Acoustic signal-based misfire detection in internal combustion engines using machine learning techniques

This research focuses on detecting misfire in a four-cylinder four-stroke gasoline engine inside an acoustic engine test cell using audio signal processing. This research proposes a smart solution by combining signal processing techniques and artificial neural networks. Misfire was created by fuel injection cut off for each cylinder at a constant speed of ۷۶۰ rpm, and the audio signals were recorded under controlled acoustic conditions. FFT, MFCC and STFT techniques were used for feature extraction. The results showed that the artificial neural network and the one-dimensional convolutional neural network with features extracted from the fast Fourier transform achieved accuracies of ۹۸.۴۰% and ۹۹.۳۶%, respectively. Also, the two-dimensional convolutional neural network using features extracted from the short-time Fourier transform achieved an accuracy of ۹۹.۷۱%. These results show that the proposed methods, especially the use of two-dimensional convolutional neural networks, have a very good performance in identifying the healthy and faulty state of the engine and can serve as an effective tool for real-time monitoring and fault diagnosis of gasoline engines.

کلیدواژه ها:

Engine Fault detection ، Misfiring ، Acoustic Signal ، Machine learning