Intelligent workbench material management and operation and maintenance method and system

Abstract

The application relates to the technical field of industrial automation and intelligent manufacturing, and particularly relates to a kind of intelligent workstation material control and operation and maintenance method and system.The method comprises: constructing a digital twin model of the intelligent workstation;based on the model, the production plan, material quality and supply chain data are fused to predict material demand and trigger replenishment;the key equipment working parameters are synchronized, the health score calculation and fault mode matching are carried out by using the abnormal detection algorithm and the equipment health knowledge base, and predictive maintenance is realized;when a fault warning is diagnosed, a linkage plan is generated and simulated in the digital twin model, the optimal scheme is selected and executed to adjust the associated equipment state and task.The application realizes active prediction of material demand, early diagnosis and root cause analysis of equipment failure, and safe and intelligent collaborative response across units, effectively avoiding production interruption and reducing unexpected downtime, and significantly improving the overall equipment efficiency, operation safety and flexibility of the workstation.

Description

Technical Field

[0001] This invention relates to the field of industrial automation and intelligent manufacturing technology, specifically to a method and system for material management and maintenance of an intelligent workbench. Background Technology

[0002] In current discrete manufacturing fields, such as electronic assembly and precision instrument assembly, intelligent workbenches serve as core production units, and their operational efficiency and stability directly impact overall output. However, existing workbench material management and equipment maintenance methods generally face the dilemmas of passive response, information silos, and a lack of collaboration. Regarding material management, most systems rely on setting fixed safety stock thresholds for alarms, a typical "hindsight" approach. This method cannot cope with changes in consumption rates caused by sudden changes in production plans, batch differences in material quality, or supply chain fluctuations, easily leading to unplanned downtime due to material shortages or capital tied up in inventory. In terms of equipment maintenance, it mainly relies on scheduled maintenance and post-failure repair. Scheduled maintenance lacks specificity, potentially leading to over-maintenance or under-maintenance; while post-failure repair results in prolonged unexpected downtime and significant production losses. Although some advanced solutions introduce anomaly detection based on sensor data, most can only provide fault alarms, failing to perform early fault prediction, precise location, and root cause analysis. Maintenance decisions still heavily rely on individual engineer experience, making knowledge difficult to accumulate and reuse. More importantly, material flow and equipment status flow are usually managed by independent systems, lacking deep collaboration. When a critical piece of equipment experiences a fault warning, the material handling system is unable to detect it and adjust the delivery plan in a timely manner. Simultaneously, the production line control system struggles to automatically, safely, and optimally execute emergency operations such as task reassignment and robot path adjustments. The entire system lacks resilience and adaptability. Therefore, there is an urgent need for an integrated intelligent workbench management and maintenance solution capable of proactively predicting material demand, diagnosing equipment health status in advance, and enabling intelligent collaborative decision-making and execution across units.

[0003] Therefore, existing technologies still need further development. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned technical deficiencies and provide an intelligent workbench material management and maintenance method and system to solve the problems existing in the prior art.

[0005] To achieve the above-mentioned technical objectives, according to a first aspect of the present invention, the present invention provides a method for intelligent workbench material management and maintenance, comprising: S1. Construct a digital twin model that includes the physical entity of the intelligent workbench and its operating environment mapping; S2. Based on the digital twin model, predictive management of workbench material consumption is performed, and material scheduling instructions are generated based on the management results. S3. Based on the digital twin model, perform health status diagnosis and predictive maintenance on key equipment of the workbench; S4. When the diagnostic results meet the preset conditions, trigger the linkage operation and maintenance command to adjust the operating status or tasks of the associated equipment.

[0006] Specifically, the predictive management of workbench material consumption and the generation of material scheduling instructions based on the management results include: The system acquires real-time inventory data, consumption rate, and production plan data from the manufacturing execution system for materials on the workbench; it also acquires material-related quality perception data and supply chain reliability data, where the quality perception data reflects key characteristic indicators of material batches and the supply chain reliability data reflects the historical fulfillment stability of material supply sources; it inputs the inventory data, consumption rate, production plan data, quality perception data, and supply chain reliability data into a demand forecasting model to predict the material demand gap within a specified future time window; when the predicted demand gap exceeds a first threshold, it automatically generates and sends a replenishment request instruction containing material identification, required quantity, and required time to the warehouse management system or supplier system.

[0007] Specifically, the health status diagnosis and predictive maintenance of key equipment on the workbench includes: A corresponding digital twin sub-model is constructed for the key equipment, and its operating parameters are synchronized to the sub-model in real time. An equipment health knowledge base associated with equipment type and process is established, which stores feature data, maintenance records, and performance degradation patterns of historical failure cases. Based on the real-time operating parameters, the instant health score of the equipment is calculated using a preset anomaly detection algorithm. The real-time operating parameters and historical parameter sequences are matched and similarity calculated with the data in the equipment health knowledge base to perform fault mode identification and root cause reasoning, and to generate diagnostic reports and maintenance recommendations. When the instant health score is lower than a second threshold or a specific fault mode is identified, a maintenance warning is generated that includes maintenance content, suggested execution time, and estimated time consumption.

[0008] Specifically, the method also includes a knowledge base evolution step: After each maintenance operation is completed, the actual operation data, replacement component information, and post-maintenance equipment performance verification data are collected. This maintenance event is treated as a new case, processed in a standardized manner, and then stored in the equipment health knowledge base to optimize subsequent fault mode identification and root cause reasoning.

[0009] Specifically, the step of calculating the device's instantaneous health score based on the real-time operating parameters using a preset anomaly detection algorithm includes: Differentiated health assessment models are configured for different types of key equipment. The assessment model defines a set of key parameters and their weights for calculating the health score. For online learning-enabled assessment models, the model parameters are continuously fine-tuned using historical normal state data of the current equipment and similar equipment to adapt to individual differences in equipment and slow performance drift. The real-time health score is generated by combining the output of the anomaly detection algorithm and the weighted statistical analysis results based on the set of key parameters.

[0010] Specifically, the triggering of the coordinated operation and maintenance command to adjust the operating status or tasks of the associated devices includes: In the digital twin model, based on the diagnostic results and preset linkage strategy rules, one or more preliminary linkage control plans are generated. The plans include operation instructions for the target equipment and path planning instructions for the collaborative robot. The preliminary linkage control plans are simulated and executed in the digital twin model, and the impact of each plan on predefined production efficiency indicators is evaluated based on the simulation results. From the one or more plans, the plan with the best comprehensive evaluation result is selected as the final linkage operation and maintenance instruction. The final linkage operation and maintenance instruction is issued to the corresponding physical entity equipment for execution.

[0011] Specifically, the linkage strategy rules include at least an equipment failure early warning strategy and a task switching strategy; the task switching strategy stipulates that when it is predicted that a certain device is about to fail, its unfinished task queue will be switched to a backup device, wherein the switching decision needs to take into account the current load, task processing capacity, and compatibility of the required materials of the backup device.

[0012] Specifically, after constructing the digital twin model, the process also includes a knowledge base construction step: By extracting events, features, and solutions related to workbench equipment failures, abnormal material consumption, and production cycle optimization from historical operation and maintenance logs, equipment manuals, and expert experience, a structured operation and maintenance knowledge graph is formed and associated with the corresponding entities in the digital twin model.

[0013] Specifically, the method further includes: real-time monitoring of the workbench's operating status and matching the current operating parameters with the event features in the operation and maintenance knowledge graph; when the matching degree exceeds the third threshold, actively pushing the associated solutions or historical processing records in the operation and maintenance knowledge graph to the operation and maintenance interface to assist in decision-making.

