Dynamic scheduling system and method for intelligent stamping production line driven by industrial large model

Abstract

The invention relates to an intelligent stamping production line dynamic scheduling system and method driven by an industrial large model. According to the technical scheme, the intelligent stamping production line dynamic scheduling system comprises the steps that S1, an intelligent agent model of equipment in a stamping production line is constructed; s2, a reinforcement learning strategy is initialized, and a space state is initially defined; s3, realizing data interaction between the agent model and the large language model through an information interaction layer; s4, a decision execution layer performs strategy optimization according to the scheduling suggestion of the large language model, and issues and executes a production instruction; and S5, collecting real-time production data through a feedback optimization layer, and continuously iterating and optimizing a scheduling strategy. The functions of the intelligent agent model are effectively enhanced by applying technologies such as a dual-trigger mechanism and reinforcement learning, the flexibility of the intelligent agent is enhanced whether learning and adaptation of the surrounding environment are combined or priority queue dynamic scheduling is carried out according to dual-trigger conditions of time and events, and division and cooperation of the intelligent agent models between the stamping production lines are promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent production scheduling in workshops, and in particular to a dynamic scheduling system and method for an intelligent stamping production line driven by an industrial large model. Background Art

[0002] Traditional scheduling methods are facing increasingly significant limitations in the production process of intelligent stamping lines. The primary issue lies in their overreliance on the fixed production cycle of the presses and the relatively fixed movement speed of the robotic arms that move the products, which limits scheduling flexibility. This rigid scheduling approach struggles to adapt flexibly to the complex and ever-changing production demands of the workshop and its actual realities, resulting in reduced production efficiency and unnecessary resource consumption.

[0003] Further analysis reveals that the actual production conditions on a shop floor are complex and ever-changing. Traditional scheduling methods often struggle to cope with these complex and ever-changing production demands, unable to quickly adjust to meet new production requirements. Furthermore, traditional scheduling methods often lack a scientific basis for resource allocation, leading to wasted resources and increased production costs. Therefore, to improve production efficiency and reduce unnecessary resource consumption, a more flexible and intelligent scheduling system is urgently needed to optimize the production process. Summary of the Invention

[0004] In response to the problems of rigidity, poor flexibility and poor adaptability of existing scheduling methods to the production environment, the present invention designs and develops a dynamic scheduling system and method for intelligent stamping production lines driven by industrial large models. Through the infrastructure layer, information interaction layer, decision execution layer, and feedback optimization layer, combined with a dual-trigger mechanism priority queue dynamic scheduling algorithm and a composite method of adaptive scheduling based on reinforcement learning, the dual-trigger mechanism, reinforcement learning and other technologies are used to effectively enhance the functions of the intelligent body model itself. Whether it is combined with learning and adaptation of the surrounding environment, or dynamic scheduling of priority queues based on dual trigger conditions of time and events, the flexibility of the intelligent body is enhanced, and the division of labor and cooperation of intelligent body models between stamping production lines is promoted.

[0005] To achieve the above objectives, the technical solution of the present invention is: The dynamic scheduling method of intelligent stamping production line driven by industrial large model includes the following steps: S1. Build an intelligent agent model of the equipment in the stamping production line; S2. Initialize the reinforcement learning strategy and initially define the spatial state; S3. Implement data interaction between the agent model and the large language model through the information interaction layer; S4. The decision-making execution layer optimizes the scheduling recommendations based on the large language model, issues, and executes production instructions. S5. Collect real-time production data through the feedback optimization layer and continuously iterate and optimize the scheduling strategy.

[0006] Specifically, the step S3 is as follows: S301. Adaptive scheduling method based on reinforcement learning to collect real-time equipment status and production environment data; S302 uses a dual trigger mechanism priority queue dynamic scheduling algorithm, through event trigger and time trigger dynamic update task priority queue; S303. Input the parsed information into the large language model, generate natural language scheduling suggestions and convert them into executable instructions.

