摘要
为解决当前带式输送机托辊故障检测主要依赖于人工巡检,标准不明确、效率较低的问题,采用数据统计分析、人工智能定位算法及图像匹配技术,研究带式输送机托辊故障检测。通过统计分析托辊异常情况产生的类型,发现托辊异常的主要形式为晃动导致与固定架之间发生角度偏差,进而影响作业安全。为此提出一种检测托辊位置偏转算法,采用基于YOLOv7模型的旋转目标检测,检测托辊位置及托辊角度。并使用采集的历史图像与实时采集图像匹配,对比判断托辊丢失状态及偏转角度,判断托辊视觉异常情况。结果表明:托辊检测精度为99.7%;在弱光拍摄条件下,托辊丢失故障检出率为99.8%,托辊角度变化大于5°的故障检出率为94.5%。
The fault detection of rollers of coal conveyor belts currently relies on manual inspection,which lacks clear standards and has low efficiency.Data statistical analysis,artificial intelligence localization algorithms,and image matching technology were used to investigate the fault detection of rollers of coal conveyor belts.Through statistical analysis of roller abnormalities,it was found that shaking was the primary form of roller abnormality,resulting in a certain angular deviation between the roller and the fixed frame and thus affecting operational safety.Therefore,a detection algorithm for roller position deviation was proposed.The rotary object detection based on the YOLOv7 model was used to detect the position and angle of the rollers.Historical images were collected and matched with captured images in real time to compare the roller loss status and deviation angles and identify visual anomalies of the rollers.The results show that the roller detection accuracy is 99.7%.Under low-light shooting conditions,the detection rate of roller loss faults is 99.8%,and the detection rate of faults where the roller angle changes by more than 5°is 94.5%.
作者
吉日格勒
柳尧
尚书宏
JIRI Gele;LIU Yao;SHANG Shuhong(National Energy Group Zhunneng Group Co.,Ltd.,Ordos Inner Mongolia 017100,China)
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2023年第S02期195-201,共7页
China Safety Science Journal
