supervised fault detection for engine misfires using acoustic signals and deep convolutional neural networks

This research focuses on detecting misfires in internal combustion engines within an acoustic engine test room using audio signal processing. this research proposes an intelligent solution by combining signal processing techniques and artificial neural networks. misfires were made on a gasoline engine at 760 rpm; acoustic signals were recorded under controlled acoustic conditions. fft, mfcc, and stft techniques were used for feature extraction. the findings revealed that a deep ann combined with feature extraction via fft achieved an accuracy of 99.32%, while a deep 1d cnn utilizing mfcc reached an accuracy of 99.69%, effectively identifying instances of incomplete combustion.