Rapid generation method of process routes based on multi-agent collaboration with LLMs

In the process of process design for manufacturing, issues such as high reliance on personal experience and knowledge, along with long design cycles, are common. This paper proposes a rapid method for generating machining process routes based on Large Language Models (LLMs) and multi-agent collaboration. The complex task of generating process routes is broken down into four subtasks: machining feature recognition, machining feature sorting, machining feature process chain and resource selection, and process route merging and optimization. Each subtask is assigned to an agent fine-tuned with LLMs, equipped with different specialized tools such as STP file parsing and process knowledge base querying, to endow each agent with distinct expertise. The agents collaborate by exchanging information to achieve the rapid, automated generation of machining process routes, offering heuristic ideas for process designers. The TOPSIS evaluation method integrating quantitative and qualitative indicators based on actual production data and expert scores is used to compare the final generated processing route with typical ones, showing that it achieves a higher closeness degree. This demonstrates the advantages of multi-agent collaboration in complex tasks, providing a new solution for the automation and intelligence of process design in intelligent manufacturing systems.