首页|期刊导航|模具技术|冲压件流水线汽车覆盖件的缺陷实时检测系统设计

冲压件流水线汽车覆盖件的缺陷实时检测系统设计OA

Design of real time defect detection system for stamping assembly line automotive covering parts

中文摘要英文摘要

针对冲压件流水线场景中传统汽车覆盖件缺陷检测存在钢带表面缺陷检测精度和效率低的问题,提出一种基于改进 YOLOv8 的缺陷检测模型.首先,以YOLOv8 算法为基础目标检测算法;然后在其基础上添加 FasterNet 网络架构、引入HS-GAM 注意力机制和 WIoUv3 损失函数;之后将改进的 YOLOv8 模型部署至设计的汽车覆盖件缺陷实时检测系统中,以提高汽车覆盖件钢带表面缺陷检测精确度和实时性.实验结果表明,提出的改进 YOLOv8 模型对划痕、凹坑、麻点的检测精度分别为 0.977、0.994 和 0.981,mAP@0.5 指标为 0.989,每秒帧数(FPS)取值为 64.56,相较于主流模型 Faster-R-CNN、SSD、YOLOv7 和 CNN-ViT 模型,本模型的检测精度和检测速度更高.由此分析证明,本模型可实现汽车覆盖件缺陷准确检测,检测精度和效率显著提升.经实际部署可知,设计的系统可实现汽车覆盖件精准、快速检测和检测结果可视化分析,满足实际应用需求,具备时效性和可靠性.

To address the limitations of traditional defect detection methods for automotive body panels in stamping production lines—particularly low accuracy and efficiency in detecting surface defects on steel strips,we propose an enhanced YOLOv8-based defect detection model.The approach first employs the YOLOv8 algorithm as the core object detection framework,then integrates the FasterNet neural architecture,introduces the HS-GAM attention mechanism,and utilizes the WIoUv3 loss function.This improved YOLOv8 model was deployed into a real-time inspection system for automotive body panels to enhance both detection precision and processing speed.Experimental results demonstrate that the proposed model achieves detection accuracies of 0.977 for scratches,0.994 for dents,and 0.981 for speckles;a mean absolute precision(mAP@0.5)of 0.989,and a frame per second(FPS)rate of 64.56.Compared to mainstream models such as Faster-R-CNN,SSD,YOLOv7,and CNN-ViT,this model exhibits superior performance in both accuracy and speed.Our analysis confirms its capability for precise defect detection with significant improvements in efficiency and reliability.Practical implementation shows the system enables accurate,rapid defect identification and visualized analysis of inspection results,fully meeting industrial application requirements..

杜婉;柴彩彩

西安汽车职业大学,陕西 西安 710600西安汽车职业大学,陕西 西安 710600

信息技术与安全科学

汽车覆盖件缺陷检测改进YOLOv8FasterNet网络注意力机制

automotive coveringsdefect detectionimprove yolov8fasternet networkattention mechanism

《模具技术》 2026 (4)

61-70,10

2024年度陕西省教育厅一般专项科学研究计划项目"城市环境下智能车辆自主换道决策机制研究"(编号:24JK0586).

评论