[0014] According to a second aspect of the present invention, an intelligent workbench material management and maintenance system is provided, comprising: The digital twin construction and management unit is used to build and maintain a digital twin model corresponding to the physical entity of the smart workbench; A predictive material management unit is used to analyze material consumption trends and predict demand based on the digital twin model, and to generate and send material scheduling instructions. The equipment health diagnosis and maintenance unit is used to monitor the operating parameters of key equipment based on the digital twin model, use a knowledge base and algorithms to assess health status and predict faults, and generate maintenance warnings. The linkage operation and maintenance decision and execution unit is used to generate and verify the linkage control plan in the digital twin environment based on the output of the equipment health diagnosis and maintenance unit, and send the optimized instructions to the physical workbench for execution; The predictive material management unit, equipment health diagnosis and maintenance unit, and coordinated operation and maintenance decision and execution unit all interact with the digital twin construction and management unit through data exchange and command collaboration.

[0015] Beneficial effects: The intelligent workbench material management and maintenance method and system provided by this invention bring multi-dimensional and multi-level beneficial effects, realizing a leap from local automation to global intelligence.

[0016] First, at the material supply level, by constructing a multi-factor demand forecasting model that integrates real-time consumption, production planning, material quality perception, and supply chain reliability, a fundamental shift from passive alarms to proactive and accurate forecasting has been achieved. This not only significantly reduces the risk of production interruptions due to material shortages and improves the overall utilization rate of equipment, but also reduces capital occupation through more precise inventory control and proactively addresses material quality fluctuations and supply risks, thereby enhancing supply chain resilience.

[0017] Secondly, regarding equipment reliability, by establishing digital twin models for equipment and combining machine learning-based anomaly detection with knowledge graph-based case reasoning, a leap from "perceiving anomalies" to "diagnosing faults and recommending solutions" has been achieved. This system can capture early trends in equipment performance degradation and predict potential failure points, transforming unplanned downtime into predictive maintenance that can be scheduled during off-production periods, significantly reducing emergency repair costs and production losses. Simultaneously, the operations and maintenance knowledge base possesses self-evolving capabilities, continuously accumulating and reusing maintenance experience, reducing reliance on senior experts, and improving the overall skill level and efficiency of the operations and maintenance team.

[0018] Third, at the system collaboration and security level, a collaborative operation and maintenance decision-making mechanism based on a digital twin model enables global intelligent collaboration across equipment, materials, and robots. When equipment failure is predicted, the system can simulate and optimize various contingency plans in virtual space, and automatically execute the safest solution with the least impact on overall production efficiency (such as intelligent task switching and dynamic path planning). This avoids secondary problems or safety accidents caused by improper handling of local failures, and greatly enhances the adaptability, flexibility, and operational resilience of the entire workbench and even the production line.

[0019] Finally, from a system architecture perspective, using a digital twin model as a unified data hub and simulation sandbox decouples and service-oriented functional modules such as prediction, diagnosis, and decision-making, giving the system high flexibility, scalability, and maintainability, thus laying a solid technical foundation for building an adaptive and self-optimizing intelligent production system. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the intelligent workbench material management and maintenance method provided in a specific embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Based on the embodiments in this application, other similar embodiments obtained by those skilled in the art without creative effort should all fall within the scope of protection of this application. Furthermore, directional terms mentioned in the following embodiments, such as "up," "down," "left," and "right," are only for reference to the directions in the accompanying drawings; therefore, the directional terms used are for illustrative purposes and not for limiting the invention.

[0022] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments.

[0023] Please see Figure 1 This invention provides a method for material management and maintenance of an intelligent workbench, comprising: S1. Construct a digital twin model that includes the physical entity of the intelligent workbench and its operating environment mapping.

[0024] It should be further explained that this method provides a complete, closed-loop intelligent workbench collaborative management and operation framework. Its specific implementation first relies on a precise 3D model of the physical workbench (which can be built using Unity3D, UE4, or a professional industrial digital twin platform). Then, through IoT gateways, PLCs, and equipment controllers, it collects real-time status data (such as coordinates, speed, current, pressure, and material level) from all intelligent electric screwdrivers, dispensing machines, collaborative robots, vision cameras, and material sensors within the workbench. Using protocols such as OPCUA and MQTT, the data is synchronized to the virtual model, thereby constructing a digital twin model that is "synchronized" with the physical world. This model serves as the data hub and simulation sandbox for all subsequent advanced functions.

[0025] S2. Based on the digital twin model, predictive management of workbench material consumption is performed, and material scheduling instructions are generated based on the management results.

[0026] It should be further explained that the specific implementation of predictive material management is as follows: In this twin model, a logical association is established for each material loading point (such as SMT feeder, vibratory feeder, and pallet), and its remaining quantity or weight is monitored in real time. Combined with future production orders synchronized from MES (Manufacturing Execution System) (including product model, quantity, and planned start time), the material demand in the next few hours is dynamically calculated through built-in predictive algorithms (such as the fusion model detailed later), and precise replenishment instructions are automatically generated and sent to WMS (Warehouse Management System) or AGV scheduling system.

[0027] S3. Based on the digital twin model, perform health status diagnosis and predictive maintenance on key equipment of the workbench.

[0028] It should be further explained that the specific implementation of equipment health diagnosis and predictive maintenance is as follows: a dedicated digital twin model is established for each key process equipment in the workbench (such as tightening shafts and dispensing valves), and its real-time operating parameters (current, temperature, vibration spectrum, etc.) are continuously fed in. The health status is calculated using an anomaly detection model based on machine learning (such as isolated forests and autoencoders), and pattern matching is performed in combination with a continuously evolving fault knowledge base. In this way, when the equipment performance deteriorates significantly or specific fault precursors occur, an early warning work order containing fault type, possible causes, recommended maintenance time and content is generated.

[0029] S4. When the diagnostic results meet the preset conditions, trigger the linkage operation and maintenance command to adjust the operating status or tasks of the associated equipment.

[0030] It should be further explained that the specific implementation of triggering the linkage operation and maintenance command is as follows: When a high-level maintenance warning is received, the system generates multiple alternative linkage plans in the digital twin environment based on the preset linkage strategy rules (such as "if the health of axis A is lower than the threshold X, then migrate its task to the backup axis B and replan the robot path"). The system also evaluates the impact of each plan on the overall production efficiency and safety through rapid simulation. After selecting the optimal plan, the system sends specific control commands (such as stop commands, task reassignment commands, and new path points) to the actual PLC, robot controller and equipment through a safety protocol to achieve a safe and smooth emergency response.

[0031] Understandably, the core benefit of this method lies in its fundamental shift from the traditional "perception-alarm-manual handling" model to an intelligent closed-loop model of "perception-prediction-simulation decision-automatic execution." By constructing a high-fidelity digital twin, it achieves transparent, visualized, and in-depth insight into the physical workbench's status. Predictive material management transforms material replenishment from a passive response (material shortage alarm) to proactive preparation, effectively avoiding production line downtime due to waiting for materials and improving equipment utilization (OEE). Predictive maintenance allows for intervention before equipment failures occur, transforming unplanned downtime into planned, preventative maintenance that can be scheduled during non-production periods, significantly reducing the impact of sudden failures on production plans and potential maintenance costs. Most importantly, the collaborative operation and maintenance mechanism ensures that the instructions executed in the physical space are optimal and safe by pre-simulating and evaluating various response strategies in virtual space, avoiding secondary problems or safety accidents caused by improper handling of localized failures, and greatly enhancing the resilience, flexibility, and intelligence level of the entire workbench and even the production line.

