected fault modes and maintenance actions, the skill and resource requirements corresponding to the hydraulic equipment maintenance tasks to be dispatched are determined. By mapping and encoding the urgency of the task, the complexity of the fault, the skill requirements, and the resource requirements, a semantic profile of the maintenance task is generated.

[0036] Specifically, after generating the fault semantic understanding results, the intelligent dispatch platform further calls the maintenance task profile generation module to map the fault object, suspected fault mode, matching confidence, and maintenance requirements to the structural relationships of the hydraulic equipment, thereby determining the structural association range corresponding to the object to be repaired. Taking a repair task of the mold-locking hydraulic system of an injection molding production line as an example, the fault semantic understanding results characterize slow mold-locking action, prolonged system pressure build-up time, and lag in proportional valve response. Suspected fault modes include proportional valve spool sticking, decreased hydraulic pump efficiency, and clogged return oil filter. Based on the equipment structural relationships, the platform maps this result to the mold-locking hydraulic circuit, pump station pressure supply unit, proportional valve control unit, actuator cylinder unit, and return oil filter unit, and determines the pressure supply, pressure regulation, execution, and return oil association relationships between the above units, forming the structural association range of the object to be repaired.

[0037] When determining the scope of structural impact, the platform does not simply pinpoint repair requests to individual components. Instead, it considers the hydraulic circuit topology and fault propagation direction to determine the affected functional range. For example, prolonged pressure build-up time could stem from decreased efficiency on the pressure supply side, or from control valve leakage or abnormal return oil resistance. Slow mold clamping action can affect the injection molding machine's mold closing cycle time and may further extend the product molding cycle. Based on the structural relationships of the hydraulic equipment, the platform designates the pump station pressure supply unit, proportional valve control unit, and mold clamping cylinder execution unit as the core impact range, and the oil tank return oil filter unit and oil temperature monitoring unit as extended impact ranges. It also configures the impact strength and fault direction for different structural units.

[0038] Subsequently, the platform calculates the task urgency based on the structural correlation range, fault propagation relationships, and operational impact status. Task urgency can be determined based on whether equipment is shut down, whether the fault affects critical production cycles, whether the anomaly poses a risk of propagation, and the on-site production scheduling status. For example, even if the mold-locking mechanism is not completely shut down, but its current operating cycle is 35% longer than the baseline cycle, the production line is under high load with continuous production orders, and the system oil temperature continues to rise, the platform increases the task urgency accordingly. If the fault semantic understanding result indicates that the pressure anomaly has affected the safety interlock or caused the equipment to automatically shut down, the platform further increases the task urgency and marks the task as a priority dispatch task.

[0039] The platform also calculates fault complexity based on the number of suspected fault modes, the extent of structural correlation, the consistency of abnormal characteristics, and the difficulty of maintenance diagnosis. The current task involves multiple functional units, including pressure supply, control, and filtration. Furthermore, there are similar external manifestations among the proportional valve spool sticking, hydraulic pump efficiency decline, and filter blockage. The platform determines that this task requires on-site testing of pressure, flow rate, valve position response, and filter differential pressure, making its fault complexity higher than that of a typical seal replacement task. If historical similar maintenance records indicate that the same model of equipment has experienced compound faults, the platform will further increase the fault complexity and introduce advanced hydraulic diagnostic capabilities into the subsequent dispatch model.

[0040] Furthermore, after determining the urgency and complexity of the task, the platform generates skill and resource requirements based on the semantic association between suspected fault modes and maintenance actions. For example, for a stuck proportional valve spool, the platform determines that skills related to proportional valve disassembly and inspection, valve control response testing, and system flushing are required; for a decrease in hydraulic pump efficiency, the platform determines that skills related to pump station pressure testing, internal leakage detection, and pump unit maintenance are required; and for a clogged return oil filter, the platform determines that skills related to filter differential pressure detection and filter element replacement are required. Resource requirements are derived from maintenance actions and include hydraulic test fittings, pressure testing tools, valve assembly / disassembly tools, spare filter elements, hydraulic oil replenishment resources, and on-site safety isolation conditions. The platform converts these skill and resource requirements into structured tags that can be recognized by the dispatch model and retains the correspondence between them and suspected fault modes.

[0041] Finally, the platform performs profiling and encoding on the task urgency, fault complexity, skill requirements, and resource requirements to generate a semantic profile of the maintenance task. This semantic profile includes task level encoding, complexity encoding, skill requirement vector, resource requirement vector, structural association range encoding, and profile credibility encoding. For the aforementioned mold-locking hydraulic system task, the platform generates a semantic profile indicating high urgency, medium-to-high complexity, the need for proportional valve diagnostics and hydraulic system testing capabilities, and the requirement to carry filter elements and pressure testing tools. This profile is then passed to the subsequent maintenance resource capability profile matching stage.

[0042] Through the above methods, the platform can further transform the semantic understanding results of faults into computable dispatch decision objects, so that maintenance tasks not only reflect the fault phenomena, but also accurately express the scope of structural impact, maintenance difficulty, personnel capability requirements and resource preparation requirements, thereby improving the accuracy of subsequent maintenance resource matching and the reliability of dispatch decisions.

[0043] In some embodiments, a maintenance resource capability profile is generated based on historical maintenance feedback, real-time operational status, and available resource conditions. Constraint matching is then performed between the maintenance task semantic profile and the maintenance resource capability profile to determine a candidate maintenance resource set, including: Based on historical maintenance feedback of maintenance resources, capability performance characteristics associated with hydraulic equipment failure modes and maintenance actions are extracted, and a maintenance resource capability profile is generated by combining real-time operating status and available resource conditions. Based on the task constraint requirements in the semantic profile of maintenance tasks, construct constraint matching conditions for the profile of maintenance resource capabilities. The maintenance resource capability profile is adapted and filtered using constraint matching conditions to exclude maintenance resources that do not meet the basic conditions for dispatching orders. The selected maintenance resources are matched based on capability similarity and resource accessibility to determine the candidate maintenance resource set.

[0044] Specifically, after generating a semantic profile of the maintenance task, the intelligent dispatch platform calls upon the maintenance resource profile service to model the capabilities of maintenance personnel, maintenance teams, and external maintenance units that can participate in the dispatch. Taking a maintenance task for the luffing hydraulic system of a port stacker-reclaimer as an example, the semantic profile of the maintenance task indicates that the task is of high urgency and medium-to-high complexity, requiring hydraulic pump station pressure testing, proportional valve response diagnosis, filter element inspection, and rapid recovery capabilities, as well as pressure testing tools, test connectors, and spare filter elements. The platform first reads maintenance records of each maintenance resource in the historical maintenance database for the same equipment type, similar hydraulic circuits, and similar fault modes, extracting the results of a single repair, average arrival time, average diagnosis time, rework status, spare parts usage accuracy, and on-site feedback results to form capability performance characteristics associated with hydraulic equipment fault modes and maintenance actions.

