基于DeepSeek-RL的生产调度优化范式:以孤岛装配为例

Large Language Models (LLMs). represented by ChatGPT and DeepSeek. are bringing revolutionary power to smart manufacturing. The fusion of LLM and production scheduling is the currently hotspot. Therefore, a production scheduling optimization method based on DeepSeek-Rl was proposed. The method introduced the natural language processing and code generation capabilities of DeepSeek-Rl into the scheduling task of the island assembly line, and a scheduling optimization process that includes three phases: prompt design, modeling and solving, and it¬erative optimization was constructed. In the prompt design phase, the island assembly scheduling problem was transformed into structured prompt, including problem assumptions, constraints, optimization objectives, etc. In the modeling solution phase. DeepSeek-Rl automatically generated the solution code based on the prompt to obtain an initial scheduling plan. During the iterative optimization phase, the user obtained the final solution that conformed to the problem assumptions and constraints through the check-feedback loop based on the initial scheduling scheme. The case study showed that the method could achieve hourly level (about 8 hours) production scheduling optimization solution for island assembly line, and the feasibility and advantages of DeepSeek-R1-enabled scheduling optimization were verified through multi-algorithm comparison.