CN-121980919-A - Digital twin fidelity monitoring and man-machine collaborative production intelligent scheduling method and system

CN121980919ACN 121980919 ACN121980919 ACN 121980919ACN-121980919-A

Abstract

The application discloses a digital twin fidelity monitoring and man-machine collaborative production intelligent scheduling method and system, and belongs to the technical field of intelligent manufacturing and industrial Internet. The method comprises the steps of constructing a digital twin model for reflecting real-time states of a physical workshop according to real-time production data, generating an initial production scheduling scheme based on the digital twin model according to production constraint and an optimization target, simulating deduction and performance evaluation of the initial production scheduling scheme, presenting deduction results and evaluation data to a target object, receiving an input scheduling adjustment instruction of the target object, fusing the scheduling adjustment instruction with the initial production scheduling scheme, generating a final production scheduling scheme, converting the final production scheduling scheme into a corresponding scheduling instruction, and then sending the corresponding scheduling instruction to a physical production system for execution, and calculating fidelity by comparing predicted data of the digital twin model with actual operation data of a physical entity in the process of executing production according to the scheduling instruction by the physical production system, wherein model updating and/or scheduling re-planning are triggered when the fidelity is lower than a preset threshold.

Inventors

YANG LEI
XIN QI
YANG FEI
LI TAO
Zhang yasen
XIONG CHENG
Song Genyuan
Qiao Zhefei
LU XIN

Assignees

中信重工机械股份有限公司
中国中信有限公司

Dates

Publication Date: 20260505
Application Date: 20251226

Claims (10)