[0045] During the capability performance feature extraction process, the platform does not simply judge maintenance capabilities based on employee type or level, but rather breaks down historical maintenance feedback into specific fault modes and maintenance action dimensions. For example, maintenance worker A has a high first-time repair rate in tasks involving abnormal pump station pressure and filter blockage, and has repeatedly completed rapid restoration of hydraulic circuits in port equipment. The platform improves this worker's capability parameters in abnormal pressure supply diagnosis, filter element replacement, and on-site emergency handling. Maintenance worker B has experience in proportional valve commissioning, but recently experienced two valve control response anomalies requiring repair. The platform lowers this worker's repair reliability parameter in complex valve control diagnosis tasks. For team-based maintenance resources, the platform also generates combined capability features based on skill complementarity, collaborative maintenance records, and task completion stability.

[0046] Subsequently, the platform generates a maintenance resource capability profile by combining real-time operation status and available resource conditions. Operation status includes current task occupancy, estimated availability window, location, continuous operation duration, and order response status; available resource conditions include tool carrying status, spare parts availability status, service area coverage, and on-site access qualifications. Taking the aforementioned variable amplitude hydraulic system task as an example, maintenance personnel A, while having a high capability match, is currently handling emergency repairs on another piece of equipment and is expected to be available in two hours; maintenance personnel C is closer to the site, possesses high-pressure hydraulic system operation qualifications, and carries pressure testing tools in their vehicle, but lacks spare filters; maintenance team D is slightly farther away, but can obtain filters from the warehouse and simultaneously arrange for two personnel to be on-site. The platform encodes capability characteristics, real-time status, and resource conditions into a maintenance resource capability profile, and assigns a profile update time and reliability identifier to each profile.

[0047] During the candidate screening phase, the platform constructs constraint matching conditions based on the task constraints in the semantic profile of the maintenance task. These constraints include basic dispatch conditions and optimized matching conditions. Basic dispatch conditions determine whether maintenance resources possess the necessary hydraulic diagnostic capabilities, on-site operation qualifications, available time windows, and service coverage. Optimized matching conditions evaluate the degree of capability matching, resource accessibility, and response adaptability between maintenance resources and the task. For high-urgency tasks, the platform increases the constraint strength of arrival time windows and safety qualifications; for medium-to-high complexity tasks, the platform increases the constraint strength of failure mode experience and diagnostic capabilities; and for tasks with significant resource dependence, the platform increases the constraint strength of the accessibility of key tools and spare parts.

[0048] The platform uses constraint matching conditions to adapt and filter maintenance resource capability profiles, excluding maintenance resources that do not meet the basic conditions for dispatching. For example, personnel without the qualifications for on-site operation of high-pressure hydraulic systems are excluded, as are personnel unable to arrive within the task window, and outsourcing units lacking basic hydraulic testing capabilities. For maintenance resources that are not excluded, the platform further calculates capability similarity and resource accessibility. Capability similarity is obtained by matching skill requirement vectors with capability vectors, while resource accessibility is determined comprehensively by considering personnel location, tool status, spare parts acquisition paths, and on-site access conditions. If a maintenance resource has a high skill match but spare parts are unavailable, the platform retains the resource but lowers its priority; if a work team has complementary skills and spare parts can arrive in a timely manner, the platform considers that team as a candidate resource for combination.

[0049] Ultimately, the platform determines a set of candidate maintenance resources and writes capability matching tags, resource reachability tags, and constraint satisfaction status for each candidate resource. Through this process, maintenance resources are no longer simply filtered based on static personnel files, but rather a dynamic capability profile is formed based on historical maintenance feedback, real-time operation status, and available resource conditions. This profile is then matched with the semantic profile of the maintenance task, thereby improving the accuracy of the candidate maintenance resource set and providing reliable input for the subsequent multi-objective intelligent dispatch model to generate a comprehensive dispatch score.

[0050] In some embodiments, a multi-objective intelligent dispatching model is used to evaluate the dispatching suitability of the candidate maintenance resource set and generate a comprehensive dispatching score, including: Input the semantic profile of the maintenance task and the maintenance resource capability profile corresponding to the candidate maintenance resource set into the multi-objective intelligent dispatch model to generate task resource adaptation features. Based on the urgency and complexity of the hydraulic equipment repair tasks to be dispatched, determine the dynamic weighting relationship between the dispatch evaluation objectives; Based on task resource adaptation characteristics and dynamic weight relationships, a collaborative evaluation of skill adaptation, response time, resource availability and repair reliability is performed on each candidate maintenance resource to generate a dispatch adaptation value for the candidate maintenance resource. The dispatch adaptation value is constrained and corrected according to the preset constraint correction rules to generate a comprehensive dispatch score.

[0051] Specifically, after determining the set of candidate maintenance resources, the intelligent dispatch platform invokes a multi-objective intelligent dispatch model to evaluate the compatibility between each candidate maintenance resource and the hydraulic equipment maintenance task to be dispatched. Taking a maintenance task for the hydraulic lubrication unit of a mining crushing station as an example, the semantic profile of the maintenance task indicates that the task is of high urgency and medium fault complexity, with suspected fault modes being insufficient hydraulic pump pressure and filter unit blockage. Skill requirements include hydraulic pump pressure testing, filter differential pressure judgment, and system recovery and debugging. Resource requirements include pressure testing tools, filter spare parts, and on-site safe operating conditions. The set of candidate maintenance resources includes maintenance personnel A, maintenance personnel B, and maintenance team C, each corresponding to different historical maintenance capabilities, current locations, tool carrying status, and current task load.