[0032] Specifically, the predictive management of workbench material consumption and the generation of material scheduling instructions based on the management results include: real-time acquisition of workbench material inventory data, consumption rate, and production plan data from the manufacturing execution system; acquisition of material-related quality perception data and supply chain reliability data, wherein the quality perception data reflects key characteristic indicators of material batches, and the supply chain reliability data reflects the historical fulfillment stability of material supply sources; inputting the inventory data, consumption rate, production plan data, quality perception data, and supply chain reliability data into a demand forecasting model to predict the material demand gap within a specified future time window; and automatically generating and sending a replenishment request instruction containing material identification, required quantity, and required time to the warehouse management system or supplier system when the predicted demand gap exceeds a first threshold.

[0033] It should be further noted that the implementation of predictive materials management in this step is extremely specific and in-depth, with specific designs including: (1) Specific implementation of data acquisition: Inventory data is acquired in real time through RFID antennas (reading material tray tags) or high-precision laser rangefinders (measuring material pile height) installed in each material slot, with an accuracy of ±1 material or ±0.5mm. Consumption rate is calculated by statistically analyzing the number of times materials are taken within a unit of time (e.g., every 5 minutes) (triggered by photoelectric sensors) or weight changes (measured by weighing sensors), and then smoothed into a consumption per minute. Production plan data is synchronized periodically (e.g., every 15 minutes) through the standard interface of MES (e.g., RESTAPI) to obtain the workbench production task queue that will be launched in the next 8-24 hours. Quality perception data comes from the upstream incoming quality control (IQC) system. The inspection report of the batch of materials currently in use on the workbench is obtained through the interface. Key performance indicators (KPIs) (e.g., chip pin coplanarity, solder paste viscosity, adhesive curing time) are quantified into a quality score of 0-100. Supply chain reliability data Then, it retrieves the data from the Supplier Relationship Management (SRM) system and calculates the on-time delivery rate (OTD) for a specific supplier over the past M months (e.g., M=12) for that material. The formula is as follows: On-time delivery is defined as the actual delivery date being no later than the promised date by T days (e.g., T=1).

[0034] (2) Specific training and implementation of the demand forecasting model: The model adopts a deep learning model that integrates attention mechanism and long short-term memory network (LSTM). The input of the model is a multivariate time series feature vector. ,in It is the real-time inventory level at time t. It is the average consumption rate N minutes before time t (e.g., N=60). It is the future The material requirements for planned production within a time window (e.g., 8 hours) (derived from MES planning breakdown). This is the quality score of the current batch of materials. This is the current reliability score of the material supplier. The model output is a prediction of the future. Inventory level at any given time The model is trained using historical data (including normal and abnormal consumption), and the loss function is mean squared error (MSE). During training, it is necessary to ensure that the data includes samples of accelerated consumption caused by quality fluctuations (low Q value) and emergency replenishment samples caused by supply delays (low S value) so that the model can learn the impact of these factors. (3) Specific implementation of replenishment triggering and instruction generation: demand gap The calculation formula is ,in This is the preset safety stock level for the material. First threshold. It is usually set to 0, but to reduce frequent restocking, it can be set to a positive number, for example... (Units), meaning it is only triggered when the predicted shortage is greater than 5 units. The reason for choosing this value is to balance inventory costs and replenishment operation costs.

[0035] Furthermore, the replenishment request instruction is generated in JSON format, for example: {"material_id":"RES-0603-10K","required_quantity":150,"requested_delivery_time":"2026-03-04T14:30:00Z","target_workbench":"WS-05"}. This instruction is sent via the Enterprise Service Bus (ESB) or by directly calling the Webhook interface of the WMS system.

[0036] Understandably, this step, by introducing quality perception (Q) and supply chain reliability (S), two dynamic factors often overlooked in traditional Material Requirements Planning (MRP), constructs a predictive model that is closer to production reality and has stronger robustness. Quality perception data The introduction of this feature allows the model to automatically perceive the "health status" of the material itself. For example, when using a batch of solder paste with a lower viscosity, the actual printing consumption may be higher than the standard value, and the model will adjust accordingly. The system automatically adjusts short-term consumption forecasts upwards when the value decreases, thus triggering replenishment in advance and preventing material shortages caused by "hidden" overconsumption. (Supply chain reliability data) The introduction of this multi-factor fusion forecasting method imbues the model with risk buffering capabilities. For suppliers with historically unstable delivery records (low S-value), the model adds an extra "safe lead time" to the calculated theoretical replenishment point, thereby placing orders earlier to hedge against potential delivery delays. This multi-factor fusion forecasting method significantly improves the accuracy and reliability of material support in complex and dynamic production environments, providing key technical support for achieving JIT (Just-in-Time) production and low-inventory operations.

[0037] Specifically, the health status diagnosis and predictive maintenance of key equipment on the workbench includes: constructing a corresponding digital twin sub-model for the key equipment and synchronizing its operating parameters to the sub-model in real time; establishing an equipment health knowledge base associated with equipment type and process, wherein the knowledge base stores feature data, maintenance records, and performance degradation patterns of historical failure cases; calculating the equipment's instant health score based on the real-time operating parameters using a preset anomaly detection algorithm; matching and calculating the similarity between the real-time operating parameters and historical parameter sequences and the data in the equipment health knowledge base to perform fault mode identification and root cause reasoning, generating a diagnostic report and maintenance recommendations; and generating a maintenance warning containing maintenance content, suggested execution time, and estimated time consumption when the instant health score is lower than a second threshold or a specific fault mode is identified.

[0038] It should be further explained that the method specifically includes: (1) Specific implementation of the construction and synchronization of the digital twin sub-model of the equipment: Taking an intelligent servo electric screwdriver as an example, its twin sub-model is not only a three-dimensional visualization model, but also a simulation model containing its dynamics and electrical characteristics. Through the built-in controller of the equipment or an external data acquisition module, its working parameters are collected in real time at a frequency of 100Hz, including: the torque curve of the tightening process. (in (for angle), motor three-phase current Motor winding temperature The spectrum of fuselage vibration acceleration (X, Y, Z axes). This data is published in real time via the OPCUA protocol. The digital twin platform subscribes to these data points and drives the corresponding parameter variables in the twin sub-model to update, achieving millisecond-level virtual-real synchronization.

[0039] (2) Specific implementation of the equipment health knowledge base construction: The knowledge base is stored using a graph database (such as Neo4j). Its node types include "equipment model", "failure mode", "failure characteristics (parameters)", "root cause", "maintenance action", and "spare parts". The relationship types include "equipment_may_fail", "failure_manifests as_characteristic", "failure_caused by_cause", "failure_can be solved by_maintenance action", and "maintenance action_requires_spare parts". For example, a knowledge item can be: (equipment model: ESD-SERVO-2000) - [may happen], (failure mode: torque out of tolerance) - [manifests as], (characteristic: torque curve peak variance > 0.2Nm), (failure mode: torque out of tolerance) - [caused by...], (root cause: clutch wear), (root cause: clutch wear) - [can be solved by...], (maintenance action: replace clutch assembly). The performance degradation mode is stored in the form of a time series template, such as the typical curve where the "torque curve peak variance" corresponding to "clutch wear" slowly increases with the number of tightening cycles.

[0040] (3) Specific implementation of real-time health score calculation: A multivariate anomaly detection method based on Mahalanobis distance is adopted. First, during the stable operation phase after equipment installation and commissioning, a large number of multidimensional parameter samples (such as peak torque, effective current, temperature, total vibration, etc.) under normal conditions are collected (e.g., 10,000 cycles) to form a normal dataset, and its mean vector is calculated. Covariance Matrix For parameter vectors acquired in real time Calculate its Mahalanobis distance : Mahalanobis distance It measures the degree to which the current state deviates from the center of a normal multivariate distribution. A health score mapped to 0-100 using a sigmoid function. : ,here It is a baseline distance (e.g., normal data) (95th percentile) It is the scaling factor. When hour, ;when Much larger hour, Approaching 0.

