Vibration Signatures, Wavelets and Principal Components Analysis in Diesel Engine Diagnostics

The vibration signatures of a normally aspirated diesel engine contain valuable information on the health of the combustion chamber components. It could be used to detect incipient faults in the engine. Several commonly occurring faults were induced in a 4-stroke diesel engine and the ensuing vibration signals recorded. Three diierent feature extraction techniques: Domain expertise, wavelet analysis and principal components analysis (PCA), were compared to evaluate the eeectiveness of each method. The extracted features were then used to develop artiicial neural nets based fault diagnostic systems. It was found that the best results were obtained with the wavelets based feature extraction technique. However, each of the three systems was shown to exhibit a high degree of accuracy. In a separate study, ensembles of 3 nets were created. The majority of ensembles performed better than any of the constituent nets, thus demonstrating the power of the technique of combining neural nets in majority voting systems.