摘要
列车配电柜是其电力系统核心,负责电能外部引入、分配、控制与保护,其安全性直接影响列车运行安全。如何快速准确检测其缺陷已成为保障行车安全的关键工作之一,但目前配电柜巡检存在依赖人工检测、主观因素大、缺乏有效智能检测手段等问题。针对以上问题,提出了一种基于MFFM-YOLO和MobileNet轻量化的高速列车配电柜缺陷智能识别方法。首先,提出基于MobileNetv3的轻量化模型设计方法,构建基于YOLOv8-Seg模型的高速列车配电柜缺陷识别模型,减少模型参数,提升缺陷识别速度;其次,设计多级特征融合模块(Multi Feature Fusion Module,MFFM),保留空间维度同时有效集成特征图,提升模型准确性和鲁棒性。最后,开展缺陷识别验证实验,所提方法实现了强噪声下器件松脱、安装错误、接线异常等配电柜缺陷识别,准确率96.38%,时间缩短46.54%,能有效缓解列车配电系统周期巡检和运维压力,提升其智能化巡检水平,强化列车电力保供支撑能力。
The train distribution cabinet is the core of its power system,responsible for the external introduction,distribution,control,and protection of electrical energy.Its safety directly affects the safety of train operation.How to quickly and accurately detect its defects has become one of the key tasks to ensure driving safety,but currently there are problems in the inspection of distribution cabinets,such as relying on manual detection,subjective factors,and lack of effective intelligent detection methods.A lightweight intelligent defect recognition method for high-speed train distribution cabinets based on MFFM-YOLO and MobileNet is proposed to address the above issues.Firstly,a lightweight model design method based on MobileNetv3 is proposed to construct a defect recognition model for high-speed train distribution cabinets based on the YOLOv8-Seg model,reducing model parameters and improving defect recognition speed;Secondly,design a Multi-Feature Fusion Module(MFFM)to preserve spatial dimensions while effectively integrating feature maps,improving model accuracy and robustness.Finally,a defect recognition verification experiment was conducted,and the proposed method achieved the identification of distribution cabinet defects such as device loosening,installation errors,and abnormal wiring under strong noise,with an accuracy rate of 96.38%and a time reduction of 46.54%.It can effectively alleviate the pressure of periodic inspection and operation of the train distribution system,improve its intelligent inspection level,and strengthen the support capability of train power supply.
作者
盛利
杨慕晨
张喆清
郭伟超
赵晨阳
张法业
SHENG Li;YANG Muchen;ZHANG Zheqing;GUO Weichao;ZHAO Chenyang;ZHANG Faye(CRRC Qingdao Sifang Locomotive and Rolling Stock Co.,Ltd.,Qingdao,Shandong 266111,China;School of Control Science and Engineering,Shandong University,Jinan,Shandong 250061,China)
出处
《自动化与仪器仪表》
2025年第10期39-44,共6页
Automation & Instrumentation
