Research On Vehicle Fault Diagnosis Algorithm Based On Back-propagation Convolutional Neural Network

According to the www.autohome.com website data,31.1% of PM2.5 pollution sources in Beijing come from motor vehicles.Once the motor vehicle fails to operate,it will lead to increased emissions of pollutants,greatly affecting the effectiveness of national air pollution control.At present,the maintenance of motor vehicles mainly relies on the maintenance personnel to use the vehicle diagnostic equipment to read the data of vehicle sensor parameters and vehicle fault code,as the basis of fault judgment to repair vehicles.The personal technical ability of maintenance personnel directly affects the accurate judgment of vehicle failure.Causes the industry to the vehicle maintenance efficiency is not high,not in time and so on.Machine learning technology can learn model parameters,and it has a good application in classification and prediction problems in image and speech recognition.Therefore,based on Python's abundant underlying data control technology,this paper collects motor vehicle data through existing post-market diagnostic equipment,collates single vehicle data as training and validation data,and proposes a new approach based on the analysis and research of vehicle data and various artificial intelligence diagnostic technologies.A vehicle fault diagnosis algorithm based on the combination of neural network for image recognition and existing diagnostic knowledge is presented.This paper mainly did the following work.(1)Through the understanding and analysis of vehicle fault diagnosis knowledge,the existing maintenance equipment is used to collect diagnostic data.Acquisition of vehicle diagnostic data and fault codes during the use of diagnostic personnel.Understand the process and basic logic of vehicle maintenance personnel using diagnostic equipment to repair vehicles,and collect and save corresponding communication data according to the process of equipment use.(2)There are many kinds of data collected from the vehicle fault diagnosis equipment,After SPSS modeler analysis,select the model P₉49_V732 as the test object.By using back propagation(BP)neural network in machine learning,the optimal BP neural network model can be obtained by adjusting the number of hidden layer nodes,activation function,learning rate and normalization,and by comparing the adjusted training results.(3)The convolutional neural network in image recognition is modified and transplanted to the prediction of vehicle fault diagnosis to obtain the training model.Then the prediction results of the trained model on the vehicle diagnosis data verify that the original Le Net-5 model is not applicable to the vehicle fault diagnosis algorithm.At the same time,according to the optimization ideas of Alex Net and VGGNet,the optimized convolution network model is proposed and the effectiveness of the optimized convolution neural network model is verified.The accuracy is 96.7%.It can be concluded that the reconstructed convolutional neural network has better fault prediction ability in vehicle fault diagnosis algorithm.In practical application,it can help vehicle maintenance personnel improve the prediction of vehicle fault.