[0007] Furthermore, the construction of the agent model in step S1 specifically includes: An intelligent device model is established through point cloud scanning. The intelligent device model includes a perception module, a reasoning module, and an action module, and is capable of receiving environmental information, performing logical reasoning, and selecting an action plan.

[0008] Furthermore, the dual trigger mechanism priority queue dynamic scheduling algorithm in step S302 includes: The event trigger conditions are sudden changes in equipment production rhythm, changes in order priority, or abnormal material supply; the time trigger conditions are periodic global status evaluations.

[0009] Furthermore, the strategy optimization in step S4 is specifically as follows: The intelligent agent model compares the various scheduling schemes generated by the large language model and selects the optimal production rhythm and task allocation scheme through reinforcement learning.

[0010] Furthermore, the data collected by the feedback optimization layer in step S5 includes production rhythm, product production steps, and product delivery time, and adaptive evolution is achieved by adjusting parameters through a large language model.

[0011] An industrial large-scale model-driven intelligent stamping production line dynamic scheduling system, including: The infrastructure layer, information interaction layer, decision execution layer and feedback optimization layer are connected in sequence to implement the steps of the dynamic scheduling method.

[0012] Furthermore, the infrastructure layer includes agent model construction and reinforcement learning strategy initialization.

[0013] Furthermore, the information interaction layer includes information input, information analysis, information request, and command generation.

[0014] Furthermore, the decision execution layer includes agent strategy optimization, instruction issuance and physical execution.

[0015] Beneficial effects of the present invention: 1. By building intelligent agent models of the press and robotic arm, connecting them to an interactive interface with a large industrial language model, and leveraging a dual-trigger priority queue dynamic scheduling algorithm and reinforcement learning to rapidly learn emergency response strategies, we achieved a shift in production scheduling from relying on a fixed production pace to dynamic adjustments. The intelligent agent can communicate with the large language model in real time, flexibly adjusting the production pace based on the actual shop floor conditions. This significantly enhances the flexibility of the scheduling system, enabling the production line to more efficiently respond to changes in production demand. 2. The intelligent agent model compares the various scheduling suggestions provided by the large language model to quickly determine and execute the optimal production execution plan. Furthermore, the intelligent interaction between the robot agent and the large model enables more precise delivery timing, reduces waiting time and ineffective movements, and significantly improves production efficiency and response speed. 3. Driven by a large industrial model, the intelligent stamping production line achieves flexible production throughout the entire process, from raw material loading to finished product delivery. The intelligent collaboration between the press and robotic arm, along with real-time scheduling recommendations from the large language model, creates a highly flexible and adaptable production system, providing strong support for the intelligent transformation of the manufacturing industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. For those skilled in the art, other drawings can be obtained based on the drawings without paying any creative work.

[0017] Figure 1 It is a system structure diagram of the present invention; Figure 2 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] like Figure 1-2 The industrial large model driven intelligent stamping production line dynamic scheduling method shown includes the following steps: S1. Build an intelligent agent model of the equipment in the stamping production line; S2. Initialize the reinforcement learning strategy and initially define the spatial state; S3. Implement data interaction between the agent model and the large language model through the information interaction layer; S4. The decision-making execution layer optimizes the scheduling recommendations based on the large language model, issues, and executes production instructions. S5. Collect real-time production data through the feedback optimization layer and continuously iterate and optimize the scheduling strategy.

[0020] Furthermore, the step S3 is specifically as follows: S301. Adaptive scheduling method based on reinforcement learning to collect real-time equipment status and production environment data; S302 uses a dual trigger mechanism priority queue dynamic scheduling algorithm, through event trigger and time trigger dynamic update task priority queue; S303. Input the parsed information into the large language model, generate natural language scheduling suggestions and convert them into executable instructions.

[0021] like Figure 1 As shown, a dynamic scheduling system for an intelligent stamping production line driven by an industrial large model includes: an infrastructure layer, an information interaction layer, a decision execution layer, and a feedback optimization layer. Each layer is connected in sequence and implements the steps of the dynamic scheduling method.