[0052] The platform first inputs the semantic profile of the maintenance task and the corresponding maintenance resource capability profile of each candidate maintenance resource into a multi-objective intelligent dispatch model. The model performs task-side encoding on task urgency, fault complexity, skill requirements, resource requirements, and structural correlation range; and resource-side encoding on personnel skills, similar fault handling records, available time windows, location distance, tool and spare parts conditions, and repair stability. It then identifies the response relationship between task requirements and resource capabilities through a cross-profile attention mechanism. For example, maintenance worker A has a high first-time repair rate in tasks involving insufficient hydraulic pump pressure, but is far from the site; maintenance worker B is close to the site and carries pressure testing tools, but needs to retrieve filter spare parts separately; maintenance team C has both pump station diagnostic and filter replacement capabilities, but their current available time window is later than the task requirements. Based on these relationships, the model generates task resource adaptation features, ensuring that subsequent evaluation no longer relies on a single distance or personnel idle status.

[0053] Furthermore, after generating task resource adaptation characteristics, the platform determines the dynamic weighting relationship between dispatch evaluation targets based on the task urgency and fault complexity of the hydraulic equipment repair task to be dispatched. For tasks with high urgency, the platform increases the weight of response time and resource availability in the overall evaluation to avoid selecting personnel who arrive too slowly simply because of high historical skill scores. Since the fault complexity of this task is medium, the platform retains the importance of skill adaptation and repair reliability, but does not regard complex diagnostic capabilities as the sole dominant factor. If the task semantic profile indicates high fault complexity, such as simultaneously involving proportional valve control abnormalities, pump internal leakage, and oil contamination, the platform will further increase the weight of skill adaptation and repair reliability. If the task is in a shutdown emergency repair state, the platform will further increase the weight of response time.

[0054] Subsequently, the platform collaboratively evaluated each candidate maintenance resource based on task resource adaptation characteristics and dynamic weight relationships. Skill adaptation evaluation determined whether the candidate maintenance resources covered the hydraulic diagnostic, disassembly, debugging, and recovery capabilities required by the task; response timeliness evaluation determined whether the expected order acceptance, departure, and arrival times of the candidate maintenance resources met the task urgency requirements; resource availability evaluation determined whether key tools, spare parts, and on-site operating conditions could be met before maintenance began; and repair reliability evaluation determined the completion quality by combining historical first-time repair rates, rework records, and performance on similar tasks. For maintenance worker A, the model gave a high skill adaptation and repair reliability score, but a low response timeliness score due to a longer arrival time; for maintenance worker B, the model gave a high response timeliness and tool availability score, but a reduced resource availability score due to a lack of filtered spare parts; for maintenance team C, the model gave a high skill adaptation and resource coverage score, but the available time window did not meet the urgent task requirements.

[0055] The platform integrates various evaluation dimensions according to dynamic weighting relationships to generate dispatch adaptation values ​​for candidate maintenance resources. After the dispatch adaptation values ​​are generated, the platform continues to correct them according to preset constraint correction rules. These constraint correction rules include safety qualification correction, task conflict correction, spare parts shortage correction, continuous operation risk correction, and collaborative feasibility correction. For example, if a candidate maintenance resource does not have the qualification for high-pressure hydraulic operation at a mine site, it will be forcibly downgraded or excluded, even if the initial model adaptation value is high; if the continuous operation time of a candidate exceeds a preset threshold, the platform deducts from the execution stability beyond response time; if critical spare parts can be quickly delivered from the warehouse to the site, the platform partially restores the deduction caused by spare parts shortage.

[0056] Finally, the platform outputs a comprehensive dispatch score for each candidate repair resource and retains the score contribution information. For the aforementioned task, Repair Personnel B received the highest comprehensive dispatch score because of their proximity, having already brought their tools, and being able to simultaneously pick up the filter from the warehouse; Repair Personnel A's ranking was lowered because their arrival time could not meet the high urgency requirements; Repair Team C was not prioritized for dispatch due to a conflict in available time windows.

[0057] Through the above implementation methods, the multi-objective intelligent dispatch model can dynamically balance skill adaptation, response time, resource availability, and repair reliability. It also eliminates dispatch results with high surface scores but are actually unexecutable through constraint correction rules, thereby improving the accuracy, timeliness, and execution reliability of hydraulic equipment maintenance dispatch.

[0058] In some embodiments, the semantic profile of the maintenance task and the maintenance resource capability profile corresponding to the candidate maintenance resource set are input into the multi-objective intelligent dispatch model to generate task resource adaptation features, including: The semantic profile of maintenance tasks and the profile of maintenance resource capabilities are homogeneously encoded to obtain task requirement vectors and resource capability vectors. Based on the cross-profile attention mechanism in the multi-objective intelligent dispatch model, the task requirement vector and resource capability vector are interactively matched to extract the correlation response between task requirements and resource capabilities. Task resource adaptation features are generated based on the degree of coverage of the associated response, the degree of resource satisfaction, and the degree of execution adaptability.

[0059] Specifically, after receiving the semantic profile of the maintenance task and the set of candidate maintenance resources, the model first performs isomorphic encoding of the profile, so that the task-side data and the resource-side data enter the same feature space. Taking the maintenance task of abnormal downward speed of the main cylinder of a large hydraulic bending machine as an example, the semantic profile of the maintenance task records information such as high urgency, medium to high fault complexity, the associated range of the main cylinder execution unit, suspected proportional valve response lag, insufficient pressure supply from the pump station, and the need for pressure detection and valve control debugging.

[0060] The candidate maintenance resource set includes maintenance personnel A, maintenance personnel B, and maintenance team C, each with different hydraulic valve debugging experience, tool carrying status, location distance, and current task load. The platform encodes task urgency, fault complexity, skill requirements, resource requirements, structural correlation scope, and profile credibility into a task requirement vector. Simultaneously, it encodes the capability performance, real-time availability, tool and spare parts conditions, historical repair quality, and on-site access status from the maintenance resource capability profile into a resource capability vector.

[0061] During isomorphic encoding, the platform uses feature slots of the same dimension for both the task side and the resource side, ensuring that task requirements and resource capabilities are aligned item by item. For example, the proportional valve response diagnosis requirement on the task side corresponds to the proportional valve debugging capability on the resource side; the pump station pressure supply detection requirement corresponds to the pump station pressure testing experience; the rapid arrival requirement corresponds to the location and available time window; and the resource preparation requirement corresponds to the tool carrying and spare parts acquisition capabilities. For numerical features, the platform converts them into continuous feature values ​​according to preset normalization rules; for semantic features, the platform performs embedding encoding based on hydraulic domain word vectors and maintenance action labels; for features with low credibility, the platform adds credibility correction markers to the vectors to avoid low-quality data having an excessive impact on the matching results.

