Enterprise production management method based on digital twinning and workflow simulation

Abstract

The application discloses an enterprise production management method based on digital twinning and workflow simulation, and particularly relates to the technical field of industrial internet and intelligent manufacturing, and comprises the following steps: constructing a physical production system digital twinning and a business process workflow model, dynamically correlating to form an integrated digital twinning model through a model fusion engine, driving by real-time events, deducing simulation on future production processes, and outputting bottleneck and conflict prediction, generating multiple alternative scheduling schemes by using a multi-objective optimization engine based on the prediction results, selecting an optimal scheme and analyzing the optimal scheme into control instructions to issue and execute, and realizing model self-correction and closed-loop optimization through real-time monitoring and feedback. The application solves the problem of disconnection between business and physical state caused by independent digital twinning and workflow, realizes full-process closed-loop management from prediction to execution, and improves the production self-adaptation and intelligent level.

Description

Technical Field

[0001] This invention relates to the field of industrial internet and intelligent manufacturing technology, and more specifically, to an enterprise production management method based on digital twins and workflow simulation. Background Technology

[0002] In existing technologies, digital twins, as a key technology for realizing cyber-physical fusion, have been widely used in the modeling and monitoring of industrial production systems. By integrating geometric, physical, behavioral, and rule-based models and utilizing sensor data, it constructs a high-fidelity virtual mapping of physical entities, thereby achieving real-time visualization and historical traceability of the physical production system's operational status. On the other hand, workflow technology has been maturely applied at the enterprise management level. It standardizes and automates the execution of business processes through predefined task nodes and logical rules (such as sequence, branching, and looping), for example, managing the entire production management process from order placement to product delivery in ERP (Enterprise Resource Planning) and MES (Manufacturing Execution System).

[0003] However, in practical use, it still has some shortcomings. For example, digital twins and workflow models are often independent of each other. Digital twins focus on the "data mapping" of the physical entity's state, while workflows focus on the "logical control" of business processes. The two fail to achieve deep integration and dynamic semantic association, resulting in the inability of business processes to adaptively adjust and precisely drive according to the real-time state of the physical system. The system lacks a simulation and deduction core that integrates data and logic, making it difficult to make forward-looking predictions of production bottlenecks and resource conflicts in future periods based on real-time events. There is a disconnect between the path from optimization decision-making to production execution. The optimization scheme lacks a verification link in a high-fidelity fusion model, and deviations during execution cannot be effectively fed back to the model for self-correction, making it difficult to form a continuously iterative optimization intelligent closed loop, which restricts the further improvement of the adaptive and intelligent level of production management. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, the present invention provides an enterprise production management method based on digital twins and workflow simulation, which solves the problems mentioned in the background art through the following solutions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an enterprise production management method based on digital twins and workflow simulation, comprising:

[0006] S1: Based on the physical elements and real-time operational data of the physical production system, construct its digital twin; based on the production management business process, construct a workflow model containing task nodes and logical rules; through the model fusion engine, dynamically associate and semantically bind the logical nodes in the workflow model with the corresponding entities in the digital twin to form an integrated digital twin model that integrates data and logic.

[0007] S2: During the operation of the integrated digital twin model, the corresponding task nodes in the workflow model are dynamically activated by the real-time status events reported by the physical production system, and the production process in the future preset time period is simulated and the prediction results of production bottlenecks and resource conflicts are output.

[0008] S3: Based on the prediction results, multiple alternative production scheduling schemes are output through the optimization engine; wherein, the optimization engine is configured to collaboratively optimize multiple objectives such as equipment utilization, order delivery time and production cost;

[0009] S4: Input the multiple alternative production scheduling schemes into the integrated digital twin model and perform simulations in parallel. By comparing the key performance indicators output by the simulation, quantitatively evaluate and select the optimal production scheduling scheme.

[0010] S5: The selected optimal production scheduling scheme is parsed into control commands that can be executed by the underlying control system and sent to the execution equipment in the physical production system through the industrial network; at the same time, the execution effect of the control commands in the physical production system is monitored in real time through the integrated digital twin model, and the monitoring data is fed back to the integrated digital twin model for model self-correction and subsequent optimization process input.