[0041] (4) Specific implementation of fault mode matching: when health score When a downward trend or abnormality in specific parameters is observed, the system extracts key time-series data (such as vibration spectrum) from the current device over a recent period (e.g., the past 10 minutes). Using the Dynamic Time Warping (DTW) algorithm, the DTW distance between this sequence and the typical fault precursor sequences stored in each "fault feature" node of the knowledge base is calculated. A smaller DTW distance indicates higher similarity. A similarity threshold is then set. When the similarity of the matched fault feature sequences exceeds this threshold, it is considered that the current device may have experienced a corresponding fault mode. The system will automatically associate possible root causes and recommended maintenance actions along the relationships in the knowledge graph.

[0042] (5) Specific implementation of maintaining early warning generation: second threshold Set the score to 70. When... When a high-risk fault mode (such as "bearing breakage") is matched, the system generates an alert. The recommended execution time is automatically determined by the system by finding the nearest available window in the future production schedule. The estimated time is obtained from the historical average time of similar maintenance actions in the knowledge base. The alert information is pushed to the CMMS (Computerized Maintenance Management System) and a work order is generated.

[0043] Understandably, this step combines statistical anomaly detection with knowledge graph-based case reasoning, achieving a leap from "knowing there's a problem" to "knowing what the problem is, why it's happening, and how to fix it." Simple Mahalanobis distance calculation (or other anomaly detection algorithms) can only provide a quantitative indicator of the degree of deviation (health score), but cannot offer any actionable diagnostic information. This method, through knowledge graph matching, can directly link abnormal numerical patterns with accumulated fault knowledge from human experts, outputting highly interpretable diagnostic reports, significantly reducing the technical threshold and diagnostic time for operations and maintenance personnel. The structured storage of the knowledge graph makes fault diagnosis logic transparent, traceable, and scalable. Newly discovered fault cases can be easily added to the graph, thereby continuously enhancing the diagnostic capabilities of the entire system as runtime increases, realizing the digital accumulation and intelligent reuse of operational experience.

[0044] Specifically, the method also includes a knowledge base evolution step: after each maintenance operation is completed, the actual operation data, replacement component information and post-maintenance equipment performance verification data are collected during the maintenance process. This maintenance event is treated as a new case, and after standardization, it is fed back into the equipment health knowledge base for optimization of subsequent fault mode identification and root cause reasoning.

[0045] It should be further explained that the knowledge base evolution process is a complete "data-knowledge" closed loop, and its specific implementation process is as follows: 1) Maintenance Data Collection: After a maintenance engineer receives an alert work order via a maintenance app on a mobile terminal (such as a tablet) and completes on-site maintenance, the app guides the engineer to fill out a structured electronic work order. This work order not only includes pre-filled alert information (such as device ID, alert time, and preliminary diagnosis), but also requires the engineer to fill in or confirm the following key fields: a) Actual observed fault phenomena (can be selected from a list or entered manually); b) The confirmed root cause (you can choose from the system's recommended list or enter a new cause); c) A list of the specific maintenance operation steps actually performed (e.g., “removing the protective cover”, “cleaning the encoder”, “replacing the motor bearing, model: XXX”). d) Spare parts / consumables parts number and quantity used; e) After maintenance, key parameters of the equipment during test mode (e.g., after bearing replacement, the vibration value decreased from 8.5 mm / s to 1.8 mm / s). Simultaneously, the App automatically records the maintenance start time, end time, and engineer ID.

[0046] 2) Standardized Case Processing: After receiving a completed electronic work order, the system's backend service initiates a standardized process. First, the "actual fault phenomenon" and "root cause" described in the text are extracted and normalized using Natural Language Processing (NLP) technology, mapped to existing standard terminology nodes in the knowledge base, or new standard terminology nodes are created. Second, the text sequence of "maintenance operation steps" is matched against a standard maintenance operation code library, transforming it into a series of standardized operation codes (e.g., "OP001-Disassembly", "OP005-Cleaning", "OP012-Replacement"). Finally, performance verification indicators are calculated, such as the percentage reduction in vibration. .

[0047] 3) Feedback and Model Optimization: Standardized case data is stored as a new "case" node in the graph database. This new case node is linked to the corresponding "root cause" node through the "confirm_root cause" relationship and to a series of "maintenance action" nodes through the "take_maintenance action" relationship. Simultaneously, the equipment parameter sequences (especially fault feature sequences) collected before and after this maintenance are extracted and stored as a new instance sample for this "fault feature" node. This new sample can be used to optimize subsequent fault matching algorithms. For example, in subsequent DTW similarity calculations, the new sample can be added to the reference sequence set of this fault feature, or used to retrain a deep learning-based sequence classification model, thereby improving the accuracy of identifying this type of fault pattern. Furthermore, the effectiveness of the maintenance actions (verified through performance metrics) is also evaluated. The results (of these results) are also recorded, and highly effective maintenance actions will receive higher weight in subsequent recommendations.

[0048] Understandably, the knowledge base evolution process is key to the system's "self-learning" and "adaptive" capabilities. It ensures that every on-site maintenance practice becomes nourishment for system knowledge accumulation and algorithm optimization. Traditional expert systems or fault diagnosis models often have static knowledge or parameters, making it difficult to adapt to equipment aging, the introduction of new processes, or the emergence of new faults. This method, however, establishes an automated feedback loop from practice to knowledge, enabling the knowledge base to continuously absorb the latest maintenance experience, and the diagnostic model to continuously optimize its judgment boundaries with the latest data. This means that the longer the system runs and the more maintenance cases it experiences, the higher its diagnostic accuracy and recommendation precision will be. This effectively solves the problem of experience loss caused by equipment iteration and personnel changes in industrial scenarios, transforming the tacit knowledge of individuals or teams into digital assets that can be inherited and reused by the enterprise, achieving sustainable evolution of operational capabilities.

[0049] Specifically, the step of calculating the device's real-time health score based on the real-time operating parameters using a preset anomaly detection algorithm includes: configuring differentiated health assessment models for different types of key devices, wherein the assessment model defines a set of key parameters and their weights for health score calculation; for online learning-enabled assessment models, continuously fine-tuning the model parameters using historical normal state data of the current device and similar devices to adapt to individual device differences and slow performance drift; and generating the real-time health score by combining the output results of the anomaly detection algorithm and the weighted statistical analysis results based on the set of key parameters.

[0050] It should be further explained that this step involves a refined and personalized design for calculating the health score, as detailed below: (1) Differentiated Model Configuration: During system initialization, a "health assessment model template" is preset for each type of equipment. This template is a configuration file in JSON or XML format. For example, for a "precision dispensing valve," its configuration file might be: {"device_type":"Dispensing_Valve","key_parameters":[{"name":"pressure_std","weight":0.30,"normal_range":[0.8,1.2 ]},{"name":"flowrate_avg","weight":0.25,"normal_range":[9.5,10.5]},{"name":"response_time","weight":0.20,"normal_ range":[5,15]},{"name":"valve_temperature","weight":0.15,"normal_range":[20,45]},{"name":"actuation_count","weig ht":0.10,"normal_range":[0,1e7]}],"anomaly_detection_algorithm":"IsolationForest","online_learning_enabled":true} Here, `key_parameters` defines five key parameters used for weighted statistical analysis and their weights, with the sum of the weights being 1. `normal_range` is the normal range defined by factory calibration or historical statistics. `anomaly_detection_algorithm` specifies the core anomaly detection algorithm used; here, Isolation Forest is used as an example.

