Fault Diagnosis and Isolation for Diesel Engine Combustion Chambers Based on Autoencoder and BP Neural Network

In order to improve the efficiency and accuracy of diesel engine combustion chamber fault isolation, a method of combining the feature dimension reduction of AutoEncoder network and the fault isolation of BP neural network was proposed based on acoustic emission signals. Taking a Z6170 diesel engine of China ZICHAI company as an example, fault simulation tests of exhaust valve and piston rings under experimental environments were carried out, and the acoustic emission signals of the cylinder head were collected, then the time-domain, frequency-domain and other characteristic parameters of different signal sections in the whole cycle were extracted. The dimension of characteristic parameters was reduced by using AutoEncoder network, BP neural network was used to fault diagnosis and fault isolation, so that a fault diagnosis and fault isolation model of combustion chamber components was established. After training and verification of the model, it shows that the proposed diagnosis and isolation method is effective with capability of identifying the faults of exhaust valve and piston ring for the combustion chamber parts of diesel engines, therefore, it is promising to detect and isolate the condition of combustion components automatically.