[0011] Preferably, the construction of the digital twin includes:

[0012] Digitalization of physical elements: Using 3D modeling software or laser scanning technology to create geometric models of equipment, materials and tools on the production line to form virtual representations;

[0013] Real-time data access and mapping: Collect equipment status, process parameters, and material information through sensors, PLCs, and manufacturing execution systems, and bind the data to the corresponding virtual entities in the digital twin in real time;

[0014] Model Validation and Calibration: The accuracy of the model is verified by comparing historical data and simulation tests. A confidence assessment mechanism based on mean absolute percentage error is introduced, and an automatic calibration alarm is triggered when the MAPE continuously exceeds 5%.

[0015] Preferably, the construction of the workflow model includes:

[0016] Business process analysis: Identify key business processes in production management, including order processing, production planning, material scheduling, quality control, and equipment maintenance;

[0017] Logical rule definition: Use workflow modeling tools to define the logical rules between nodes, including sequential flow, conditional branching, event response, timing and deadline logic.

[0018] Preferably, the integrated digital twin model includes:

[0019] Establishment of the association mapping table: The model fusion engine maintains the "entity-node" mapping table to clarify the correspondence between logical nodes in the workflow model and entities in the digital twin;

[0020] Semantic binding implementation: Achieving semantic consistency between entities and nodes through unified attribute tags;

[0021] The operation of the fusion model enables dynamic interaction and synchronous updates between business process logic and physical entity status.

[0022] Preferably, the prediction result includes:

[0023] Based on real-time monitoring of task queue length and device load rate during the simulation process, potential bottlenecks and resource conflicts are identified.

[0024] Generate a structured report that includes the bottleneck location, estimated time of occurrence, scope of impact, and types of conflicting resources.

[0025] Preferably, the alternative production scheduling scheme includes:

[0026] A Pareto optimal solution set is generated through a multi-objective optimization engine, and each solution achieves the optimal trade-off between equipment utilization, order delivery time and production cost.

[0027] A multi-objective evolutionary algorithm is used to solve the problem, and multiple non-dominated solutions are output as alternatives.

[0028] Preferably, the key performance indicators include:

[0029] Equipment utilization rate, on-time order delivery rate, total production cost, work-in-process inventory level, and production efficiency;

[0030] The calculation is based on time-series data output from parallel simulation, and a comprehensive evaluation is performed through standardization and weighted scoring.

[0031] Preferably, the real-time monitoring includes:

[0032] The optimal production scheduling scheme is parsed into control commands and sent to the execution equipment through the industrial network;

[0033] By continuously receiving feedback data from the physical system through an integrated digital twin model, the virtual model is updated synchronously, enabling real-time monitoring of execution results.

[0034] Preferably, the model self-calibration and subsequent optimization include:

[0035] Deviation analysis: Compare the actual execution data with the simulation expected data to identify execution deviations;

[0036] Parameter calibration: Automatic or assisted calibration of model parameters in the digital twin based on deviation data, which is used as input for subsequent optimization processes.

[0037] The technical effects and advantages of this invention are as follows:

[0038] 1. It has achieved deep integration of data and logic, and built a "digital brain" for intelligent decision-making: Through the model fusion engine, the digital twin, which focuses on "data mapping" of physical entity status, is dynamically associated and semantically bound with the workflow model, which focuses on "logic control" of business processes, breaking down the barriers between the two independent of each other in the traditional system.

[0039] 2. It possesses forward-looking prediction and simulation capabilities, realizing a leap from passive response to proactive management: driven by real-time status events, it uses an integrated digital twin model to simulate and extrapolate the production process for a future preset period, enabling it to predict production bottlenecks and resource conflicts in advance and accurately.

[0040] 3. It provides scientific and objective optimization decision support, and improves the overall efficiency of production scheduling: It generates a set of alternative solutions that achieve Pareto optimality in multiple objectives such as equipment utilization, order delivery time and production cost through a multi-objective optimization engine, and further conducts quantitative evaluation and optimization through parallel simulation in a high-fidelity fusion model.

[0041] 4. A complete intelligent closed loop of "decision-execution-monitoring-correction" has been formed, realizing the continuous self-evolution of the system: the optimal solution is analyzed and sent to physical devices for execution, while the execution effect is monitored and fed back in real time through an integrated digital twin model;

[0042] 5. Improved transparency and precision in production management: Through the visualization of integrated digital twin models and the quantitative KPI comparison of parallel simulation outputs, the operating status, bottlenecks, and advantages and disadvantages of the entire production system are readily apparent. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0044] Figure 2 This is a schematic diagram of the integrated digital twin model construction structure of the present invention.

[0045] Figure 3 This is a schematic diagram of the real-time event-driven inference simulation and prediction structure of the present invention.