1. The intelligent scheduling method for the production of the digital twin fidelity monitoring and man-machine cooperation is characterized by comprising the following steps of: collecting real-time production data of a physical workshop, and constructing a digital twin model for reflecting the real-time state of the physical workshop according to the real-time production data; Based on the digital twin model, generating an initial production scheduling scheme according to production constraint and an optimization target, and performing simulation deduction and performance evaluation on the initial production scheduling scheme to obtain corresponding simulation deduction results and performance evaluation data; presenting the simulation deduction result and the performance evaluation data to a target object in a visual form, and receiving a scheduling adjustment instruction input by the target object; fusing the scheduling adjustment instruction with the initial production scheduling scheme through a preset decision fusion algorithm to generate a final production scheduling scheme, converting the final production scheduling scheme into a corresponding scheduling instruction, and then issuing the corresponding scheduling instruction to a physical production system for execution; And in the process that the physical entity of the physical production system executes production according to the scheduling instruction, performing fidelity calculation and monitoring by comparing the predicted data of the digital twin model with the actual operation data of the physical entity, and triggering the updating of the digital twin model and/or the re-planning of the scheduling when the fidelity is lower than a preset threshold.
2. The method of claim 1, wherein the real-time production data comprises equipment operation data reported in real time by physical equipment in the physical plant, bill of materials and process work orders of a manufacturing execution system, inventory and distribution information of a warehouse management system, and production orders of an enterprise resource planning system; an operation of constructing a digital twin model for reflecting the real-time status of the physical plant from the real-time production data, comprising: Establishing an equipment state model for dynamically reflecting the equipment operation state, the processing capacity and the health degree in a virtual space according to the equipment operation data, wherein the equipment operation data comprises an on-off state, a spindle rotating speed, a feeding rate, an alarm signal and a program execution section; According to the bill of materials and the inventory and distribution information, a material circulation model for tracking the identity, the quantity, the position and the consumption state of materials in real time is established in a virtual space; Establishing an order progress model for dynamically calculating order completion progress, predicting delivery time and identifying bottleneck procedures in a virtual space according to the production order and the procedure work order; Integrating the equipment state model, the material circulation model and the order progress model through a unified data interface and a clock synchronization mechanism to form a digital twin model which is mapped with the physical workshop in real time.
3. The method of claim 1, wherein generating an initial production schedule based on the digital twin model according to production constraints and optimization objectives comprises: Acquiring available processing equipment, a bill of materials to be produced and a production order to be processed from the digital twin model as basic elements of production scheduling; Setting the equipment processing capacity, the material packing time and the order delivery period as production constraints, and setting at least one of maximizing production efficiency, minimizing production cost or minimizing order delay as an optimization target; And calling a preset scheduling algorithm, generating a plurality of candidate production scheduling schemes based on the basic elements and the production constraint, and screening the candidate production scheduling schemes according to the optimization target to obtain the initial production scheduling scheme.
4. The method of claim 2, wherein performing the simulation deduction and the performance evaluation on the initial production schedule to obtain corresponding simulation deduction results and performance evaluation data comprises: Inputting the initial production scheduling scheme into the digital twin model, driving the equipment state model, the material circulation model and the order progress model, performing collaborative simulation operation according to time logic and resource constraint defined by the initial production scheduling scheme, and simulating a complete production execution process; In the collaborative simulation operation process, recording simulation state data which is output by the digital twin model and is related to a time sequence as a corresponding simulation deduction result, wherein the simulation state data at least comprises a device load rate sequence, a material consumption time sequence and an order progress time sequence; And calculating a group of evaluation indexes for quantifying the performance of the initial production scheduling scheme based on all the recorded simulation state data as corresponding performance evaluation data, wherein the evaluation indexes at least comprise equipment average utilization rate, total production period, order on-time delivery rate and total production capacity.
5. The method of claim 1, wherein the operation of fusing the schedule adjustment instructions with the initial production schedule to generate a final production schedule via a predetermined decision fusion algorithm comprises: analyzing the scheduling adjustment instruction, and identifying adjustment content and adjustment intention appointed by the target object aiming at equipment, materials or orders; Locally modifying the initial production schedule based on the adjustment content and the adjustment intent to generate one or more modified candidate schedules; The initial production scheduling scheme and the one or more corrected candidate scheduling schemes are used as a scheduling scheme set to be decided together; and calling the decision fusion algorithm, comprehensively grading each scheme in the scheduling scheme set based on the optimization target, and selecting the scheme with the highest comprehensive grading as the final production scheduling scheme.
6. The method of claim 1, wherein performing the operations of fidelity calculation and monitoring by comparing the predicted data of the digital twin model with actual operational data of the physical entity comprises: acquiring actual operation data of physical entities in the physical production system in real time, wherein the actual operation data at least comprises equipment actual state data, material actual position data and order actual progress data; comparing the actual running data with the predicted data which are output by the digital twin model at the same moment and correspond to the same physical entity, and respectively calculating fidelity deviation values of equipment states, material circulation and order progress in a preset time window; Based on the equipment state, the material circulation and the fidelity deviation value of the order progress, the overall fidelity of the digital twin model is obtained through weighted average calculation; And monitoring the overall fidelity and the fidelity deviation values of each item in real time.
7. The method of claim 1, wherein triggering the digital twin model update when the fidelity is below a preset threshold comprises: Identifying one or more key fidelity deviation terms that result in overall fidelity being below a preset threshold; Positioning a corresponding physical entity and a target sub-model in the digital twin model according to the identified key fidelity deviation term; recalibrating and calibrating parameters of the target sub-model based on the currently acquired actual operation data; and replacing the corresponding part in the digital twin model by using the target sub-model after recalibration to finish model updating.
8. The method of claim 1, wherein triggering the operation of scheduling re-planning when the fidelity is below a preset threshold comprises: Suspending a scheduling instruction which is currently being executed, and recording the current state of the physical production system as an initial state of re-planning; inputting the current state into the updated digital twin model, and regenerating a new production scheduling scheme based on the latest production constraint and optimization target; And performing simulation deduction and performance evaluation on the new production scheduling scheme, and if the evaluation result meets the preset requirement, issuing and executing the new production scheduling scheme as a new final production scheduling scheme.
9. A storage medium comprising a stored program, wherein the method of any one of claims 1 to 8 is performed by a processor when the program is run.
10. The intelligent scheduling system for the production of the digital twin fidelity monitoring and man-machine cooperation is characterized by comprising a data acquisition and integration layer, a digital twin construction layer, an intelligent scheduling core layer, a scheduling execution and interaction layer and a monitoring and analysis application layer; The data acquisition and integration layer is used for acquiring real-time production data of the physical workshop; The digital twin construction layer is used for constructing a digital twin model for reflecting the real-time state of the physical workshop according to the real-time production data; the intelligent scheduling core layer is used for generating an initial production scheduling scheme based on the digital twin model according to production constraint and optimization targets, and performing simulation deduction and performance evaluation on the initial production scheduling scheme to obtain corresponding simulation deduction results and performance evaluation data; The scheduling execution and interaction layer is used for presenting the simulation deduction result and the performance evaluation data to a target object in a visual form and receiving a scheduling adjustment instruction input by the target object; The intelligent scheduling core layer is further used for fusing the scheduling adjustment instruction with the initial production scheduling scheme through a preset decision fusion algorithm to generate a final production scheduling scheme; the scheduling execution and interaction layer is also used for converting the final production scheduling scheme into a corresponding scheduling instruction and then issuing the corresponding scheduling instruction to a physical production system for execution; And the monitoring and analyzing application layer is used for performing fidelity calculation and monitoring by comparing the predicted data of the digital twin model with the actual operation data of the physical entity in the process that the physical entity of the physical production system executes production according to the scheduling instruction, and triggering the updating of the digital twin model and/or the re-programming of the scheduling when the fidelity is lower than a preset threshold value.