[0022] The infrastructure layer includes agent model construction and reinforcement learning strategy initialization.

[0023] The intelligent agent model construction mainly refers to the construction of intelligent agent models for basic equipment in the stamping production line, such as presses and robotic arms, through modeling methods such as point cloud scanning. The intelligent agent can receive and interpret information from the environment, perform logical reasoning based on this information, predict results, and select the best action plan, including perception modules, reasoning modules and action modules.

[0024] The reinforcement learning strategy initialization mainly refers to the initial definition of the space state, including equipment load, order queue, equipment energy consumption, equipment production rhythm, task allocation, etc.

[0025] The information interaction layer is the mechanism and process for information exchange and communication between intelligent agents such as presses and robotic arms and large language models, including input information, information parsing, information request, and command generation.

[0026] The input information mainly refers to the adaptive scheduling method based on reinforcement learning adopted by the intelligent agent model. The intelligent agent learns the optimal decision by interacting with the environment. For example, the intelligent agent learns the equipment status (production rhythm, equipment temperature, equipment vibration, etc.) and the production environment (production time, materials used in production, order priority, etc.). Based on the working results of the stamping equipment, the intelligent agent understands the production status and learns and stores this information.

[0027] Information parsing primarily involves the agent employing a dual-trigger priority queue dynamic scheduling algorithm. Priority queues are set up with one trigger being an event, such as a sudden change in equipment production cycle status, a change in order priority, or an abnormal material supply. The second trigger is a time-based one, such as performing global status assessments at fixed intervals to detect or predict the occurrence of faults. The agent model dynamically calculates task priorities and updates the queue based on real-time status.

[0028] The information interaction mainly refers to the interaction between the intelligent agent model and the large language model, which exchanges the collected information with the large language model.

[0029] The command generation mainly refers to the large language model generating natural language suggestions based on the information learned and collected by the intelligent agent according to the real-time status, converting it into data and providing it to the intelligent agent, performing pre-training and inference, and providing a large number of scheduling solutions.

[0030] The decision-making and execution layer is responsible for formulating and implementing organizational strategies and operational plans, serving as a bridge between the upper and lower levels of the digital smart factory's operations. This includes optimizing agent strategies, issuing commands, and implementing them physically.

[0031] Agent strategy optimization primarily involves the agent generating multiple judgments and solutions based on data exchange and functional integration with a large language model. Through multiple comparisons and decisions, the agent ultimately determines the optimal solution. For example, the press agent compares multiple scheduling suggestions from the large language model and determines the optimal production execution plan. This involves both action selection and online learning.

[0032] The issuance of instructions and physical execution mainly refer to the intelligent agent performing production actions according to the production plan and production rhythm after decision-making.

[0033] In the feedback optimization layer, during the production process, the intelligent model continuously collects production data, such as production rhythm, product production steps, product delivery time, etc., and feeds it back to the large language model for continuous optimization. The entire production scheduling system will gradually become more efficient and flexible.

[0034] like Figure 2As shown, the first step is to build an intelligent agent model capable of perceiving the environment, effectively receiving environmental information, and executing actions based on instructions. This also requires clarifying the agent's goals and functional scope. After completing this foundational work, the architecture should be designed and technology selected. Next, the agent's perception, reasoning, and action capabilities should be developed. Once these tasks are complete, an initial definition of the spatial state is required, specifically encompassing aspects such as equipment load, order queue status, equipment energy consumption, equipment production rhythm patterns, and task allocation mechanisms.

[0035] After connecting the interactive interface between the agent model and the large language model, the scheduling mode shifts to a dialogue between the agent and the large language model regarding the calculation method for product production cycle time. The agent model uses an adaptive scheduling method based on reinforcement learning. The agent learns optimal decisions by interacting with the environment. The agent learns the equipment status, including production cycle time, equipment temperature, and equipment vibration, and the production environment, including production time, materials used, and order priority.