[0046] Figure 4 This is a schematic diagram of the multi-objective optimization generation of alternative solutions according to the present invention.

[0047] Figure 5 This is a schematic diagram of the parallel simulation evaluation and decision-making structure of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] refer to Figures 1-5 The enterprise production management methods shown are based on digital twins and workflow simulation, including:

[0050] S1: Based on the physical elements and real-time operational data of the physical production system, construct its digital twin; based on the production management business process, construct a workflow model containing task nodes and logical rules; through the model fusion engine, dynamically associate and semantically bind the logical nodes in the workflow model with the corresponding entities in the digital twin to form an integrated digital twin model that integrates data and logic.

[0051] S2: During the operation of the integrated digital twin model, the corresponding task nodes in the workflow model are dynamically activated by the real-time status events reported by the physical production system, and the production process in the future preset time period is simulated and the prediction results of production bottlenecks and resource conflicts are output.

[0052] S3: Based on the prediction results, multiple alternative production scheduling schemes are output through the optimization engine; wherein, the optimization engine is configured to collaboratively optimize multiple objectives such as equipment utilization, order delivery time and production cost;

[0053] S4: Input the multiple alternative production scheduling schemes into the integrated digital twin model and perform simulations in parallel. By comparing the key performance indicators output by the simulation, quantitatively evaluate and select the optimal production scheduling scheme.

[0054] S5: The selected optimal production scheduling scheme is parsed into control commands that can be executed by the underlying control system and sent to the execution equipment in the physical production system through the industrial network; at the same time, the execution effect of the control commands in the physical production system is monitored in real time through the integrated digital twin model, and the monitoring data is fed back to the integrated digital twin model for model self-correction and subsequent optimization process input.

[0055] S1: Constructing an integrated digital twin model: Through a model fusion engine, the digital twin of the physical production system is deeply integrated with the workflow model of the business process to form a unified, integrated digital twin model that combines data and logic.

[0056] S1.1: Constructing a digital twin of the physical production system:

[0057] Digitalization of physical elements: Using 3D modeling software (such as SolidWorks, Blender) or laser scanning technology, geometric models are created for physical elements such as equipment, materials, and tools on the production line to form their virtual representation.

[0058] Real-time data access and mapping: By deploying sensors, PLCs, and manufacturing execution systems on the equipment, real-time operational data such as equipment status (e.g., running, stopped, faulty), process parameters (e.g., speed, temperature), and material information are collected. Using an IoT platform or industrial data bus, this data is bound and synchronized in real-time with the corresponding virtual entities in the digital twin, ensuring that the virtual model is an accurate mirror of the physical system.

[0059] Model Validation and Calibration: The accuracy of the digital twin is verified through historical data comparison and simulation testing. A model confidence assessment mechanism is introduced, which is based on the calculation of the mean absolute percentage error (MAPE). When the MAPE continuously exceeds a preset threshold (5%), a model calibration alarm is automatically triggered.

[0060] Calculation formula: Model confidence assessment: ;

[0061] in, denoted as the sample size, where the actual values ​​are historical data reported by the physical system and the predicted values ​​are simulation output data from the digital twin under the same initial conditions. A lower MAPE value indicates a higher model confidence level.

[0062] S1.2: Constructing a workflow model for production management business processes:

[0063] A workflow model is an abstract representation of production management business processes, including task nodes (such as activities and decision points) and logical rules (such as sequence, branching, and looping). The implementation process includes:

[0064] Business process analysis: Identify key business processes in production management, such as order processing, production planning, material scheduling, quality control, and equipment maintenance.

[0065] Logical rule definition: Using workflow modeling standards or visualization tools, define the "logical rules" between nodes, including sequential flow (such as "material completeness check must be performed after work order is issued"), conditional branching (such as "if the quality inspection is qualified, it flows into the next process, otherwise the rework process is triggered"), and event response (such as "equipment failure event triggers maintenance application node").

[0066] S1.3: Achieve dynamic association and semantic binding through a model fusion engine:

[0067] The association mapping table is established: The model fusion engine maintains an "entity-node" mapping table. This table clearly defines which entity or type of entity in the digital twin corresponds to each logical node in the workflow model.

[0068] Semantic binding implementation: Semantic consistency is achieved by defining unified attribute tags (such as "equipment ID", "process type", "required materials") for entities and nodes.

