汽车电子电器接触不良的深度学习图像诊断方法

中图分类号：U469.72 文献标识码：A 文章编号：1003-8639（2026）03-0117-02

Deep Learning Image Diagnosis Method for Poor Contact of Automobile Electrical Appliances

Weng Wenjun (Shanghai General Motors Company Limited Wuhan Branch,Wuhan 43O2Oo, China)

【Abstract】 The article proposes a deep learning image diagnosis method based on the improved YOLOv11. By constructing an image diagnosis model forcontactfailure,the detectionabilityfortypical defects such ascontactburing, oxidative corrosionand mechanical loosening is enhanced.The experiments show that the model achieves an mAP @0.5 of （20 97.2% at a resolution of 640 1× 640, with an inference speed of 150 FPS,and the recall rate for detecting O.5mm-level microdefects reaches 97.8% .This method,through lightweight design,reduces the parameter quantity by 32% and is suitable for embedded devices,meeting the real-time detection requirements of industrial production lines.

【Key words】automotive electrical appliances;poor contact;deep learning; image diagnosis

0 引言

汽车电子电器系统的可靠性直接影响整车性能与行车安全，其中接触不良故障占比超过 60% ，典型表现为触点烧蚀、氧化腐蚀及机械松动等。（剩余3273字）