The invention discloses a
machining part process
planning method and
system based on a
process knowledge base and an Agent-RAG architecture, and belongs to the technical field of intelligent manufacturing and precision
machining. Aiming at the problems of insufficient intelligence, poor
interpretability, low
utilization rate of multi-source heterogeneous process data and the like of process planning in small-batch, multi-variety and customized production of
machining parts, the invention proposes that a
process knowledge graph is automatically constructed based on a large
language model, and interpretable process planning is realized in combination with Agent-RAG. The method comprises the following steps: automatically
processing a
workflow based on process data of a
large model, and completing OCR (
Optical Character Recognition), cleaning, segmentation and automatic labeling of original process data; the method comprises the following steps: constructing a machining part process planning domain ontology DOPP, realizing end-to-end
process knowledge extraction by adopting a LoRA fine-tuning
large model, and forming a mixed
knowledge base of a
graph database and a vector
database; and designing query
decomposition and planning, self-adaptive retrieval routing and scheme generation three-Agent collaborative Agent-RAG strategies, and realizing
decomposition, accurate retrieval and interpretable scheme generation of complex
process requirements. According to the method, the labor cost is greatly reduced, the
knowledge extraction and process planning efficiency, accuracy and
interpretability are improved, and the method is suitable for machining part intelligent
process design and agile manufacturing scenes.