[0069] Operation of the fusion model: At this point, an "integrated digital twin model" has been formed. In this model, the logic of the business process drives the state changes of the physical entity, while the real-time data of the physical entity, in turn, affects the progress path of the business process.

[0070] It is necessary to further explain that the core types and advanced forms of logical rules include:

[0071] Execution logic (sequential and parallel): defines the basic progression mode of task nodes.

[0072] Serial order: The specified tasks must be executed strictly in the order of "A->B->C".

[0073] Parallel branch: Defined after a specific node, multiple tasks can be started simultaneously.

[0074] Synchronous aggregation: requires that the next critical node be activated only after all tasks in multiple parallel branches have been completed.

[0075] Decision logic (conditional branching): This is the core of the rules, enabling the workflow to make intelligent judgments. Its conditional judgments directly rely on real-time or historical data provided by the digital twin.

[0076] State-based conditions: For example, IF the “Status” attribute of device A in the digital twin == “Fault” THEN activate the “Request Repair” node ELSE activate the “Assign Work Order” node.

[0077] Parameter-based conditions: For example, if the "real-time yield" attribute of the quality inspection unit in the digital twin is less than a preset threshold, then a "full inspection" or "process adjustment" process is triggered.

[0078] Resource-based conditions: For example, if the "current quantity" of the material inventory in the digital twin is less than the safety stock, then activate the "automatic replenishment" node.

[0079] Event response logic: This defines how the workflow responds to unexpected events from the physical world or within the system, and is key to achieving real-time performance.

[0080] External events, such as "equipment failure alarm" events, trigger "emergency equipment shutdown" and "production rescheduling" processes.

[0081] Internal events: such as the "simulation predicts a bottleneck" event triggering the "early warning information push" or "assisted decision analysis" process.

[0082] Time sequence and deadline logic: Injects a time dimension into business processes for monitoring and managing timeliness.

[0083] Time-triggered: For example, the "Production Daily Report Generation" node is automatically activated at the end of each shift.

[0084] Deadline monitoring: For example, set a "maximum time limit" for the "processing step" node; if the time limit is exceeded, an "overtime alarm" will be automatically triggered and relevant personnel will be notified.

[0085] It should be further explained that in most manufacturing industries, if the fluctuations of key parameters such as production cycle time and equipment hours can be controlled within 5%, the model is generally considered to have extremely high confidence and can accurately reflect the operating state of the physical system. This level of accuracy is sufficient to support subsequent simulation, prediction and optimization decisions, ensuring that the "predicted results" and "optimal production scheduling schemes" derived from the model are reliable and credible.

[0086] Setting the threshold too low (e.g., less than 2%) will cause model calibration alarms to be triggered frequently. Production environments contain a large number of random, normal, and minor fluctuations. Treating these as model inaccuracies will cause the system to perform unnecessary calibrations continuously, increasing the computational burden and potentially reducing the model's generalization ability due to "overfitting" noisy data.

[0087] Setting the threshold too high (e.g., greater than 10%) means that the system will not issue an alarm even when the model has already shown significant deviation. By the time the deviation accumulates to such a large extent, the resulting erroneous predictions and scheduling schemes may have already had a substantial impact on actual production. At this point, it is too late to correct the error, and the predictive maintenance and forward-looking scheduling have lost their meaning.

[0088] S2: Real-time event-driven simulation and prediction: Using the integrated digital twin model constructed by S1, predictive analysis of future production status is achieved.

[0089] S2.1: Real-time state event capture and driving:

[0090] Event Listening and Receiving: The integrated digital twin model continuously monitors "real-time status events" reported by the physical production system through message queues or API interfaces via event listeners deployed in the system. These events carry key information such as event type, occurrence entity ID, and timestamp.

[0091] Event-driven node activation: When an event is received, the model fusion engine immediately parses the event and dynamically activates the task nodes related to the event in the workflow model according to the mapping relationship established in S1.

[0092] S2.2: Simulation of the future preset time period:

[0093] Simulation engine startup: Starting with the activated node, the discrete event simulation engine built into the integrated digital twin model begins to work.

[0094] Accelerated simulation process: Based on the current system's complete state (from the digital twin) and activated process logic (from the workflow model), the simulation engine simulates the production process for a preset period (e.g., the next 8 hours) in a virtual environment at a speed faster than real time. During this process, the estimated start / end time and resource usage of each task are calculated.

[0095] S2.3: Output the prediction results:

[0096] Bottleneck and conflict detection algorithm: During the simulation process, the simulation engine runs the analysis algorithm to monitor the task queue length and device load rate in real time.

