A Multi-Agent Framework for Multi-Source Manufacturing Knowledge Integration and Question Answering

The integration of next-generation information technologies in the manufacturing sector has resulted in the generation of extensive and complex knowledge data by organizations. Effectively managing and utilizing this dispersed knowledge presents a significant challenge. This paper proposes a Multi-Agent-based, Multi-Source, and Multimodal Retrieval-Augmented Generation (RAG) framework specifically designed for the manufacturing industry. By employing advanced methodologies such as multimodal document parsing, multi-source data integration, and intelligent agent-driven querying, the proposed framework aims to address the interconnected challenges of knowledge management and intelligent question answering in complex manufacturing environments. The principal contributions of this research include: (1) the advancement of sophisticated multimodal document parsing techniques, (2) the establishment of robust strategies for multi-source data integration, (3) the development of a comprehensive multi-agent intelligent question-answering system, and (4) the enhancement of the Thinking RAG model. This framework provides a holistic solution to the complexities associated with knowledge management and improves retrieval efficiency in the manufacturing sector.