[0051] (2) Online Learning and Adaptation: For models with online learning enabled (such as Isolation Forest), the system periodically (e.g., every morning) initiates a model update task. This task collects data from all devices marked as "normal operation" over a past period (e.g., the past 7 days) (typically data from periods when no alarms were triggered and production yield met standards), and uses this new data to incrementally train or retrain the original Isolation Forest model. For the weighted statistical analysis section, the "normal range" of key parameters... It will also be dynamically updated based on recent normal data, for example, updating it to the 5th percentile of the parameter value in recent data ( ) and the 95th percentile ( This dynamic adjustment allows the definition of "normal" to evolve as the device ages naturally (such as a slow increase in response time), thus avoiding false alarms.

[0052] (3) Comprehensive score generation: final real-time health score It is a weighted composite of two parts. The first part is the anomalous score derived from the core anomaly detection algorithm (such as Isolation Forest). (The algorithm output is usually an anomaly probability, which needs to be converted into a health score of 0-100, for example) The second part is the performance score based on weighted statistics of key parameters. Performance score The calculation is as follows: for each key parameter Calculate its current value Relative to the dynamic normal range deviation :like Within the scope, ;like Out of range ,in Is it away The most recent boundary value. Then the individual score for this parameter is... Overall performance score ,in This is the weight of that parameter. Final health score. .here It is a harmonic weight, with a preferred value of 0.7. The reason is that the core anomaly detection algorithm ( The weighted statistical score is more sensitive to sudden and complex failures and should account for the majority of the score; while the weighted statistical score ( It can reflect the chronic deviation trend of each key parameter and serves as an important supplement.

[0053] Understandably, this step, through differentiated configuration and online learning mechanisms, achieves "personalization" and "adaptability" in health assessment, greatly improving the accuracy and practicality of the assessment. Differentiated configuration acknowledges that different devices, due to their different operating principles and key components, should have different evaluation systems and weights for their health status. For example, the pressure stability of a dispensing valve has a high weight, while the torque curve characteristics of an electric screwdriver have a high weight, making the scoring more professional. The online learning mechanism enables the model to follow the unique operating characteristics and slow performance drift of each device, dynamically adjusting its "normal" baseline. This effectively reduces "false alarms" caused by normal aging of equipment or seasonal environmental changes, ensuring that the health score always reflects the "relative health" of the equipment relative to its current life stage, rather than a rigid absolute standard. The comprehensive scoring method combines the sensitivity of algorithms with the robustness of statistics. It can quickly capture complex failure modes composed of minor anomalies in multiple parameters (through the algorithmic part) and intuitively reflect the clear degradation of individual key parameters (through the statistical part), providing maintenance personnel with a comprehensive and detailed equipment health profile, supporting the decision-making upgrade from "condition monitoring" to "precise maintenance."

[0054] Specifically, triggering the linkage operation and maintenance command to adjust the operating status or task of the associated equipment includes: generating one or more preliminary linkage control plans in the digital twin model based on the diagnostic results and preset linkage strategy rules, wherein the plans include operation instructions for the target equipment and path planning instructions for the collaborative robot; simulating the execution of the preliminary linkage control plans in the digital twin model, and evaluating the impact of each plan on predefined production efficiency indicators based on the simulation results; selecting the plan with the best comprehensive evaluation result from the one or more plans as the final linkage operation and maintenance command; and issuing the final linkage operation and maintenance command to the corresponding physical entity equipment for execution.

[0055] It should be further explained that the joint operation and maintenance decision-making and execution is a complex, simulation-based optimization process, which is implemented as follows: (1) Contingency Plan Generation: The system maintains an editable linkage strategy rule base. Rules adopt the "condition-action" form. For example, a rule can be: "IF (device health score < 60) AND (device type == 'tightening shaft') AND (device current task queue length > 0) THEN Generate Contingency Plan: 1. Pause the current task of this device; 2. All unstarted tasks in the equipment queue are checked for process compatibility and assigned to other spare tightening shafts in the same workstation; 3. Send a command to the AGV system to move the material rack of the equipment out of the workstation; 4. Mark the device area in the digital twin as 'Under Maintenance, Entry Prohibited'. When the diagnostic module issues an alert message (containing device ID, health score, and fault code), the rule engine matches all matching rules. Each rule may generate one or more alternative plans. These plans are structured lists of instruction objects.

[0056] (2) Digital Twin Simulation and Evaluation: After generating the contingency plan, the system will not execute it immediately, but will instead launch a high-speed, discrete-event digital twin simulator. The simulator uses the current digital twin model (reflecting the real-time state of the physical world) as its initial state, injects the instruction sequence from the contingency plan, and then rapidly extrapolates the operation of the workbench over a future period (such as the next production cycle or 30 minutes) at simulation time steps (e.g., 1 second). During the simulation, a series of predefined production efficiency indicators will be calculated in real time, such as: a) Changes in task completion time The difference between the total completion time of all affected tasks and the original planned completion time after the contingency plan is implemented.

[0057] b) Work-in-process inventory level The number of work-in-process items in each buffer zone of the workbench at the end of the simulation.

[0058] c) Total length of robot path The total length of the new path executed by the collaborative robot.

[0059] d) Energy consumption forecast Total energy consumption of all devices during the simulation.

[0060] e) Security Risk Scoring The risk value is calculated based on the degree of overlap between the robot path and the personnel activity area and maintenance area. Each contingency plan i will receive an evaluation vector. .

[0061] (3) Optimal decision: The system pre-sets a multi-objective optimization function. This is used to calculate the overall score for each contingency plan. A common weighted summation form is: Weight arrive Configure based on production priorities, such as in scenarios where efficiency is paramount. The highest weight is assigned; in scenarios where safety is paramount, The system with the highest weight. The highest-level contingency plan serves as the final directive. Before its final issuance, the plan and assessment results can be displayed on the control interface for operator confirmation.

[0062] (4) Command Issuance: The final command is converted into a protocol command that can be executed by the specific device. For example, a "control command" frame conforming to the PROFINET protocol is sent to the PLC to disconnect the main contactor of the faulty device; a new path point sequence is sent to the collaborative robot controller through the SDK provided by the robot manufacturer (such as UR Cap, KUKA, Ethernet RSI); and a handling task is issued by calling the AGV scheduling system through the standard REST API. All commands are executed synchronously or in an orderly manner through timestamps or sequence numbers.

[0063] Understandably, this step, through "pre-simulation" in a digital twin environment, elevates the collaborative decision-making process from simple, potentially problematic reflexes to intelligent decisions that have undergone global optimization and safety verification. It avoids the common dilemma in traditional automation systems where "solving one problem leads to another." For example, simply switching the task of a faulty device to another might overload that device, creating a new bottleneck; or the new robot path might intersect with personnel access routes, posing safety hazards. By rapidly simulating in a virtual twin, the system can proactively assess these potential risks and select the overall optimal solution. This significantly enhances the resilience, safety, and efficiency of the entire workbench system in the face of sudden failures, enabling the production system to automatically, quickly, and safely recover to a highly efficient operating state after a disturbance. This is a core technology for achieving a truly adaptive and self-optimizing production system.

[0064] Specifically, the linkage strategy rules include at least an equipment failure early warning strategy and a task switching strategy; the task switching strategy stipulates that when it is predicted that a certain device is about to fail, its unfinished task queue will be switched to a backup device, wherein the switching decision needs to take into account the current load, task processing capacity, and compatibility of the required materials of the backup device.

