Research On Feature Extraction Of Diesel Engine Noise Based On Acoustic Array And Wavelet Transform

Acoustic condition monitoring of machinery has received considerable attention in recent years.The main advantage of using sound signal in diagnosis is that the acoustic measurements can be performed at a distance from the machine,and sound signal is sensitive to initial faults.And the wavelet transform of time-frequency analysis can analyze the characteristics of non-stationary signals.The feature extraction of the valve-train system of diesel engine based on the acoustic array and wavelet transform is performed in this dissertation.According to the array layout research,the appropriate acoustic array spacing、number and measurement distance are obtained and used it to collect signal.Secondly,the array data is processed by the fixed beamforming algorithm and the quality improvement effect of enhanced acoustic signal is evaluated by signal enhancement index.It is aim to improve the signal-to-noise ratio of the acoustic signal.The experimental result of measured signal show that the proposed method can improve the signal-to-noise ratio of the acoustic signal.Then,the basic theoretical knowledge of wavelet transform is also introduced in detail and its advantage in signal decomposition is analyzed in this dissertation.After that,the method of wavelet basis function selection based on principal component analysis is proposed for the subjectivity of the wavelet basis function selection method.Finally,combining with the traditional wavelet basis selection method,the analysis software for wavelet basis function selection is obtained.Finally,according to the characteristics of signal non-stationary,wavelet transform is introduced into the feature extraction of signal.The method of the principal component analysis is used to the feature extraction of continuous wavelet transform.The characteristic parameters of the diesel engine signal are obtained and the high-dimensional matrix information is decomposed to the principal component which contains a lot of signal information.The signal frequency components of the small diesel engine are analyzed again by the principal component,and its change of the characteristic frequency is obtained after the change of the valve clearance.After that,the suitable level of wavelet packet decomposition is determined and the difference of signal frequency band energy in different states is obtained by using the commonly wavelet packet frequency band energy method and the previous spectrum analysis.Finally,the experimental results show that the feature extraction method based on the combination of acoustic array signal and wavelet transform are effective for the mechanical diagnosis.