Research On Fault Diagnosis System Of Mechanical Belt Buckle Based On Vision

PVC conveyor belt is widely used in underground coal transmission work,PVC conveyor belt used in belt conveyor must have a buckle to connect it into a ring for normal use.Once the belt buckle breaks during the operation of the conveyor belt,it is easy to cause safety production liability accidents.So it is necessary to check and maintain the mechanical buckle of PVC conveyor belt on a regulai basis.But for many years,it is mostly through the manual detection of underground workers.This detection method is not only inaccurate but also time-consuming and laborious.In addition,the high-speed operation of conveyor belt and the impact of underground environmental factors bring great difficulties to the detection of underground workers.Therefore,it is urgent to realize the visual inspection of the belt buckle of the conveyor,realize the image processing of the mechanical belt buckle,and timely find out the failure of the U-shaped nail pulling out,the U-shaped part bending,breaking,and realize the failure early warning to avoid the accidents of the belt buckle breaking.Therefore,it is very important to design a visual inspection system to replace the human inspection.This paper studies the fault detection system of belt buckle joint of conveyor belt based on machine vision,puts forward a design scheme of image acquisition of belt buckle joint,designs a framework based on LabVIEW,constructs a platform for image acquisition and image processing of mechanical belt buckle,and realizes the recognition and classification of belt buckle joint fault by convolution neural network.The system takes the mechanical belt buckle on PVC conveyor belt as the research object,collects the image of the mechanical belt buckle on the running conveyor belt by CCD linear camera,realizes the image preprocessing and image segmentation of the mechanical belt buckle image by the specific image processing algorithm,and then realizes the image capture of the belt buckle joint area by the image correction and the belt buckle joint calibration.In this paper,the traditional network model is optimized and improved,dropout method is introduced to solve the over fitting problem in training and learning,adagrad algorithm is used to optimize the network training parameters,and relu activation function is used to solve the gradient dispersion problem in recognition training,The method of Parzen+CNN is used to identify the fault of buckle joint,Experiments show that Parzen classifier can realize recognition and classification effectively.Compared with the traditional Softmax classification method,the recognition accuracy of Parzen classifier on MNIST is improved by 2.40%and the recognition efficiency is improved by 20.07%.The recognition accuracy on Buckle increases by 1.14%and the recognition efficiency increases by 33.22.%,showing that the recognition algorithm has strong generalization.machine vision based PVC belt mechanical buckle fault diagnosis system is finally constructed.The system runs fast and can get accurate classification results.The system can realize self diagnosis and fault early warning in the process of high-speed operation of conveyor.Once the hidden trouble of buckle fault is found,it can automatically detect the type of fault,automatically generate fault report,reduce the time of equipment troubleshooting,improve the efficiency of fault handling,and avoid the occurrence of major hidden trouble.