[0065] It should be further explained that the specific implementation of the linkage strategy rules is the cornerstone of the system's intelligent decision-making, among which the task switching strategy is particularly critical, and its specific decision-making logic is implemented as follows: (a) Equipment Failure Early Warning Strategy Classification: The early warning strategy is based on the health score. The severity of the failure mode is classified. For example: Level 1 (Observational): Or it may be matched to a low-risk mode (such as "minor filter clogging"). Action: The device will only be highlighted in the digital twin interface and dashboard, and a log will be recorded; no physical action will be triggered. Level 2 (Warning Level): Or it may match a medium-risk mode (such as "mid-stage bearing wear"). Action: Issue an audible and visual warning on the interface to notify maintenance personnel to prepare. The system begins pre-calculating possible switchover plans in the background, but does not issue them. Level 3 (Action Level): Or it may match a high-risk mode (such as "frequent triggering of motor overcurrent protection"). Action: Immediately initiate the task switching decision process and execute the optimal contingency plan verified by simulation.

[0066] (b) Multi-factor evaluation of task switching decision: When a Level 3 warning is triggered and the faulty equipment has unfinished tasks, the system executes the following decision-making process: 1) Identify candidate backup equipment: Based on the process route and BOM, identify all backup equipment in the workbench or adjacent workstations that can perform the same or similar processes.

[0067] 2) Calculate the handover suitability score: for each candidate device Calculate the score The calculation formula is as follows: .in, It is a candidate device The matching degree of the ability to handle the task to be switched is a value of 0-1, which is determined by the degree of matching of process parameters (such as accuracy and torque range), with a perfect match being 1. It is a candidate device The estimated completion time (in minutes) for the current task queue. It is a penalty for differences in ability; if the abilities are perfectly matched ( If the value is 0, then it is 0; otherwise, it is a value proportional to the difference in ability. This is a material compatibility penalty; if the material required for the task to be switched to is already available on the candidate equipment. If the material is in the faulty equipment station, the penalty value is 0; if AGV or manual transfer of materials from the faulty equipment station is required, the penalty value is the estimated transfer time (minutes). These are weighting coefficients used to balance the importance of time, capacity, and material factors. A typical optimal value setting is... The reason is that ensuring production continuity (minimizing queue time) is usually the highest priority, therefore The highest priority is ensuring task quality (capability matching); the next highest priority is ensuring task quality (capability matching); the time cost of material transportation is relatively controllable and has the lowest weight.

[0068] 3) Decision-making and execution: selection The candidate device with the highest value is selected as the switching target. Subsequently, the system simulates and verifies the feasibility of this switching scheme (including task transfer, material handling, path replanning, etc.) in a digital twin. If there are no conflicts, a specific switching instruction set is generated and issued.

[0069] Understandably, this step explicitly defines the intelligent decision-making logic for task switching, transforming it from a simple "finding free time" problem into a global optimization problem that comprehensively considers "whether it can be done," "how quickly," and "what the cost is." This is achieved by introducing capability matching. This ensured that the switched-out equipment was technically competent, avoiding quality issues caused by insufficient equipment capabilities. This was achieved by considering the current load. This achieved load balancing, preventing failover from causing backup equipment overload to become a new bottleneck. Material compatibility penalties were assessed. By incorporating the practical constraints of material supply into the decision-making process, the solution becomes feasible. This refined switching strategy, based on multi-objective evaluation, maximizes the use of redundant resources within the system. In the event of sudden equipment failure, it can restore the production process with the fastest speed, the least quality risk, and the least additional logistical cost, significantly improving the availability and flexibility of the production system.

[0070] Specifically, after constructing the digital twin model, the process also includes a knowledge base construction step: extracting events, features, and solutions related to workbench equipment failures, abnormal material consumption, and production cycle optimization from historical operation and maintenance logs, equipment manuals, and expert experience, forming a structured operation and maintenance knowledge graph, and associating it with the corresponding entities in the digital twin model.

[0071] It should be further explained that the construction of the initial knowledge base is the knowledge foundation for the system's intelligence, and its specific implementation is a systematic project, with the following steps: 1) Extraction of multi-source heterogeneous knowledge: a) Historical Maintenance Logs: Export historical records from CMMS, SCADA, or digitized paper work orders databases. Process these logs using ETL (Extract, Transform, Load) tools and Natural Language Processing (NLP) techniques. For example, from the unstructured maintenance description "Replaced the servo motor of the third axis, encoder alarm cleared," use a Named Entity Recognition (NER) model to extract the entities "third axis," "servo motor," and "encoder," along with the action "replacement," and use a relation extraction model to establish the relationships "third axis - occurred - encoder alarm" and "encoder alarm - passed - replaced servo motor - resolved."

[0072] b) Equipment Manual: Parse PDF equipment manuals, using OCR and text parsing technology to locate the "Troubleshooting" section. Extract structured triples from the list-style or paragraph-style content of "Fault Phenomenon," "Possible Causes," and "Solutions" using rule templates or machine learning models (such as BERT-based sequence labeling). For example, extract "Motor Overheating - Possible Cause - Overload," "Motor Overheating - Possible Cause - Poor Heat Dissipation," and "Motor Overheating - Solution - Check Load" from "Phenomenon: Motor overheating. Possible Causes: Overload; Poor Heat Dissipation. Solution: Check Load; Clean Fan."

[0073] c) Expert Experience: By designing structured questionnaires or interviews, guide domain experts (such as senior equipment engineers and process engineers) to contribute their experience in the format of "If... (phenomenon), it may be... (cause), should... (action)". These experiences can be directly entered into the system or converted into structured data through speech recognition and text analysis.

[0074] 2) Knowledge Graph Construction and Storage: The extracted structured triples are imported into a graph database (such as Neo4j or Nebula Graph). In the graph database, entities (such as "servo motor", "overload alarm", and "torque sensor") are used as nodes, and relationships (such as "cause", "manifests as", and "requires inspection") are used as edges connecting the nodes. Attributes are defined for nodes and edges; for example, the "fault mode" node can have attributes such as "severity level" and "occurrence frequency"; the "solution" node can have attributes such as "mean time to repair" and "required skill level". This constructs a queryable and reasoning-based operations and maintenance knowledge graph.

[0075] 3) Association with the Digital Twin Model: During digital twin modeling, each virtual device object (such as "Robot_Arm 1" or "Dispensing Machine_Main Valve") is assigned a unique identifier (ID). In the knowledge graph, a corresponding "Device Instance" node is created and linked to the "Device Model" node via the "is_instance_of" relationship. Simultaneously, the twin ID of this device instance is used as its attribute. Thus, when "Robot_Arm 1" in the digital twin model malfunctions, the system can find the corresponding "Device Instance" node in the knowledge graph using its ID, and then query all relevant fault modes, causes, and solutions along relationships such as "may fail".

[0076] Understandably, the operations and maintenance (O&M) knowledge graph constructed in this step is a crucial process for transforming the implicit, fragmented experience accumulated by humans over a long period into explicit, structured knowledge that machines can understand, relate, and reason about. It solves the problems of scattered, difficult-to-find, and reusable knowledge in traditional O&M. Through strong associations with specific device instances in the digital twin model, real-time data can directly trigger relevant knowledge, achieving instantaneous connection from "data" to "information" to "usable knowledge." This provides rich background knowledge and logical chains for subsequent intelligent diagnosis, root cause analysis, and maintenance decisions, enabling the system not only to discover problems but also to understand the possible causes behind them and recommend validated solutions. This is equivalent to equipping the O&M system with a "digital brain," giving it associative and reasoning abilities similar to human experts, and is a crucial cornerstone for realizing the transition of O&M assisted decision-making from "perceptual intelligence" to "cognitive intelligence."

[0077] Specifically, the method further includes: real-time monitoring of the workbench's operating status and matching the current operating parameters with the event features in the operation and maintenance knowledge graph; when the matching degree exceeds the third threshold, actively pushing the associated solutions or historical processing records in the operation and maintenance knowledge graph to the operation and maintenance interface to assist in decision-making.