[0097] Resource load factor and bottleneck prediction:

[0098]

[0099]

[0100] in, For resources In the future Load rate at any given time This refers to the busy time of the resource within the simulation time slice. This represents the total simulation time. For the work center exist The queue length at any given time (i.e., the number of work orders waiting to be processed or the total working hours). When Persistently greater than 85% or When the trend is monotonically increasing, the system determines that the resource or work center is a potential bottleneck.

[0101] Structured report generation: After the simulation is completed, the system automatically generates a "prediction result" report that includes the specific bottleneck location, the expected occurrence time, the scope of impact, and the type of conflicting resources.

[0102] It should be further noted that the event types include: production process execution events, resource status change events, quality and anomaly events, external instruction and plan change events, and timing and periodic events.

[0103] The production process execution events include:

[0104] Work order initiation: A production work order is officially initiated at a specific piece of equipment or workstation;

[0105] Work order completion: A production work order is successfully completed at a specific equipment or workstation;

[0106] Process changeover: The transfer of materials or semi-finished products from one process to the next.

[0107] Batch creation / closing: In batch production mode, the creation or completion of a new batch;

[0108] Order receipt / shipment: Raw materials entering the warehouse or finished products leaving the warehouse;

[0109] Among them, resource status change events:

[0110] Equipment failure: Critical production equipment stops operating due to abnormality;

[0111] Equipment restoration: The faulty equipment is repaired and put back into use;

[0112] Equipment downtime: Planned downtime, such as maintenance, mold replacement, and cleaning;

[0113] Equipment startup: The equipment stops shutting down and enters the ready state;

[0114] Material shortage: Detection indicates that the inventory of materials required for the production line is below the safety threshold or has been depleted;

[0115] Tool Ready / Out of Service: The tools and fixtures required for production are ready or worn out and unusable.

[0116] Quality and abnormal events include:

[0117] Quality spot check trigger: When the preset spot check point or quantity is reached;

[0118] Quality Inspection Results (Pass / Fail): Report on the results of quality inspection;

[0119] Process parameters out of tolerance: Real-time monitored process data (such as temperature and pressure) exceed the allowable range;

[0120] Triggering rework: Due to quality issues, a work order is determined to require reprocessing;

[0121] Trigger scrapping: The material or semi-finished product is determined to be unusable.

[0122] External instructions and plan change events include:

[0123] Urgent order insertion: A new order that requires priority processing has been received;

[0124] Order Change / Cancellation: Information about an existing order has been modified or cancelled;

[0125] Production planning and scheduling: New production plans are issued from the ERP / MES system to the production units;

[0126] Priority adjustment: The priority of a work order or order is manually adjusted;

[0127] Among them, time-series and periodic events include:

[0128] Shift start / end: ​​Marks the start or end of a work shift;

[0129] Timed data acquisition: The system collects status data (which can also be regarded as an event) at fixed time intervals.

[0130] S3: Generate alternative production scheduling solutions based on prediction results: Transform predictive insights into specific and actionable action plans.

[0131] S3.1: Optimization Engine Input and Problem Definition:

[0132] Input data preparation: The optimization engine receives the "prediction results" from S2, as well as the real-time system status and preset production constraints (such as process routes, equipment capacity, and delivery time) from the integrated digital twin model.

[0133] Multi-objective function establishment: The optimization engine is configured to simultaneously optimize three objectives: "equipment utilization," "order delivery time," and "production cost." Mathematically, this is expressed as finding a scheduling scheme X such that... It reaches the Pareto optimal frontier.

[0134]

[0135]

[0136]

[0137] S3.2: Collaborative Optimization and Alternative Solution Generation

[0138] This transforms predictive insights into a set of feasible and efficient scheduling solutions. At its core is a collaborative optimization mechanism, which defines a multi-objective optimization problem and employs advanced algorithms to ensure that the output solution comprehensively balances multiple key performance indicators.

[0139] 1. Problem Modeling and Input Definition

[0140] The optimization engine first formalizes the scheduling problem as a multi-objective optimization model:

[0141] Decision variables: mainly the start time of the work order on the equipment, resource allocation sequence, etc.

[0142] Input data: real-time system status (such as equipment availability and work-in-process inventory) provided by the integrated digital twin model, prediction results from S2 (such as bottleneck resource information), and basic production constraints (such as process routes, equipment capacity, and order delivery dates).

