Research on Automobile Fault Diagnosis Model Based on Neural Network

With the continuous development and popularization of automotive technology, modern automobiles are equipped with a large number of electronic and mechatronic systems, which makes the diagnosis of automotive faults more complex. The traditional manual detection methods can no longer meet the increasing demand for automobile maintenance. In this study, a neural network-based fault diagnosis algorithm model for automobile external deformation is proposed. In this paper, a dataset containing 8000 images of faulty family minivans and 4000 images of normal cars is constructed. In order to improve the image processing effect, this paper adopts bilateral filtering technique for image denoising to improve the image quality and make it more suitable for subsequent processing. In this paper, ordinary convolutional neural network and DenseNet network models are established respectively. And through model validation, it is found that when DenseNet-264 network is chosen, the model has the lowest error rate and higher accuracy, and the prediction error rate for faulty cars is only 20.6%.