[0062] Subsequently, the cross-profile attention mechanism in the multi-objective intelligent dispatch model interactively matches the task requirement vector and resource capability vector. The model uses the task requirement vector as the query end and the resource capability vector as the response end, calculating the attention response strength of the task requirement dimension to the resource capability dimension. If the confidence level of abnormal main cylinder downward speed and proportional valve response lag is high, the model will strengthen the response between proportional valve debugging capabilities, valve position feedback detection experience, and valve assembly / disassembly tools; if the confidence level of insufficient pump station pressure is also high, the model will further strengthen the response between pump station pressure testing capabilities and pressure detection tools. Through interactive matching, the model can identify the coverage relationship of candidate maintenance resources for different needs in the same maintenance task, rather than simply comparing personnel levels or distance.

[0063] In the aforementioned example, maintenance worker A has extensive experience in proportional valve commissioning and a high first-time repair rate, but is currently located far away and does not carry pressure testing tools; maintenance worker B is closer to the site and carries pressure gauges and test connectors, but has a higher return rate when handling complex proportional valve faults; maintenance team C includes valve control commissioning personnel and pump station maintenance personnel, covering the main skill requirements, and can obtain valve core cleaning tools from a nearby spare parts cabinet, but their arrival time is slightly longer. A cross-profile attention mechanism extracts the correlated responses of each candidate resource in terms of skill coverage, resource satisfaction, and execution adaptation, forming a response matrix. High response positions in the response matrix indicate that task requirements are effectively covered, while low response positions indicate capability gaps, resource gaps, or execution conflicts.

[0064] The platform generates task resource adaptation characteristics based on associated responses. These characteristics include capability coverage, resource fulfillment, and execution adaptation. Capability coverage reflects the level of coverage of candidate maintenance resources for hydraulic fault diagnosis, disassembly, debugging, and recovery needs; resource fulfillment reflects whether tools, spare parts, operator qualifications, and on-site support conditions meet the maintenance initiation requirements; and execution adaptation reflects the matching status of candidate maintenance resources with the current task in terms of response time, task load, collaborative feasibility, and repair stability. For maintenance personnel A, the task resource adaptation characteristics are high capability coverage, moderate resource fulfillment, and low execution adaptation; for maintenance personnel B, the characteristics are high response adaptation, relatively high resource fulfillment, and risky capability coverage; and for maintenance team C, the characteristics are high capability coverage, high resource fulfillment, and moderate execution adaptation. These task resource adaptation characteristics are input into the subsequent dispatch adaptation evaluation stage to calculate the comprehensive dispatch score.

[0065] Through the above methods, the platform can convert task semantic profiles and maintenance resource capability profiles into alignable and interactive adaptive features, enabling the multi-objective intelligent dispatch model to identify the deep matching relationship between maintenance needs and resource capabilities. This reduces dispatch deviations caused by relying solely on human experience, fixed rules, or single distance factors, and improves the matching accuracy and dispatch reliability of hydraulic equipment maintenance tasks and maintenance resources.

[0066] In some embodiments, the target maintenance resource is determined based on the comprehensive dispatch score, and a maintenance task allocation result carrying the dispatch basis is generated, including: The candidate maintenance resource set is sorted according to the comprehensive dispatch score, and the target maintenance resource that meets the allocation conditions is determined in combination with the dispatch constraints of the hydraulic equipment maintenance task to be dispatched. When the semantic profile of a maintenance task indicates a need for collaborative maintenance, a combination of collaborative maintenance resources is generated based on the complementary capabilities and matching time windows of candidate maintenance resources. Based on the target maintenance resources or the combination of collaborative maintenance resources, generate maintenance task allocation results that include task execution objects, expected response status and resource preparation information; The dispatch criteria are generated based on the contribution characteristics of the comprehensive dispatch score, and the dispatch criteria are linked with the maintenance task allocation results for output.

[0067] Specifically, after obtaining the comprehensive dispatch score of each candidate maintenance resource, the intelligent dispatch platform calls the dispatch decision module to generate the maintenance task allocation result. Taking the maintenance task of abnormal pressure fluctuation in the pitching hydraulic system of a port quay crane as an example, the candidate maintenance resource set obtained from the previous processing includes maintenance personnel A, maintenance personnel B, and maintenance team C. The comprehensive dispatch score shows that maintenance personnel A has a high score in hydraulic pump station maintenance and pressure testing, but their current location is far from the site, and the estimated arrival time exceeds the task requirements; maintenance personnel B is closer to the site, has the qualifications for high-pressure hydraulic system operation, and carries pressure testing tools in their vehicle, but lacks experience in handling complex proportional valve faults; maintenance team C consists of pump station maintenance personnel and valve control debugging personnel, with complete skill coverage, and can obtain test connectors and valve core cleaning components from the spare parts warehouse, and the estimated arrival time meets the emergency repair window. Therefore, the platform first sorts the candidate maintenance resource set according to the comprehensive dispatch score, and then performs an executability review in combination with the dispatch constraints.

[0068] Dispatch constraints include task window, site access, operational safety, resource readiness, and current task conflicts. If a candidate maintenance resource has a high overall dispatch score but does not meet the necessary access conditions, the platform will not designate that resource as a target maintenance resource. For example, although maintenance worker A has a high historical repair reliability, if the estimated arrival time exceeds the allowable downtime for quay crane repairs, the platform will downgrade him to a backup resource. Maintenance worker B can arrive quickly, but the task semantic profile shows a dual diagnostic requirement for proportional valve response anomalies and pump station pressure fluctuations, indicating a capability gap for single-person handling; therefore, the platform will not directly designate him as a single target maintenance resource. Maintenance team C meets the requirements for arrival window, skill coverage, resource readiness, and safety qualifications, and the platform will designate this team as a priority target maintenance resource.

[0069] When the semantic profile of a maintenance task indicates a need for collaborative maintenance, the platform further generates a collaborative maintenance resource combination based on the complementary capabilities and time window matching relationships between candidate maintenance resources. Collaborative maintenance needs can be triggered by fault complexity, structural correlation, and maintenance action relationships. For example, if the current task involves pump station pressure testing, proportional valve response testing, and on-site recovery and commissioning, the platform determines that at least one pump station maintenance worker and one valve control commissioning worker are needed for collaborative processing. The platform performs a combined evaluation of the candidates' historical collaboration records, skill complementarity, arrival time difference, and tool resource sharing conditions, ultimately determining that the pump station maintenance worker in maintenance team C is the primary execution target, the valve control commissioning worker is the collaborative execution target, and maintenance worker B is designated as the on-site rapid support target for initial safety isolation and preliminary pressure measurement point confirmation.

