摘要
为保证输煤线路平稳运行,实现托辊智能化监控和故障预警,采用智能巡检机器人搭载红外相机进行故障巡检,巡检机器人采用定点拍摄方式,实时采集红外图像,通过目标识别算法,判断托辊特征、定位托辊位置;采集托辊运行的测温信息,通过AI目标检测算法,判断托辊轴承高温异常情况,完成托辊温度的异常检测。利用托辊检测结果,结合提取的温度值及实际里程值,对运行人员发出告警信号,满足故障预警要求。在托辊异常检测中,借鉴Transformer网络的思想,引入递归门控卷积(g^(n)Conv),改进YOLOv7算法。结果表明:改进后平均精度为0.98,满足实时处理要求;改进的YOLOv7算法准确率提升3.1个百分点,召回率提升0.4个百分点,平均精度提升0.03个百分点,改进YOLOv7算法具有更优检测效果。
In order to ensure the steady operation of the coal transportation line and realize the intelligent monitoring and fault early warning of the roller,the intelligent inspection robot equipped with an infrared camera was used to perform fault inspection.The inspection robot adopted fixed-point shooting to collect infrared images in real time and used the target recognition algorithm to determine the characteristics of the roller and locate the roller.The inspection robot collected the temperature measurement information during roller operation,and the AI target detection algorithm was used to judge the abnormally high temperature of the roller bearing and detect abnormal roller temperature.Based on the detection results of the roller,the extracted temperature value,and the actual mileage value,an alarm signal was issued to the operator to achieve fault early warning.The detection of abnormal roller drew on the idea of the Transformer network and introduced recursive gated convolution(g^(n)Conv)to improve the YOLOv7 algorithm.The results show that the average precision of the improved YOLOv7 algorithm is up to 0.98 and meets the requirements of real-time processing.The accuracy of the improved YOLOv7 algorithm is increased by 3.1%;the recall rate is increased by 0.4%;the average precision is increased by 0.03%,and the improved YOLOv7 algorithm has a better detection effect.
作者
张磊
吉日格勒
尚书宏
ZHANG Lei;JIRI Gele;SHANG Shuhong(Chnenergy Zhunneng Group Co.,Ltd.,Ordos Inner Mongolia 010300,China)
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2023年第S02期222-227,共6页
China Safety Science Journal
