Blind Separation Of Surface Vibration Signal From Cylinder Head

Lots of information can be read from vibration signals of Internal Combustion Engine, so vibration signal can be used to estimate combustion process and diagnose fault, and this is of great significance to the development of working process detection and diagnostic technique of Internal Combustion Engine. The vibration signal also contains a lot of noise sources information, and it’s difficult to detect and diagnose working process of ICE based on measured vibration signal. Therefore, Time-Frequency Analysis and Separation of vibration signals is important to improve the availability of the vibration signal.The time-frequency analysis methods were compared in this paper, in order to obtain a suitable vibration signals processing method, and the time-frequency distribution of vibration signals were analyzed based on S-transform, then vibration response of different excitation signal were separated based on EEMD and EMD-ICA. At last, EEMD was used to extract fault information in order to verify the validity of the method. The main works of this paper are described in the following sections.Experiments are conducted on the195diesel engine. The cylinder pressure signal, cylinder head surface vibration signal, synchronization signal and crank angle signal are measured. The characteristics of the measured surface vibration signal are analyzed on the time domain and frequency domain.The basic principles and algorithms of five methods of time-frequency analysis are studied, such as Short-time Fourier transform, Wigner-Ville distribution, Hilbert-Huang transform, Wavelet transform and S-transform, and their merits and defects of the methods were compared. The conclusion is that S-transform is best suited for time-frequency analysis of vibration signal. Then, the S-transform method is adopted to analyze the cylinder head surface vibration signal.The EEMD and EMD-ICA methods are adopted to separate various excitation source response signals. Vibration signals were separated effectively by EEMD and EMD-ICA before the on the peak pressure, and signal to noise ratio (SNR) of the vibration signal excited by the in-cylinder pressure is significantly improved.The acceleration signal processed by EEMD is used for fault identification. This paper sets the abnormal lash and injection advanced angle faults on the test bench, and vibration acceleration signal was measured under two fault conditions. The vibration acceleration signal containing fault information was extracted by EEMD, and precision of the faults recognition was therefore improved.