[0070] Subsequently, the platform generates maintenance task allocation results based on the target maintenance resources or a combination of collaborative maintenance resources. The maintenance task allocation results include the task execution targets, expected response status, and resource preparation information. Task execution targets include primary maintenance personnel, collaborative maintenance personnel, and backup support personnel; expected response status includes expected order acceptance time, expected departure time, expected arrival time, and expected start time of maintenance; resource preparation information includes the pressure testing tools, hydraulic test fittings, valve control debugging terminals, spare seals, and cleaning components to be carried. The platform can also automatically generate dispatch priorities based on task urgency and write spare parts collection points, site entrances, and safety precautions into the dispatch results, enabling maintenance resources to directly perform preparation actions based on the dispatch results after accepting the order.

[0071] While generating dispatch results, the platform generates dispatch criteria based on the contribution characteristics of the comprehensive dispatch score. The dispatch criteria explain the technical reasons for selecting target maintenance resources, including contribution from skill suitability, response timeliness, resource availability, and repair reliability. For the aforementioned task, the dispatch criteria generated by the platform indicate that maintenance team C possesses complementary capabilities in pump station maintenance and valve control debugging, has a high first-time repair rate for similar historical tasks, can arrive on-site within the task window, and can simultaneously prepare key tools and spare parts; maintenance personnel B is assigned to on-site support because they are closest to the site and possess safety isolation and basic pressure testing capabilities. The platform correlates the dispatch criteria with the maintenance task allocation results and outputs them to the work order system and mobile maintenance terminal. Dispatchers can view the score contribution and constraint satisfaction status, while maintenance personnel can view the execution target, arrival time limit, and prepared resources.

[0072] Through the above implementation methods, the platform can determine target maintenance resources based on comprehensive dispatch scoring and dispatch constraints, and generate collaborative maintenance resource combinations in complex task scenarios. This ensures that dispatch results take into account skill coverage, response timeliness, resource readiness, and on-site executability. Furthermore, by outputting dispatch criteria, the dispatch results are interpretable and traceable, helping to reduce the need for repeated manual confirmation and improving the response efficiency and first-time success rate of hydraulic equipment repair tasks.

[0073] In some embodiments, maintenance execution feedback is obtained, and the hydraulic fault semantic graph, maintenance resource capability profile, and dispatch decision parameters are updated based on the maintenance execution feedback, including: Obtain maintenance execution feedback corresponding to the maintenance task allocation results, and perform structured parsing of the maintenance execution feedback to generate actual fault results, actual maintenance results, and execution quality representations; The actual fault results are compared with the fault semantic understanding results, and the fault association weights in the hydraulic fault semantic map are updated based on the comparison results. Based on actual maintenance results and performance quality characteristics, revise the capability parameters of the target maintenance resources in the corresponding failure modes and update the capability profile of the maintenance resources. Based on the updated fault association weights and capability parameters, feedback correction is performed on the dispatch decision parameters in the multi-objective intelligent dispatch model.

[0074] Specifically, the intelligent dispatch platform continuously acquires maintenance execution feedback corresponding to the maintenance task allocation results during and after the maintenance task is executed. Taking the maintenance task of a hydraulic pressing system of a metallurgical rolling mill as an example, the platform previously determined the suspected fault modes as proportional valve core jamming and filter unit blockage based on the fault semantic understanding results, and dispatched maintenance team A as the target maintenance resource to the site. After the maintenance personnel arrived, they recorded the order acceptance time, arrival time, pressure test results, valve assembly disassembly and inspection results, filter element differential pressure detection results, actual replaced parts, system recovery time, and retest status through a mobile maintenance terminal. The platform simultaneously receives the pressure curves before and after maintenance, valve position response curves, oil temperature change curves, and action cycle recovery data transmitted back from the equipment monitoring system, and uses the manually entered content, sensor retest data, and work order closure information as maintenance execution feedback.

[0075] The platform performs structured analysis on maintenance execution feedback, generating actual fault results, actual maintenance results, and execution quality representations. Actual fault results represent the fault objects and modes ultimately confirmed on-site. For example, in this task, disassembly and inspection confirmed a slight sticking of the proportional valve spool, while the return oil filter was severely clogged, although the hydraulic pump's pressure supply capacity was normal. Actual maintenance results represent the actual maintenance actions performed and resource consumption, such as cleaning the proportional valve, replacing the filter element, venting the system, and recalibrating the pressure. Execution quality representations represent the effectiveness of the maintenance task, such as whether on-site arrival was delayed, whether the first repair was successful, whether the retested parameters returned to the baseline range, whether secondary alarms occurred, and whether spare parts were temporarily requisitioned. The platform correlates these structured results with the original maintenance task allocation results to form feedback samples that can be used for model reprocessing.

[0076] Subsequently, the platform performs a consistency comparison between the actual fault results and the fault semantic understanding results. If the high-confidence fault modes in the fault semantic understanding results are consistent with the on-site confirmation results, the platform increases the association weights between the corresponding fault phenomena, operational anomalies, and fault modes in the hydraulic fault semantic map. For example, the association weight between slow pressing action, lag in valve position response, and proportional valve spool sticking is increased, as is the association weight between oil temperature rise, abnormal return oil pressure difference, and filter unit blockage.

[0077] If the efficiency decline of a previously ranked hydraulic pump is not confirmed on-site, the platform reduces the correlation weight between pressure establishment slowness and internal leakage of the hydraulic pump under the same operating conditions, and retains the comparison result as evidence for graph correction. For fault patterns that appear in actual fault results but have low confidence in the original fault semantic understanding results, the platform marks the fault pattern as a supplementary inference sample for subsequent expansion of similar fault paths.

[0078] When updating the hydraulic fault semantic graph, the platform also verifies the association weights based on the post-repair operational recovery effects. If the return oil pressure differential decreases, the oil temperature recovers, and the action cycle shortens after replacing the filter element, it indicates a strong causal relationship between the filter unit blockage and the current anomaly, and the platform increases the credibility level of this path. If the valve position feedback lag disappears after the proportional valve is cleaned, the platform further increases the weight of the path corresponding to the proportional valve jamming. For tasks where pressure fluctuations persist after repair, the platform does not directly confirm the original reasoning is correct, but instead marks the task as a sample to be reviewed to avoid erroneous feedback contaminating the graph.

