Author(s)

Kai Chen, Yuxiang Wang

Corresponding Author

Kai Chen

ABSTRACT

For the most critical feature extraction problem in vehicle engine abnormal sound fault diagnosis, this paper selects a certain type of transport vehicle as the research object, and collects the audio signal of the engine in normal operation and lack of cylinder operation as a sample. Firstly, the original signal is preprocessed, including pre-emphasis, Sub-frame, plus windows. Then the short-time energy of the frame signal is obtained in the time domain. Finally, the power spectrum is calculated in the frequency domain, which is passed through the Mel filter bank, and the MFCC is obtained by DCT. The research results show that the time and frequency domain characteristic parameters obtained by this method can reflect the operating state information of the engine, and can effectively distinguish whether there is a lack of cylinder operation. It is suitable as the feature vector for vehicle engine fault diagnosis and prediction. This will lay a foundation for future model training and matching recognition.

KEYWORDS

Engine, Acoustic signal characteristics, extraction methods