基于改进YOLOv11n模型与DeepSeek-R1-7b的FPC缺陷智能检测系统OA

To address the problems such as difficulty of achieving high accuracy detection in flexible printed circuit board(FPC)defect under resource constraints,the lack of causal analysis and process diagnosis capabilities,and the low efficiency in generating FPC defect intelligent detection reports,an FPC defect intelligent detection system integrating an improved you only look once version 11 nano(YOLOv11n)model with DeepSeek release 1 7-billion parameter reasoning large language model(DeepSeek-R1-7b)was proposed.Firstly,the backbone,neck,and detection head of YOLOv11n model was reconstructed.Subsequently,the model was compressed using a pruning strategy.Finally,DeepSeek-R1-7b was optimized by retrieval-augmented generation(RAG)strategy for analytical reasoning.The results demonstrated that the mean average precision of improved YOLOv11n model achieved 2.98%higher than the original YOLOv11n model with a target pruning ratio of 2.0,while the parameter count and model size were 20.5%and 32.7%of the original model,respectively.Under lightweight deployment conditions,the system achieved high-precision defect detection and real-time generation of FPC defect intelligent detection reports containing diagnostic conclusions and process improvement measures consistent with industrial logic.