[0079] Simultaneously, the platform adjusts the capability parameters of the target maintenance resources under the corresponding fault modes based on actual maintenance results and execution quality characteristics. If maintenance team A arrives within the specified time window and completes proportional valve cleaning, filter replacement, and system retesting in one go, the platform increases the team's capability parameters in proportional valve response anomalies, filter blockage, and hydraulic system recovery and debugging. If insufficient spare parts preparation, excessively long diagnosis time, or recurrence shortly after maintenance occurs, the platform reduces the repair reliability parameters under the corresponding maintenance actions and records the specific factors causing the deduction. For collaborative maintenance tasks, the platform evaluates the contributions of the lead maintenance personnel and collaborators separately. For example, if the lead maintenance personnel complete fault diagnosis and pressure testing, and the collaborators complete filter replacement and system venting, the platform adjusts the capability profiles of different personnel according to the execution records.

[0080] Finally, based on the updated fault association weights and capability parameters, the platform provides feedback correction to the dispatch decision parameters in the multi-objective intelligent dispatch model. If multiple feedbacks indicate that arrival time has a more significant impact on recovery time in high-urgency hydraulic shutdown tasks, the platform increases the weight of response time. If the ability to handle similar historical faults contributes more to the first-time repair rate in complex valve control tasks, the platform increases the weights of skill adaptation and repair reliability. If insufficient spare parts preparation frequently leads to maintenance delays, the platform increases the strength of resource accessibility constraints.

[0081] Through the above feedback and update process, the platform can transform the results of each maintenance execution into a basis for map updates, personnel capability corrections, and dispatch parameter calibrations, making fault identification more accurate, resource matching more reasonable, and dispatch decisions more reliable for subsequent similar hydraulic equipment maintenance tasks.

[0082] The following are system embodiments of this application, which can be used to execute the method embodiments of this application. For details not disclosed in the system embodiments of this application, please refer to the method embodiments of this application.

[0083] Figure 2 This is a schematic diagram of the intelligent dispatch system for the maintenance of hydraulic equipment based on semantic profiling and multi-constraints, provided in an embodiment of this application. Figure 2 As shown, the system includes: The acquisition module 201 is used to acquire multi-source maintenance data associated with the hydraulic equipment maintenance tasks to be dispatched, and to standardize the multi-source maintenance data to obtain unified maintenance data. The parsing module 202 is used to fuse and parse the repair semantics, operational anomalies and historical maintenance information in the unified maintenance data based on the hydraulic fault semantic graph and the domain knowledge enhanced semantic model, and generate fault semantic understanding results that represent fault objects, suspected fault modes and maintenance needs. The generation module 203 is used to generate a maintenance task semantic profile that includes task urgency, fault complexity, skill requirements and resource requirements based on the relationship between the fault semantic understanding results and the hydraulic equipment structure. The determination module 204 is used to generate a maintenance resource capability profile based on the historical maintenance feedback, real-time operation status and available resource conditions of maintenance resources, and to perform constraint matching based on the maintenance task semantic profile and the maintenance resource capability profile to determine the candidate maintenance resource set. Evaluation module 205 is used to evaluate the candidate maintenance resource set by using a multi-objective intelligent dispatch model, generate a comprehensive dispatch score, determine the target maintenance resource based on the comprehensive dispatch score, and generate maintenance task allocation results carrying the dispatch basis. The update module 206 is used to obtain maintenance execution feedback and update the hydraulic fault semantic graph, maintenance resource capability profile and dispatch decision parameters based on the maintenance execution feedback.

[0084] In some embodiments, Figure 2 The acquisition module 201 parses the task identifier and equipment identifier in the hydraulic equipment maintenance task to be dispatched, and obtains the task-side data, equipment-side data and resource-side data associated with the maintenance task based on the preset data mapping relationship; according to the unified time base and equipment structure index, it performs association and alignment of multi-source maintenance data, and performs semantic normalization processing on unstructured data based on the semantic rules of the hydraulic field; it performs quality verification, source marking and credibility encoding on the association and alignment of multi-source maintenance data, and generates unified maintenance data carrying task association relationship, equipment association relationship and data credibility representation.

[0085] In some embodiments, Figure 2The parsing module 202 extracts semantic features of maintenance tasks and abnormal equipment operation features from unified maintenance data, and maps the extraction results to equipment structure nodes and fault mechanism nodes in the hydraulic fault semantic graph. It uses domain knowledge to enhance the semantic model and performs contextual semantic encoding on the mapped features to generate fault semantic representations corresponding to the hydraulic equipment maintenance tasks to be dispatched. Based on the association paths in the hydraulic fault semantic graph, it performs graph constraint reasoning and historical similar maintenance evidence fusion on the fault semantic representations to determine the matching confidence between fault objects, suspected fault modes and maintenance needs. It generates fault semantic understanding results based on the matching confidence.

[0086] In some embodiments, Figure 2 The generation module 203 maps the fault semantic understanding results to the corresponding equipment units and hydraulic functional units in the structural relationship of the hydraulic equipment, and determines the structural association range of the object to be repaired; based on the structural association range, fault propagation relationship and operating impact state, it calculates the task urgency and fault complexity of the hydraulic equipment repair task to be dispatched; based on the semantic association between suspected fault modes and repair actions, it determines the skill requirements and resource requirements corresponding to the hydraulic equipment repair task to be dispatched; and it performs profile encoding on the task urgency, fault complexity, skill requirements and resource requirements to generate a semantic profile of the repair task.

[0087] In some embodiments, Figure 2 The determination module 204 extracts capability performance features associated with hydraulic equipment failure modes and maintenance actions based on historical maintenance feedback of maintenance resources, and generates a maintenance resource capability profile by combining real-time operation status and available resource conditions; constructs constraint matching conditions for the maintenance resource capability profile according to the task constraint requirements in the maintenance task semantic profile; uses the constraint matching conditions to adapt and filter the maintenance resource capability profile, excluding maintenance resources that do not meet the basic conditions for dispatching; and performs capability similarity and resource reachability matching on the filtered maintenance resources to determine the candidate maintenance resource set.

[0088] In some embodiments, Figure 2 The evaluation module 205 inputs the semantic profile of the maintenance task and the capability profile of the maintenance resources corresponding to the candidate maintenance resource set into the multi-objective intelligent dispatch model to generate task resource adaptation features; based on the urgency and fault complexity of the hydraulic equipment maintenance task to be dispatched, it determines the dynamic weight relationship between dispatch evaluation objectives; based on the task resource adaptation features and dynamic weight relationship, it conducts a collaborative evaluation of skill adaptation, response timeliness, resource accessibility and repair reliability for each candidate maintenance resource to generate dispatch adaptation value of the candidate maintenance resource; and it performs constraint correction on the dispatch adaptation value according to the preset constraint correction rules to generate a comprehensive dispatch score.