[0078] It should be further explained that this step is an active application of knowledge graphs, realizing "knowledge actively finding people," and its specific implementation process is as follows: 1) Real-time Monitoring and Feature Matching: A "knowledge matching engine" runs in the system background. This engine continuously monitors the data and event streams in the digital twin model. For continuous numerical parameters (such as vibration values ​​and temperature), the engine calculates their short-term (e.g., the past minute) features (such as mean, variance, and spectral features) and performs similarity calculations with the numerical feature templates stored in the "fault feature" nodes of the knowledge graph. The calculation can use cosine similarity, Euclidean distance (which needs to be normalized), or more complex sequence matching algorithms (such as DTW). For discrete events (such as alarm codes from the PLC or defect codes entered by the operator on the HMI), the engine directly performs precise or fuzzy matching with the names or codes of the "fault phenomenon" nodes in the knowledge graph. The matching process is parallel, meaning that the real-time status at the same moment may match to varying degrees with multiple "feature" or "phenomenon" nodes in the knowledge graph.

[0079] 2) Matching score calculation and threshold determination: For each matching pair, the engine calculates a matching score. The range is between 0 and 1. For numerical features, the matching degree can be (1 - normalized distance); for event matching, a perfect match is 1, and a partial match (such as containing the same keywords) can be 0.7, etc. The system sets a global third threshold. This threshold is used to determine whether to trigger an active push notification. The preferred value for this threshold is 0.75. The rationale is as follows: a threshold that is too low (e.g., 0.6) will result in a large amount of unrelated information being pushed, causing "information overload" and interfering with operations and maintenance personnel; a threshold that is too high (e.g., 0.9) may miss some important but not perfectly matching early signs. 0.75 is an empirical value that strikes a good balance between recall and precision in practice. When any matching... Upon triggering, the push notification process is initiated. 3) Knowledge Push: Once triggered, the system immediately performs the following operations: a) Aggregating information: The engine will traverse along the relationship edges of the knowledge graph to find nodes such as "failure mode", "root cause", "solution" and "historical case" that are connected to nodes with high matching degree features / phenomena.

[0080] b) Generate push card: Organize the aggregated information into an easy-to-read card, including: description of the matched phenomenon, possible failure modes (sorted by matching degree or historical frequency), recommended root cause troubleshooting steps, relevant standard operating procedure (SOP) link, and a list of similar cases handled in the past (links can be used to view detailed maintenance reports).

[0081] c) Multi-channel push: This knowledge card will be pushed to relevant responsible persons in real time through multiple channels: on the digital twin visualization screen in the workshop, the relevant equipment model will be highlighted and flashing and a card summary will pop up; on the engineer's mobile inspection app or computer workstation, the complete card will be pushed in the form of a pop-up window or message list; at the same time, this information will also be linked to the inspection or maintenance work order automatically generated in CMMS.

[0082] Understandably, this step transforms a static knowledge base into a proactive, context-aware decision support system, significantly improving the speed and accuracy of operational response. In the traditional model, when faced with an anomaly, engineers rely on memory or manually consulting numerous documents to find clues, a time-consuming and labor-intensive process prone to missing crucial information. This method, through real-time matching and proactive push, delivers the most relevant and potentially resolvable knowledge directly to engineers the moment an anomaly occurs. This is equivalent to equipping each on-site engineer with a tireless, knowledgeable "expert assistant," significantly shortening the Mean Time to Diagnose (MTTD) and Mean Time to Repair (MTTR). Especially for inexperienced new employees, this proactive push greatly enhances their problem-solving capabilities, ensuring the consistency and stability of the overall operational team's performance. Simultaneously, the pushed historical cases provide valuable practical experience references, enabling problem-solving to build upon past successes, increasing the first-time success rate of maintenance operations, and reducing the risk of secondary damage due to misjudgment or improper operation.

[0083] This invention provides another embodiment, which provides an intelligent workbench material management and maintenance system, the intelligent workbench material management and maintenance system comprising: (1) Digital twin construction and management unit, used to construct and maintain a digital twin model corresponding to the physical entity of the smart workbench.

[0084] It should be further explained that this system is a hardware and software integration platform for implementing the aforementioned method embodiments. Its specific implementation architecture is as follows: The digital twin construction and management unit serves as the core and foundation of the system and is deployed on a high-performance graphics server or cloud platform. Its software can be developed using the Unity3D / UE4 engine in combination with an industrial IoT platform (such as PTCThing Worx or Siemens MindSphere). It is responsible for the 3D modeling of the physical workbench, the synchronization of virtual and real data (through OPCUAServer / Client and MQTT Broker / Subscriber), model rendering and driving, and provides a unified API data access interface and model service for upper-layer applications.

[0085] (2) Predictive material management unit, used to analyze material consumption trends and predict demand based on the digital twin model, and generate and send material scheduling instructions.

[0086] It should be further explained that the predictive material management unit is deployed as an independent microservice, subscribing to material status topics and MES work order events published by the digital twin unit. Internally, it encapsulates a trained demand prediction model (such as the Attention-LSTM model), runs the prediction algorithm periodically (e.g., every 5 minutes), and sends replenishment instructions through the Enterprise Service Bus (ESB) or by directly calling the WMS RESTful API.

[0087] (3) Equipment health diagnosis and maintenance unit, which is used to monitor the working parameters of key equipment based on the digital twin model, use knowledge base and algorithm to assess health status and predict faults, and generate maintenance warnings.

[0088] It should be further explained that the equipment health diagnosis and maintenance unit is also deployed in the form of microservices. It subscribes to the sensor data streams of various devices published by the digital twin unit, and has built-in multiple real-time streaming data processing and machine learning analysis engines (such as Apache Flink integrated with Scikit-learn or TensorFlow Serving). It calculates the equipment health score in real time and performs fault mode matching. It accesses an independent graph database (storing the health knowledge base) and a time series database (storing historical case data). When an alert is generated, it publishes the alert event through message queues (such as RabbitMQ, Kafka).

[0089] (4) The linkage operation and maintenance decision and execution unit is used to generate and verify the linkage control plan in the digital twin environment based on the output of the equipment health diagnosis and maintenance unit, and send the optimized instructions to the physical workbench for execution; The predictive material management unit, equipment health diagnosis and maintenance unit, and coordinated operation and maintenance decision and execution unit all interact with the digital twin construction and management unit through data exchange and command collaboration.

[0090] It should be further explained that the coordinated operation and maintenance decision-making and execution unit, acting as the system's "intelligent brain," listens to early warning messages issued by the equipment health diagnosis unit. Internally, it includes a rule engine (such as Drools), a lightweight discrete event simulation engine, and an optimization decision module. Upon receiving an early warning, it performs rapid simulation and deduction on a real-time model copy provided by the digital twin unit. After determining the optimal contingency plan, it sends a sequence of control commands to the PLCs, robot controllers, and other equipment in the field for execution via an industrial protocol gateway (such as a soft PLC or gateway device supporting PROFINET and EtherCAT / IP). These four units are encapsulated and orchestrated using container-based technologies (such as Docker) and utilize a service mesh (such as Istio) for service discovery, load balancing, and security management. They collaborate to form a loosely coupled, highly cohesive, and scalable intelligent operation and maintenance system.