[0143] Optimization Objective: Simultaneously optimize the following three key objectives to construct a set of objective functions:

[0144] Maximize equipment utilization: reduce equipment idle time and increase overall output.

[0145] Minimize order delays: Ensure orders are delivered on time or as early as possible.

[0146] Minimize production costs: comprehensively consider energy consumption, material consumption, and switching costs.

[0147] 2. Execution of multi-objective optimization algorithm

[0148] To resolve potential conflicts between the aforementioned objectives (e.g., pursuing maximum utilization may lead to delays in some orders), the optimization engine employs a multi-objective evolutionary algorithm (such as NSGA-II) for solution. Its execution flow is as follows:

[0149] Population initialization: Randomly generate an initial population representing different scheduling schemes.

[0150] Iterative evolution: Through operations such as selection, crossover, and mutation, it simulates the natural evolutionary process and continuously generates new solutions.

[0151] Pareto sorting and elite retention: In each generation, the algorithm sorts the solutions according to the Pareto dominance relationship, selects the non-dominated solutions that perform well on multiple objectives and cannot be completely surpassed by other solutions, and retains them to the next generation.

[0152] 3. Output: Pareto optimal solution set

[0153] After multiple iterations and convergence, the optimization engine finally outputs a Pareto optimal solution set. Each "alternative production scheduling scheme" in this solution set represents an optimal trade-off between the three objectives of "equipment utilization", "order delivery time" and "production cost".

[0154] Option A: This option may prioritize ensuring delivery time, allowing for a slight decrease in equipment utilization.

[0155] Option B: This option may prioritize maximizing equipment utilization and accept delays in some orders.

[0156] Option C: Seek the production path with the lowest cost.

[0157] It should be further explained that multi-objective evolutionary algorithms like NSGA-II were chosen because they are particularly suitable for solving production scheduling problems with discrete variables, complex constraints, and nonlinear relationships. Through a population evolution mechanism, it can explore different regions of the solution space in parallel, thus efficiently approximating the entire Pareto optimal front. To ensure its practicality in real-time decision-making, the system typically sets upper limits on the number of iterations or computation time to ensure that a set of high-quality alternative solutions is output within an acceptable timeframe, meeting the timeliness requirements of decision-making in production settings.

[0158] S4: Scheme evaluation and decision-making based on parallel simulation: verifying and selecting the optimal scheme in a virtual environment.

[0159] S4.1: Parallel simulation execution:

[0160] Simulation environment replication: For each "alternative production scheduling scheme" from S3, the system creates an identical simulation environment copy in the integrated digital twin model. The copy creation process supports dynamic resource allocation, avoiding resource contention when multiple schemes are simulated in parallel.

[0161] Scheme Injection and Derivation: Inject all alternative schemes into their corresponding simulation replicas, and drive all replicas to perform parallel simulation derivation of future production cycles based on their respective schemes, starting from the same initial state; a new simulation time synchronization calibration mechanism is added to ensure that the simulation progress deviation of all replicas does not exceed 1%.

[0162] S4.2: Quantitative Assessment and Selection:

[0163] KPI Calculation: After the simulation ends, the system automatically calculates a set of predefined "key performance indicators" based on the data recorded during the operation of each scheme replica, including the overall utilization rate of equipment (…). On-time delivery rate of orders It includes metrics such as total production cost and bottleneck mitigation rate; it also supports user-defined KPI weights to adapt to different business priority requirements.

[0164] Comparison and Decision-Making: The system visualizes and compares the KPI results of all options on a decision dashboard. Managers, based on business objectives (such as prioritizing delivery) or through system-preset decision rules (such as weighted scoring), "quantitatively evaluate and select" the "optimal production scheduling plan" with the best overall performance from the alternatives.

[0165] It needs to be further explained that,

[0166] KPI Calculation Basis and Data Sources: All KPIs are not estimates or theoretical values, but are calculated based on high-fidelity time-series data recorded during parallel simulation operations. This data originates from the integrated digital twin model's complete record of the status changes of all entities (equipment, materials, orders) in the virtual production environment when simulating the execution of various alternative scenarios, ensuring the authenticity and traceability of the evaluation data.

[0167] Definition and calculation logic of core KPIs: This system typically includes, but is not limited to, the following core KPIs, which directly correspond to the optimization objectives in S3 and cover the core concerns of production management:

[0168] Overall Equipment Effectiveness (OEE): Measures the actual efficiency of equipment value creation; it is the product of availability, performance, and yield.

