A Novel Intelligent Compound Fault Diagnosis Method for Piston Engine Valves Using Improved Deep Convolutional Neural Network

Abstract

In recent years, intelligent data-driven deep learning-based fault diagnosis methods play a vital role in the reliability and security for modern industrial machinery. However, in real industrial application, the issues of large amounts of data and environmental noises make it difficult to detect the compound fault for piston engine based on traditional intelligent fault diagnosis method. To break through the problem and further enhance efficacy, one-dimension hierarchical joint convolutional neural network (1-DHJCNN), an improved deep convolutional neural network is proposed to extract directly useful information for vibration signal and then give diagnosis results in this paper. The effectiveness of the proposed method is verified through the dataset of valve fault for piston engine, which is composed of single valve fault and compound valve fault with different fault degrees. Benefiting from this novel network structure, the experimental results show that the proposed method can not only outperform the traditional diagnosis methods such as integral deep convolutional neural network, but also present a small amount of model parameter calculation. In the end, feature visualization is adopted as a promising tool to analyse the mechanism behind the good diagnosis result of the proposed model.KeywordsCompound fault diagnosisPiston engineDeep convolutional neural network