[0091] Understandably, this system architecture, through a clear modular design and a data collaboration mechanism centered on digital twins, achieves decoupling and efficient integration of complex functions. The digital twin construction and management unit, acting as a unified "digital mirror" and data bus, ensures that the prediction, diagnosis, and decision-making units operate on a consistent, real-time, and high-fidelity world model, avoiding decision conflicts caused by data inconsistencies. Each functional unit is deployed and expanded independently as a microservice, giving the system high flexibility and maintainability. For example, the prediction algorithm can be upgraded independently without affecting the diagnostic service, and computing resources can be dynamically scaled as needed to handle high data analysis loads. The coordinated operation and maintenance decision-making and execution unit, by introducing a simulation layer, conducts "virtual trial runs" before issuing instructions, greatly improving the safety and reliability of automated decision-making. This allows the entire system to handle various abnormal operating conditions safely and efficiently with minimal or no human intervention. This system is not only the implementation carrier of the method but also a key infrastructure for building future adaptive, self-optimizing, and self-healing intelligent production systems, bringing enterprises multiple significant benefits such as increased equipment utilization, reduced maintenance costs, reduced production interruptions, and improved safety levels.

[0092] In a preferred embodiment, this application also provides an electronic device, the electronic device comprising: The computer device includes a memory and a processor, wherein the memory stores computer-readable instructions that, when executed by the processor, implement the intelligent workbench material control and maintenance method. The computer device can be broadly categorized as a server, terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, memory, network interface, communication interface, etc., connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device may include non-volatile storage media and internal memory. The non-volatile storage media may store an operating system, computer programs, etc. The internal memory can provide an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface and communication interface of the computer device can be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, it performs the steps of the method of the present invention.

[0093] This invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the methods of embodiments of the invention to be performed. In one embodiment, the computer program is distributed across multiple network-coupled computer devices or processors, such that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, may be executed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations may be executed by one or more computer devices or processors, and one or more other method steps / operations may be executed by one or more other computer devices or processors. One or more computer devices or processors may execute a single method step / operation, or execute two or more method steps / operations.

[0094] Those skilled in the art will understand that the method steps of this invention can be performed by a computer program instructing related hardware, such as a computer device or processor, to perform the steps of this invention when executed. Depending on the context, any references herein to memory, storage, databases, or other media may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state drive, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0095] The technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification, provided that such combination does not contain contradictions.

[0096] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for material control and maintenance of an intelligent workbench, characterized in that, Includes the following steps: S1. Construct a digital twin model that includes the physical entity of the intelligent workbench and its operating environment mapping; S2. Based on the digital twin model, predictive management of workbench material consumption is performed, and material scheduling instructions are generated based on the management results. S3. Based on the digital twin model, perform health status diagnosis and predictive maintenance on key equipment of the workbench; S4. When the diagnostic results meet the preset conditions, trigger the linkage operation and maintenance command to adjust the operating status or tasks of the associated equipment.

2. The intelligent workbench material management and maintenance method according to claim 1, characterized in that, The predictive management of workbench material consumption and the generation of material scheduling instructions based on the management results include: The system acquires real-time inventory data, consumption rate, and production plan data from the manufacturing execution system for materials on the workbench; it also acquires material-related quality perception data and supply chain reliability data, where the quality perception data reflects key characteristic indicators of material batches and the supply chain reliability data reflects the historical fulfillment stability of material supply sources; it inputs the inventory data, consumption rate, production plan data, quality perception data, and supply chain reliability data into a demand forecasting model to predict the material demand gap within a specified future time window; when the predicted demand gap exceeds a first threshold, it automatically generates and sends a replenishment request instruction containing material identification, required quantity, and required time to the warehouse management system or supplier system.

3. The intelligent workbench material management and maintenance method according to claim 1, characterized in that, The aforementioned health status diagnosis and predictive maintenance of key equipment on the workbench includes: A corresponding digital twin sub-model is constructed for the key equipment, and its operating parameters are synchronized to the sub-model in real time. An equipment health knowledge base associated with equipment type and process is established, which stores feature data, maintenance records, and performance degradation patterns of historical failure cases. Based on the real-time operating parameters, the instant health score of the equipment is calculated using a preset anomaly detection algorithm. The real-time operating parameters and historical parameter sequences are matched and similarity calculated with the data in the equipment health knowledge base to perform fault mode identification and root cause reasoning, and to generate diagnostic reports and maintenance recommendations. When the instant health score is lower than a second threshold or a specific fault mode is identified, a maintenance warning is generated that includes maintenance content, suggested execution time, and estimated time consumption.

4. The intelligent workbench material management and maintenance method according to claim 3, characterized in that, The method also includes a knowledge base evolution step: After each maintenance operation is completed, the actual operation data, replacement component information, and post-maintenance equipment performance verification data are collected. This maintenance event is treated as a new case, processed in a standardized manner, and then stored in the equipment health knowledge base to optimize subsequent fault mode identification and root cause reasoning.

5. The intelligent workbench material management and maintenance method according to claim 3, characterized in that, The step of calculating the device's instantaneous health score based on the real-time operating parameters using a preset anomaly detection algorithm includes: Differentiated health assessment models are configured for different types of key equipment. The assessment model defines a set of key parameters and their weights for calculating the health score. For online learning-enabled assessment models, the model parameters are continuously fine-tuned using historical normal state data of the current equipment and similar equipment to adapt to individual differences in equipment and slow performance drift. The real-time health score is generated by combining the output of the anomaly detection algorithm and the weighted statistical analysis results based on the set of key parameters.

6. The intelligent workbench material management and maintenance method according to claim 1, characterized in that, The triggering of the linkage operation and maintenance command to adjust the operating status or tasks of the associated devices includes: In the digital twin model, based on the diagnostic results and preset linkage strategy rules, one or more preliminary linkage control plans are generated. The plans include operation instructions for the target equipment and path planning instructions for the collaborative robot. The preliminary linkage control plans are simulated and executed in the digital twin model, and the impact of each plan on predefined production efficiency indicators is evaluated based on the simulation results. From the one or more plans, the plan with the best comprehensive evaluation result is selected as the final linkage operation and maintenance instruction. The final linkage operation and maintenance instruction is issued to the corresponding physical entity equipment for execution.

7. The intelligent workbench material management and maintenance method according to claim 6, characterized in that, The linkage strategy rules include at least an equipment failure early warning strategy and a task switching strategy. The task switching strategy stipulates that when it is predicted that a certain device is about to fail, its unfinished task queue will be switched to a backup device. The switching decision needs to take into account the current load, task processing capacity, and compatibility of the required materials of the backup device.

8. The intelligent workbench material management and maintenance method according to claim 1, characterized in that, Following the construction of the digital twin model, the process also includes a knowledge base construction step: By extracting events, features, and solutions related to workbench equipment failures, abnormal material consumption, and production cycle optimization from historical operation and maintenance logs, equipment manuals, and expert experience, a structured operation and maintenance knowledge graph is formed and associated with the corresponding entities in the digital twin model.

9. The intelligent workbench material management and maintenance method according to claim 8, characterized in that, The method further includes: real-time monitoring of the workbench's operating status and matching the current operating parameters with the event features in the operation and maintenance knowledge graph; when the matching degree exceeds the third threshold, actively pushing the associated solutions or historical processing records in the operation and maintenance knowledge graph to the operation and maintenance interface to assist in decision-making.

10. A smart workbench material management and maintenance system, characterized in that, include: The digital twin construction and management unit is used to build and maintain a digital twin model corresponding to the physical entity of the smart workbench; A predictive material management unit is used to analyze material consumption trends and predict demand based on the digital twin model, and to generate and send material scheduling instructions. The equipment health diagnosis and maintenance unit is used to monitor the operating parameters of key equipment based on the digital twin model, use a knowledge base and algorithms to assess health status and predict faults, and generate maintenance warnings. The linkage operation and maintenance decision and execution unit is used to generate and verify the linkage control plan in the digital twin environment based on the output of the equipment health diagnosis and maintenance unit, and send the optimized instructions to the physical workbench for execution; The predictive material management unit, equipment health diagnosis and maintenance unit, and coordinated operation and maintenance decision and execution unit all interact with the digital twin construction and management unit through data exchange and command collaboration.

Images