[0089] In some embodiments, Figure 2The evaluation module 205 performs isomorphic encoding on the semantic profile of the maintenance task and the profile of the maintenance resource capability to obtain the task requirement vector and the resource capability vector. Based on the cross-profile attention mechanism in the multi-objective intelligent dispatch model, the task requirement vector and the resource capability vector are interactively matched to extract the correlation response between the task requirement and the resource capability. Based on the correlation response, task resource adaptation features representing the degree of capability coverage, the degree of resource satisfaction and the degree of execution adaptation are generated.

[0090] In some embodiments, Figure 2 The evaluation module 205 sorts the candidate maintenance resource set according to the comprehensive dispatch score and determines the target maintenance resource that meets the allocation conditions based on the dispatch constraints of the hydraulic equipment maintenance task to be dispatched. When the semantic profile of the maintenance task indicates that there is a need for collaborative maintenance, a collaborative maintenance resource combination is generated based on the capability complementarity relationship and operation time window matching relationship between the candidate maintenance resources. Based on the target maintenance resource or collaborative maintenance resource combination, a maintenance task allocation result containing task execution object, expected response status and resource preparation information is generated. The dispatch basis is generated according to the contribution characteristics of the comprehensive dispatch score, and the dispatch basis is associated with the maintenance task allocation result and output.

[0091] In some embodiments, Figure 2 The update module 206 acquires the maintenance execution feedback corresponding to the maintenance task allocation result, performs structured parsing on the maintenance execution feedback, and generates actual fault results, actual maintenance results, and execution quality representations; it compares the actual fault results with the fault semantic understanding results for consistency, and updates the fault association weights in the hydraulic fault semantic graph based on the comparison results; it corrects the capability parameters of the target maintenance resources in the corresponding fault modes based on the actual maintenance results and execution quality representations, and updates the maintenance resource capability profile; based on the updated fault association weights and capability parameters, it performs feedback correction on the dispatch decision parameters in the multi-objective intelligent dispatch model.

[0092] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0093] Figure 3 This is a schematic diagram of the electronic device 3 provided in an embodiment of this application. Figure 3 As shown, the electronic device 3 of this embodiment includes: a processor 301, a memory 302, and a computer program 303 stored in the memory 302 and executable on the processor 301. When the processor 301 executes the computer program 303, it implements the steps in the various method embodiments described above. Alternatively, when the processor 301 executes the computer program 303, it implements the functions of each module / unit in the various system embodiments described above.

[0094] Electronic device 3 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 3 may include, but is not limited to, processor 301 and memory 302. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or different components.

[0095] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0096] The memory 302 can be an internal storage unit of the electronic device 3, such as a hard disk or memory of the electronic device 3. The memory 302 can also be an external storage device of the electronic device 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 3. The memory 302 can also include both internal and external storage units of the electronic device 3. The memory 302 is used to store computer programs and other programs and data required by the electronic device.

[0097] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0098] If integrated modules / units are implemented as software functional units and sold or used as independent products, they can be stored in a readable storage medium (e.g., a computer-readable storage medium). Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program may include computer program code, which may be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium may include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0099] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for intelligent dispatching of maintenance orders for hydraulic equipment based on semantic profiling and multi-constraints, characterized in that, include: Obtain multi-source maintenance data associated with hydraulic equipment maintenance tasks pending dispatch, and standardize the multi-source maintenance data to obtain unified maintenance data; Based on the hydraulic fault semantic graph and the domain knowledge-enhanced semantic model, the repair reporting semantics, operational anomalies and historical maintenance information in the unified maintenance data are fused and analyzed to generate fault semantic understanding results that characterize fault objects, suspected fault modes and maintenance needs. Based on the relationship between the fault semantic understanding results and the hydraulic equipment structure, a maintenance task semantic profile is generated, which includes task urgency, fault complexity, skill requirements, and resource requirements. Based on historical maintenance feedback, real-time operation status, and available resource conditions, a maintenance resource capability profile is generated. Then, constraint matching is performed between the maintenance task semantic profile and the maintenance resource capability profile to determine a candidate maintenance resource set. The candidate maintenance resource set is evaluated for dispatch adaptation using a multi-objective intelligent dispatch model, a comprehensive dispatch score is generated, and the target maintenance resource is determined based on the comprehensive dispatch score, thereby generating a maintenance task allocation result carrying the dispatch basis. Obtain maintenance execution feedback, and update the hydraulic fault semantic graph, maintenance resource capability profile, and dispatch decision parameters based on the maintenance execution feedback.

2. The method according to claim 1, characterized in that, The process of acquiring multi-source maintenance data associated with hydraulic equipment maintenance tasks awaiting dispatch and standardizing the multi-source maintenance data to obtain unified maintenance data includes: Analyze the task identifier and equipment identifier in the hydraulic equipment maintenance task to be dispatched, and obtain the task-side data, equipment-side data and resource-side data associated with the maintenance task based on the preset data mapping relationship; The multi-source maintenance data are correlated and aligned according to a unified time base and equipment structure index, and the unstructured data is semantically normalized based on the semantic rules of the hydraulic field. After the multi-source maintenance data is aligned and associated, quality verification, source labeling, and credibility encoding are performed to generate unified maintenance data that carries task association, equipment association, and data credibility representation.

3. The method according to claim 1, characterized in that, The hydraulic fault semantic graph and domain knowledge-enhanced semantic model are used to fuse and analyze the reported repair semantics, operational anomalies, and historical maintenance information in the unified maintenance data, generating fault semantic understanding results that characterize fault objects, suspected fault modes, and maintenance needs, including: Semantic features of maintenance tasks and abnormal features of equipment operation are extracted from the unified maintenance data, and the extraction results are mapped to equipment structure nodes and fault mechanism nodes in the hydraulic fault semantic map. By leveraging domain knowledge to enhance the semantic model, contextual semantic encoding is performed on the mapped features to generate fault semantic representations corresponding to the hydraulic equipment maintenance tasks to be dispatched. Based on the associated paths in the hydraulic fault semantic graph, graph-constrained reasoning and historical similar maintenance evidence fusion are performed on the fault semantic representation to determine the matching confidence between the fault object, suspected fault mode and maintenance requirement. The fault semantic understanding result is generated based on the matching confidence.