[0036] According to the working results of the stamping equipment, the production status is understood, and this information is learned and stored. The intelligent agent model adopts a dual-trigger mechanism priority queue dynamic scheduling algorithm to set up a priority queue. One of them is event triggering, such as a sudden change in the equipment production rhythm status, a change in order priority, or an abnormal material supply. The second is time triggering, such as performing a global status assessment in a fixed time period to check for faults or predict the time of fault occurrence. The intelligent agent model dynamically calculates the task priority and updates the queue based on the real-time status. The interface connection allows the large language model to receive the production information provided by the intelligent agent, and then summarizes and analyzes it. The large language model generates natural language suggestions based on the real-time status based on the information learned and collected by the intelligent agent, converts it into data and provides it to the intelligent agent, and performs pre-training and reasoning to provide a large number of scheduling solutions.

[0037] The agent generates multiple judgments and solutions based on data exchange and functional integration within the large language model. Through multiple comparisons and decisions, the agent develops the optimal solution. For example, the press agent compares various scheduling suggestions from the large language model and determines the optimal production execution plan.

[0038] During the production process, the intelligent agent and the large language model continue to exchange information. The large language model analyzes the intelligent agent's production rhythm and other information based on real-time production, and adjusts the data to achieve the best production effect.

[0039] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A dynamic scheduling method for intelligent stamping production lines driven by industrial large models, characterized by: The following steps are involved: S1. Build an intelligent agent model of the equipment in the stamping production line; S2. Initialize the reinforcement learning strategy and initially define the spatial state; S3. Implement data interaction between the agent model and the large language model through the information interaction layer; S4. The decision-making execution layer optimizes the scheduling recommendations based on the large language model, issues, and executes production instructions. S5. Collect real-time production data through the feedback optimization layer and continuously iterate and optimize the scheduling strategy.

2. The dynamic scheduling method according to claim 1, characterized in that: The step S3 is specifically as follows: S301. Adaptive scheduling method based on reinforcement learning to collect real-time equipment status and production environment data; S302 uses a dual trigger mechanism priority queue dynamic scheduling algorithm, through event trigger and time trigger dynamic update task priority queue; S303. Input the parsed information into the large language model, generate natural language scheduling suggestions and convert them into executable instructions.

3. The dynamic scheduling method according to claim 1, characterized in that: The construction of the agent model in step S1 specifically includes: An intelligent device model is established through point cloud scanning. The intelligent device model includes a perception module, a reasoning module, and an action module, and is capable of receiving environmental information, performing logical reasoning, and selecting an action plan.

4. The dynamic scheduling method according to claim 2, characterized in that: The dual trigger mechanism priority queue dynamic scheduling algorithm in step S302 includes: Event trigger conditions include sudden changes in equipment production rhythm, changes in order priority, or abnormal material supply; The time trigger condition is the periodic global state evaluation.

5. The dynamic scheduling method according to claim 1, characterized in that: The strategy optimization in step S4 is specifically as follows: The intelligent agent model compares the various scheduling schemes generated by the large language model and selects the optimal production rhythm and task allocation scheme through reinforcement learning.

6. The dynamic scheduling method according to claim 1, characterized in that: The data collected by the feedback optimization layer in step S5 include production rhythm, product production steps, and product delivery time, and adaptive evolution is achieved by adjusting parameters through a large language model.

7. An industrial large model driven intelligent stamping production line dynamic scheduling system, characterized by: include: The infrastructure layer, information interaction layer, decision execution layer and feedback optimization layer are connected in sequence and implement the steps of any one of the methods described in claims 1-6.

8. The dynamic scheduling system according to claim 7, characterized in that: The infrastructure layer includes agent model construction and reinforcement learning strategy initialization.

9. The dynamic scheduling system according to claim 7, characterized in that: The information interaction layer includes information input, information analysis, information request, and command generation.

10. The dynamic scheduling system according to claim 7, characterized in that: The decision execution layer includes agent strategy optimization, instruction issuance and physical execution.

Images