The Research On Fault Diagnosis Of Tread Extrusion Linkage Production Line Based On Convolutional Neural Network

Since the 1980 s,with the rapid development of artificial intelligence technology,especially the further application of knowledge engineering,expert system and artificial neural network in the field of diagnosis,people have been forced to study the intelligent fault diagnosis technology more deeply and systematically.This is because,on the one hand,intelligent fault diagnosis has incomparable advantages compared with traditional diagnosis methods;on the other hand,the fault diagnosis of complex equipment needs to rely on a large number of effective process history data to a large extent.Therefore,domestic and foreign experts and scholars have developed a large number of fault diagnosis systems based on historical data,and various diagnostic methods and techniques have also been applied in the diagnosis system.This paper is based on the research,design,installation and debugging of the project “Intelligent control system for compound extrusion process of tire tread”,the convolution neural network in machine learning is introduced into fault diagnosis.The convolution neural network is good at extracting effective features from high-dimensional complex data structure.A network model is built for fault diagnosis of tire tread extrusion production line,and its fault diagnosis is verified by comparing laboratory tests.There is a higher accuracy rate than traditional algorithms.The main research work is as follows:(1)The technological process,electrical control technical index and temperature index of tread extrusion linkage production line were summarized in turn,and the types,scope and causes of faults in the production line were analyzed.(2)The generation,development,classification,research status at home and abroad and main technical means of fault diagnosis were summarized.The data acquisition system and data denoising method of tread extrusion linkage production line were introduced.(3)The data classification technology is summarized,and the basic structure and working principle of convolution neural network are introduced.The general process of fault diagnosis for tire tread extrusion production line based on convolution neural network model is studied.The historical data set is formed by data acquisition system.Convolution operation is carried out between data samples using several convolution cores of multiple convolution layers to realize feature extraction.Local features of convolution layers are pooled through the pooled layer and recanalized.The probability of each category is obtained by Softmax normalization,and all local features are taken into account to realize fault diagnosis.The model is tested on the collected historical data sets,and good results are obtained.(4)The traditional machine learning algorithm BP neural network,decision tree and logistic regression are introduced in turn,and a comparative experiment is carried out on the same historical data set.The results of the experiment are compared with the convolution neural network model proposed in this paper.It is further proved that the algorithm model has certain advantages in fault diagnosis of tread extrusion linkage production line.After that,the advantages and disadvantages of the algorithm model based on convolution neural network are summarized and analyzed,and the future research directions are prospected.