Description

Digital twin fidelity monitoring and man-machine collaborative production intelligent scheduling method and system Technical Field The application relates to the technical field of intelligent manufacturing and industrial Internet, in particular to a digital twin fidelity monitoring and man-machine collaborative production intelligent scheduling method and system. Background Production scheduling is a key element of the manufacturing industry. The existing production scheduling technology generally has the following problems: (1) Static and curing the scheduling plans generated by conventional APS systems are typically static and difficult to handle in-plant real-time disturbances. (2) Information lag and inaccuracy-scheduling computation dependent data tends to be outdated or incomplete, resulting in a planned and realistic disjoint. (3) AI model drift although there have been studies attempting to schedule using Artificial Intelligence (AI) algorithms, AI models are trained based on historical data. As the physical equipment wears, the process parameters change, the AI Model becomes progressively misaligned, creating a Model Drift (Model Drift), resulting in a less-than-optimal decision. (4) Black box decision is in contrast to man-machine, and AI scheduling (especially deep reinforcement learning) results are often black boxes, lacking in interpretability. Workshop management personnel and operators do not understand the basis of AI decision, or AI decision does not consider implicit experience (Tacit Knowledge, such as that a certain machine is out of pair today), so that the scheduling of AI is not trusted and executed, and the man-machine opposition is caused, and the system cannot fall to the ground. (5) The inefficiency of simulation deduction is that when disturbance such as equipment failure occurs, a manager needs to perform 'What-if' analysis, but traditional simulation modeling is slow and long-running, and cannot meet real-time decision requirements. Therefore, how to solve the problems of static solidification, easy drift and misalignment of the AI model, lack of trust coordination between human and machine, and slow disturbance deduction in the existing production scheduling scheme is a problem to be solved in the application. Disclosure of Invention The embodiment of the disclosure provides a production intelligent scheduling method and system for digital twin fidelity monitoring and man-machine cooperation, which at least solve the technical problems of static solidification of a production scheduling scheme, easy drifting and misalignment of an AI model, lack of trust cooperation of a man-machine and slow disturbance deduction in the prior art. According to one aspect of the embodiment of the disclosure, a production intelligent scheduling method with collaborative digital twin fidelity monitoring and man-machine is provided, and the production intelligent scheduling method comprises the steps of collecting real-time production data of a physical workshop, constructing a digital twin model for reflecting the real-time state of the physical workshop according to the real-time production data, generating an initial production scheduling scheme according to production constraint and optimization targets based on the digital twin model, carrying out simulation deduction and performance evaluation on the initial production scheduling scheme to obtain corresponding simulation deduction result and performance evaluation data, presenting the simulation deduction result and the performance evaluation data to a target object in a visual mode, receiving scheduling adjustment instructions input by the target object, fusing the scheduling adjustment instructions with the initial production scheduling scheme through a preset decision fusion algorithm, generating a final production scheduling scheme, converting the final production scheduling scheme into corresponding scheduling instructions, then sending the corresponding scheduling instructions to a physical production system for execution, and carrying out comparison between the physical entity of the physical production system and the physical entity of the physical production system according to the scheduling instructions, and triggering the real-time accuracy of the real-time scheduling data to be updated when the digital fidelity monitoring and the twin model is triggered, and the real-time accuracy of the real-time data is updated or the real-time scheduling data is calculated. According to another aspect of the embodiments of the present disclosure, there is also provided a storage medium including a stored program, wherein the method described above is performed by a processor when the program is run. According to another aspect of the disclosed embodiments, there is also provided a production intelligent scheduling system with cooperation of a digital twin-type fidelity monitoring and man-machine, comprising a data acquisition and integratio