4. The method according to claim 1, characterized in that, The process involves generating a maintenance task semantic profile based on the relationship between the fault semantic understanding results and the hydraulic equipment structure. This profile includes information on task urgency, fault complexity, skill requirements, and resource requirements. The fault semantic understanding results are mapped to the corresponding equipment units and hydraulic functional units in the structural relationship of hydraulic equipment to determine the structural association range of the object to be repaired. Based on the structural association range, fault propagation relationship, and operational impact status, calculate the task urgency and fault complexity of the hydraulic equipment maintenance task to be dispatched. Based on the semantic association between suspected fault modes and maintenance actions, the skill and resource requirements corresponding to the hydraulic equipment maintenance tasks to be dispatched are determined. The task urgency, fault complexity, skill requirements, and resource requirements are profiled and encoded to generate a semantic profile of the maintenance task.

5. The method according to claim 1, characterized in that, The process involves generating a maintenance resource capability profile based on historical maintenance feedback, real-time operational status, and available resource conditions. Then, a constraint matching process is performed between the maintenance task semantic profile and the maintenance resource capability profile to determine a candidate maintenance resource set, including: Based on historical maintenance feedback of maintenance resources, capability performance characteristics associated with hydraulic equipment failure modes and maintenance actions are extracted, and a maintenance resource capability profile is generated by combining real-time operating status and available resource conditions. Based on the task constraint requirements in the semantic profile of the maintenance task, construct constraint matching conditions for the profile of maintenance resource capabilities. The aforementioned constraint matching conditions are used to adapt and filter the maintenance resource capability profile, excluding maintenance resources that do not meet the basic conditions for dispatching orders. The selected maintenance resources are matched based on capability similarity and resource accessibility to determine the candidate maintenance resource set.

6. The method according to claim 1, characterized in that, The step of using a multi-objective intelligent dispatch model to evaluate the dispatch adaptation of the candidate maintenance resource set and generate a comprehensive dispatch score includes: Input the semantic profile of the maintenance task and the maintenance resource capability profile corresponding to the candidate maintenance resource set into the multi-objective intelligent dispatch model to generate task resource adaptation features. Based on the urgency and complexity of the hydraulic equipment repair tasks to be dispatched, determine the dynamic weight relationship between the dispatch evaluation objectives; Based on the task resource adaptation characteristics and dynamic weight relationship, a collaborative evaluation of skill adaptation, response time, resource availability and repair reliability is performed on each candidate maintenance resource to generate a dispatch adaptation value for the candidate maintenance resource. The dispatch adaptation value is constrained and corrected according to the preset constraint correction rules to generate a comprehensive dispatch score.

7. The method according to claim 6, characterized in that, The step of inputting the semantic profile of the maintenance task and the maintenance resource capability profile corresponding to the candidate maintenance resource set into the multi-objective intelligent dispatch model to generate task resource adaptation features includes: The semantic profile of the maintenance task and the profile of the maintenance resource capability are homogeneously encoded to obtain the task requirement vector and the resource capability vector. Based on the cross-profile attention mechanism in the multi-objective intelligent dispatch model, the task requirement vector and resource capability vector are interactively matched to extract the correlation response between task requirements and resource capabilities. Based on the associated response, task resource adaptation features are generated to represent the coverage of capabilities, the degree of resource satisfaction, and the degree of execution adaptability.

8. The method according to claim 6, characterized in that, The step of determining the target maintenance resource based on the comprehensive dispatch score and generating a maintenance task allocation result carrying the dispatch basis includes: The candidate maintenance resource set is sorted according to the comprehensive dispatch score, and the target maintenance resource that meets the allocation conditions is determined in combination with the dispatch constraints of the hydraulic equipment maintenance task to be dispatched. When the semantic profile of the maintenance task indicates a need for collaborative maintenance, a collaborative maintenance resource combination is generated based on the complementary capabilities and matching time windows of the candidate maintenance resources. Based on the target maintenance resources or collaborative maintenance resource combination, a maintenance task allocation result containing task execution objects, expected response status and resource preparation information is generated. The dispatch criteria are generated based on the contribution characteristics of the comprehensive dispatch score, and the dispatch criteria are associated with the maintenance task allocation results and output.

9. The method according to claim 1, characterized in that, The step of obtaining maintenance execution feedback and updating the hydraulic fault semantic graph, maintenance resource capability profile, and dispatch decision parameters based on the maintenance execution feedback includes: Obtain the maintenance execution feedback corresponding to the maintenance task allocation result, and perform structured parsing on the maintenance execution feedback to generate actual fault results, actual maintenance results, and execution quality characterization; The actual fault results are compared with the fault semantic understanding results, and the fault association weights in the hydraulic fault semantic map are updated based on the comparison results. Based on the actual maintenance results and performance characterization, the capability parameters of the target maintenance resources in the corresponding fault modes are corrected, and the capability profile of the maintenance resources is updated. Based on the updated fault association weights and capability parameters, feedback correction is performed on the dispatch decision parameters in the multi-objective intelligent dispatch model.

10. A smart dispatching system for the maintenance of hydraulic equipment based on semantic profiling and multi-constraints, characterized in that, include: The acquisition module is used to acquire multi-source maintenance data associated with hydraulic equipment maintenance tasks to be dispatched, and to standardize the multi-source maintenance data to obtain unified maintenance data. The parsing module is used to fuse and parse the repair reporting semantics, operational anomalies, and historical maintenance information in the unified maintenance data based on the hydraulic fault semantic graph and the domain knowledge-enhanced semantic model, and generate fault semantic understanding results that characterize fault objects, suspected fault modes, and maintenance needs. The generation module is used to generate a maintenance task semantic profile that includes task urgency, fault complexity, skill requirements, and resource requirements based on the fault semantic understanding results and the relationship between the hydraulic equipment structure. The determination module is used to generate a maintenance resource capability profile based on the historical maintenance feedback, real-time operation status and available resource conditions of the maintenance resources, and to perform constraint matching between the maintenance task semantic profile and the maintenance resource capability profile to determine a set of candidate maintenance resources. The evaluation module is used to evaluate the candidate maintenance resource set by using a multi-objective intelligent dispatch model, generate a comprehensive dispatch score, determine the target maintenance resource based on the comprehensive dispatch score, and generate a maintenance task allocation result carrying the dispatch basis. The update module is used to obtain maintenance execution feedback and update the hydraulic fault semantic graph, maintenance resource capability profile, and dispatch decision parameters based on the maintenance execution feedback.

Images