[0169] Calculation: In the simulation, the system accurately records the planned start-up time, actual running time, ideal cycle time, and number of qualified products for each device. The calculation formula is: OEE = Availability × Performance Rate × Yield Rate. This KPI can help identify solutions that appear highly productive but may mask frequent downtime, speed losses, or quality issues.

[0170] On-Time Delivery Rate (OTD): The percentage of production orders completed within the promised delivery period out of the total number of orders.

[0171] Calculation: OTD = (Number of on-time delivered orders / Total number of orders) × 100%. The simulation engine will label each order as "on-time" or "delayed". This KPI is the most direct indicator for evaluating whether the solution can meet customer needs.

[0172] Total production cost: The sum of all direct and indirect costs incurred in executing this scheduling scheme during the simulation period.

[0173] Calculation: This is an aggregated calculated value, which consists of:

[0174] Fixed costs: such as equipment depreciation and the amortization of factory rent.

[0175] Variable costs: such as direct material consumption and energy consumption (calculated based on equipment power consumption and running time in simulation).

[0176] Additional costs: such as setup costs due to product switching, outsourcing processing costs, and penalty costs for order delays.

[0177] Work-in-process inventory level: The quantity (or value) of materials and semi-finished products being processed or awaiting processing on the production line.

[0178] Calculation: During the simulation, the system periodically samples the queue lengths between all processes and calculates their average or peak values. Lower work-in-process inventory levels generally mean a smoother production flow and shorter manufacturing cycles.

[0179] Production efficiency: The number of qualified products produced by the system per unit of time.

[0180] Calculation: Production efficiency = Total qualified output / Total simulation time. This KPI directly reflects the production capacity and output speed of the solution.

[0181] KPI Standardization and Comprehensive Evaluation: To conduct comprehensive comparisons across different options, the system employs data standardization methods (such as Min-Max normalization) to convert KPIs of different dimensions (such as percentages, monetary units, and numbers) into a uniform range of [0,1]. Subsequently, based on preset decision weights, each KPI for each option is weighted and scored, ultimately calculating the comprehensive performance score for each option.

[0182] S5: Solution Execution and Closed-Loop Monitoring: Putting decisions into practice and forming a learning loop.

[0183] S5.1: Command Parsing and Issuance:

[0184] Instruction conversion: The system parses the selected "optimal production scheduling scheme" through the "instruction parser" and converts the scheduling instructions into "control instructions" that can be recognized and executed by the underlying control system (such as PLC).

[0185] Command issuance: These control commands are safely and accurately issued to the execution equipment in the physical production system through a highly reliable "industrial network" in accordance with the command sequence and timing requirements.

[0186] S5.2: Real-time monitoring and feedback:

[0187] Execution effect monitoring: During instruction execution, the integrated digital twin model continuously receives monitoring data from the physical system (such as actual equipment status, output, and energy consumption) and drives the virtual model to update synchronously, thereby achieving "real-time monitoring" of the "execution effect of control instructions in the physical production system".

[0188] Data feedback: This monitoring data is fed back to the integrated digital twin model in real time.

[0189] S5.3: Model self-calibration:

[0190] Deviation analysis: The system compares the actual execution data with the simulation expected data to identify execution deviations (such as the actual processing time being longer than expected).

[0191] Parameter calibration: Based on continuously accumulated deviation data, the system automatically or with manual assistance calibrates the model parameters (such as equipment standard working hours) in the digital twin, achieving "model self-correction". The calibrated model will be used in subsequent optimization cycles, making the system's predictions and decisions increasingly accurate.

[0192] It is necessary to further explain that the virtual-real synchronization channel built by real-time monitoring is far more than traditional data collection. It constructs a "neural transmission" pathway connecting the physical entity and the digital twin, which is the cornerstone for achieving virtual-real synchronization.

[0193] Data-driven model updates: Monitoring data reported by the physical production system is injected in real time into the corresponding virtual entity in the digital twin. This makes the digital twin no longer a static model, but a "living" mirror that can synchronously or even proactively reflect the true state of the physical system.

[0194] Providing a factual basis for deviation analysis: This high-fidelity virtual-real synchronization enables the system to perform fine-grained time-series data comparisons.

