Research On Condition Monitoring System Of Mine Conveyor Belt State On Vision

As an indispensable part of coal mine belt conveyor system,mining conveyor belt plays a vital role in coal production.With the continuous increase in the development of coal energy and the continuous improvement of production efficiency,the requirements of the mining conveyor belt in terms of load,speed and transmission distance continue to increase.Ensuring the safe and stable operation of the mining conveyor belt is important for ensuring the production capacity and safety of the mine.Production is very important,so it is of great significance to strengthen the safety monitoring and fault warning of mining conveyor belts.In view of the current design deficiencies and functional defects of the current mining conveyor belt condition monitoring system,in-depth analysis and experience are summarized.This paper analyzes and studies the two most common faults of mining conveyor belts,and proposes related detection methods in combination with today's popular visual inspection technology,and designs a vision-based mine conveyor belt state monitoring system.Firstly,the mechanism of longitudinal tearing failure and deviation failure of mining conveyor belt is studied and analyzed,and then the overall scheme design of condition monitoring system for mining conveyor belt is designed according to the site conditions,including system design requirements,overview of overall functions,and system software and hardware.Partial selection.For the longitudinal tear detection of mining conveyor belts: First,a longitudinal tear detection method based on image processing is studied,and algorithms are optimized and improved in the steps of image enhancement,edge detection,and feature extraction.In order to improve the detection accuracy of the surface cracks of the conveyor belt,a longitudinal tear detection method based on deep learning is studied.Through the study of convolutional neural network theory,the YOLOv3 model is selected to realize the conveyor belt longitudinal tear detection.Finally,the effectiveness of the method is verified by testing the test set images after learning and training.In addition,in order to improve the accuracy of fault detection,a longitudinal tear detection method based on auxiliary line laser is studied.After image preprocessing,the line laser stripe extraction algorithm and corner detection algorithm are optimized and improved through analysis and verification.For mining conveyor belt deviation detection: Three vision-based conveyor belt deviation detection methods are studied.In the deviation detection method based on the straight line feature,the LSD algorithm is selected to detect and screen the edge of the conveyor belt,and the deviation fault is judged according to the target straight line feature.In the deviation detection method based on area comparison,the image ROI is determined according to the running characteristics of the conveyor belt,and the ROI extraction algorithm and the area filling algorithm are optimized and improved.Finally,the detection method is verified by analyzing the conveyor belt running video.In the deviation detection method based on semantic segmentation network,the PSPNet model is selected to achieve the target segmentation of the conveyor belt image through the study of the theory of full convolutional neural network,and the segmented feature image is obtained through deep learning training.Finally,the conveyor belt images with different operating state characteristics are selected to verify the detection method.Finally,the software platform of the conveyor belt condition monitoring system is designed through the Visual Studio 2019 integrated development environment and MySQL database.Under the existing conditions of the laboratory,a conveyor belt condition monitoring experiment platform was designed and built,and the effectiveness of the monitoring system was verified through the experiment platform.The thesis has 97 figures,13 tables and 82 references.