Application Of Wavelet Analysis In Abnormal Sound Diagnosis Of Motorcycle Engines

With the rapid development of the motorcycle industry in our country and improved the requirement of quality to products by customers, the problem of the motorcycle quality that caused by the abnormal sound in the engine have been becoming a vital issue, which most of companies are nagged by it. In order to resolve this question, most of firms have been requiring detecting the engine that could product abnormal sound. However, the way of detection is still using the subjective assessment, and there is a strong subjectivity in this method, which is hard to measure the quantitative value. Therefore, an objective way to detect the abnormal sound in the engine is provided in this thesis.The paper centers on abnormal sound detections, based on reading lots of related materials and comparing different methods that analysis the fault signal form the engine, the wavelet analysis is applied, which have a good effect. Thus, this method is researched and gets an ideal effect in analyzing and detecting the abnormal sound.Firstly, the related approaches of analyzing the abnormal sound engine are introduced, such as the Fourier transforms, short time Fourier transforms and order analysis. And, according to the features of the engine signals, there are some deficiencies in proposed ways. So, the time-frequency analysis is a suitable tool to analysis the engine signal that contains non-stationary and instantaneous components. Then, comparing with most of time-frequency analysis, the wavelet transform is the best way to analysis the abnormal sound signal in the engine.Secondly, although the wavelet analysis is considered as an important breakthrough in the signal processing, the selection of mother wavelet restricts development of the wavelet analysis. When selecting different mother wavelet, the result of the signal analysis varies wildly. So founding a method of the wavelet energy spectrum meets this problem. Form 15 types of mother wavelet functions in the Matlab software, the Complex Morlet wavelet is the best wavelet for analyzing the fault features. Meanwhile, there are two parameters in this wavelet, which are the bandwidth parameter-Fb and the wavelet center frequency parameter-Fc. Thus, for the purpose of choose suitable parameters of Fb and Fc, The maximum criterion of the Kurtosis coefficient and the minimum criterion of the wavelet Shannon entropy are operated. In the end, using the real engine signal of abnormal validates this selected mother wavelet, which can prove this way correctly and effectively.Finally, for the issues of abnormal sounds in the 110cc engines, several of engine sound signals, such as oil pump faults, balancer shaft faults and normal engines, are transformed into the wavelet energy spectrums. By it, the transient pulses can be found in the wavelet energy spectrums, which can demonstrated more convincingly than any other ways. It can distinguish the fault features from normal features; at the same time, the wavelet coefficients are extracted and considered as feature parameters, which could support the diagnosis tool of abnormal sound engines. Hence, combining C5.0 algorithm and extracted wavelet features, the model of abnormal sound diagnosis are built, and are verified by diverse of sound signals of engines.