[0195] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0196] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An enterprise production management method based on digital twins and workflow simulation, characterized in that, include: S1: Construct a digital twin of the physical elements and real-time operational data of the physical production system; Based on the production management business process, a workflow model containing task nodes and logical rules is constructed; through the model fusion engine, the logical nodes in the workflow model are dynamically associated and semantically bound with the corresponding entities in the digital twin, forming an integrated digital twin model that integrates data and logic. S2: During the operation of the integrated digital twin model, the corresponding task nodes in the workflow model are dynamically activated by the real-time status events reported by the physical production system, and the production process in the future preset time period is simulated and the prediction results of production bottlenecks and resource conflicts are output. S3: Based on the prediction results, multiple alternative production scheduling schemes are output through the optimization engine; wherein, the optimization engine is configured to collaboratively optimize multiple objectives such as equipment utilization, order delivery time and production cost; S4: Input the multiple alternative production scheduling schemes into the integrated digital twin model and perform simulations in parallel. By comparing the key performance indicators output by the simulation, quantitatively evaluate and select the optimal production scheduling scheme. S5: The selected optimal production scheduling scheme is parsed into control commands that can be executed by the underlying control system and sent to the execution equipment in the physical production system through the industrial network; at the same time, the execution effect of the control commands in the physical production system is monitored in real time through the integrated digital twin model, and the monitoring data is fed back to the integrated digital twin model for model self-correction and subsequent optimization process input.

2. The enterprise production management method based on digital twin and workflow simulation according to claim 1, characterized in that, The construction of the digital twin includes: Digitalization of physical elements: Using 3D modeling software or laser scanning technology to create geometric models of equipment, materials and tools on the production line to form virtual representations; Real-time data access and mapping: Collect equipment status, process parameters, and material information through sensors, PLCs, and manufacturing execution systems, and bind the data to the corresponding virtual entities in the digital twin in real time; Model Validation and Calibration: The accuracy of the model is verified by comparing historical data and simulation tests. A confidence assessment mechanism based on mean absolute percentage error is introduced, and an automatic calibration alarm is triggered when the MAPE continuously exceeds 5%.

3. The enterprise production management method based on digital twin and workflow simulation according to claim 1, characterized in that, The construction of the workflow model includes: Business process analysis: Identify key business processes in production management, including order processing, production planning, material scheduling, quality control, and equipment maintenance; Logical rule definition: Use workflow modeling tools to define the logical rules between nodes, including sequential flow, conditional branching, event response, timing and deadline logic.

4. The enterprise production management method based on digital twin and workflow simulation according to claim 1, characterized in that, The integrated digital twin model includes: Establishment of the association mapping table: The model fusion engine maintains the "entity-node" mapping table to clarify the correspondence between logical nodes in the workflow model and entities in the digital twin; Semantic binding implementation: Achieving semantic consistency between entities and nodes through unified attribute tags; The operation of the fusion model enables dynamic interaction and synchronous updates between business process logic and physical entity status.

5. The enterprise production management method based on digital twin and workflow simulation according to claim 1, characterized in that, The prediction results include: Based on real-time monitoring of task queue length and device load rate during the simulation process, potential bottlenecks and resource conflicts are identified. Generate a structured report that includes the bottleneck location, estimated time of occurrence, scope of impact, and types of conflicting resources.

6. The enterprise production management method based on digital twin and workflow simulation according to claim 1, characterized in that, The alternative production scheduling schemes include: A Pareto optimal solution set is generated through a multi-objective optimization engine, and each solution achieves the optimal trade-off between equipment utilization, order delivery time and production cost. A multi-objective evolutionary algorithm is used to solve the problem, and multiple non-dominated solutions are output as alternatives.

7. The enterprise production management method based on digital twin and workflow simulation according to claim 1, characterized in that, The key performance indicators include: Equipment utilization rate, on-time order delivery rate, total production cost, work-in-process inventory level, and production efficiency; The calculation is based on time-series data output from parallel simulation, and a comprehensive evaluation is performed through standardization and weighted scoring.

8. The enterprise production management method based on digital twin and workflow simulation according to claim 1, characterized in that, The real-time monitoring includes: The optimal production scheduling scheme is parsed into control commands and sent to the execution equipment through the industrial network; By continuously receiving feedback data from the physical system through an integrated digital twin model, the virtual model is updated synchronously, enabling real-time monitoring of execution results.

9. The enterprise production management method based on digital twin and workflow simulation according to claim 1, characterized in that, The model self-calibration and subsequent optimization include: Deviation analysis: Compare the actual execution data with the simulation expected data to identify execution deviations; Parameter calibration: Automatic or assisted calibration of model parameters in the digital twin based on deviation data, which is used as input for subsequent optimization processes.

Images