刮板输送机断链智能监测技术研究  

Research on intelligent technology for broken chain monitoring on scraper conveyors

作　　者：李灵锋[1] 张洁[1] 陈茁 查天任 尹瑞 LI Lingfeng;ZHANG Jie;CHEN Zhuo;ZHA Tianren;YIN Rui(Mechanical and Electrical Engineering Department,Hebei Construction Material Vocational and Technical College,Qinhuangdao 066004,China;China Coal Zhangjiakou Coal Mining Machinery Co.,Ltd.,Zhangjiakou 076250,China;Hebei Province High-end Intelligent Mine Equipment Technology Innovation Center,Zhangjiakou 076250,China)

机构地区：[1]河北建材职业技术学院机电工程系,河北秦皇岛066004 [2]中煤张家口煤矿机械有限责任公司,河北张家口076250 [3]河北省高端智能矿山装备技术创新中心,河北张家口076250

出　　处：《工矿自动化》2025年第3期63-69,77,共8页Journal Of Mine Automation

基　　金：国家重点研发计划项目(2017YFF0210606);河北省高等学校科学研究项目(ZD2022018)。

摘　　要：针对现有基于AI算法的煤矿井下刮板输送机断链监测技术在线学习能力低、检测精度差、稳定性低、复杂场景适应性和可靠性差等问题,通过在极限学习机(ELM)中增加增量式在线训练,设计了可实现离线样本和实时在线样本训练的在线贯序极限学习机(OSELM)网络,进而提出了基于OSELM的刮板输送机断链智能监测技术。将经过大量煤矿井下刮板输送机链条监控图像(离线样本)训练的OSELM网络算法写入AI摄像仪,将AI摄像仪安装于刮板输送机机尾,实时感知刮板输送机链条运行状态并进行在线学习,由AI摄像仪输出控制决策,并通过刮板输送机集中控制系统平台实时显示识别结果。井下工业性试验结果表明,OSELM网络具有较高的自主学习能力、较强的泛化性和鲁棒性,对刮板输送机断链识别的平均精度均值、准确率和精确率分别为98.6%,99.3%,91.7%,检测速度达205.6帧/s,整体效果优于深度神经网络融合网络、RT-DETR、YOLOv5、YOLOv8、ELM等模型,实现了刮板输送机链条状态的精准、实时检测。To address the issues of existing AI algorithm-based broken chain monitoring technologies for underground coal mine scraper conveyors,including poor online learning ability,low detection accuracy,instability,and inadequate adaptability and reliability in complex scenarios,an online sequential extreme learning machine(OSELM)network was developed by integrating incremental online training into the extreme learning machine(ELM).This approach enabled both offline and real-time online learning.Based on this,an OSELMbased intelligent broken chain monitoring technology for scraper conveyors was proposed.The OSELM network algorithm,trained on a large dataset of underground scraper conveyor chain monitoring images(offline samples),was embedded into an AI camera.The AI camera was installed at the tail of the scraper conveyor to monitor the operation status of the chain in real-time while performing continuous online learning.The AI cameras output control decisions,with recognition results displayed in real-time on the centralized control system platform for the scraper conveyor.The results of industrial tests in underground mining environments demonstrated that the OSELM network exhibited strong self-learning ability,high generalization ability,and robustness.The mean average precision,accuracy,and precision for chain breakage identification on the scraper conveyor reached 98.6%,99.3%,and 91.7%,respectively,with a detection speed of 205.6 frames per second.The overall performance outperforms models such as Deep Neural Network Fusion Network,RT-DETR,YOLOv5,YOLOv8,and ELM,achieving precise and real-time detection of the chain status of scraper conveyors.

关 键 词：刮板输送机 链条状态识别 断链监测 AI摄像仪 在线贯序极限学习机网络 

分 类 号：TD67[矿业工程—矿山机电]