A vibration-based approach for diesel engine fault diagnosis

Abstract

As diesel engines play a critical role in various applications, the ability for a health monitoring system to perform early fault diagnosis is of vital importance for the reliable functioning of a diesel engine throughout its service. In regard to the criticality of diesel engine fault diagnosis, an abundance of research efforts have been conducted by adopting injection-based, vibration-based, or instantaneous speed-based methods. The study presented in this paper discusses an integrated approach of diagnosing combustion faults and valve leakage, by combining vibration signal with cylinder pressure and revolution speed signals. Signal processing methods, including time frequency domain analysis such as wavelet decomposition are applied to extract features from data collected under different health conditions. The dimension of the feature set is further reduced based on discriminant analysis. Classification techniques are subsequently evaluated as fault diagnosis tools. The proposed approach is validated on a small-scale diesel engine test bed, and is to be applied in a real world health monitoring system.