Application Of Wavelet Filtering In Abnormal Sound Diagnosis Of Motorcycle Engine

Engine fault diagnosis based on the analysis of abnormal sound signal is anon-contact fault diagnosis method，which has some advantages: easily operating, highaccuracy, high degree of automation etc. However, there will be noise in the acousticsignal acquisition and analysis process inevitably. As this background，this paper studiesthe application of wavelet filtering techniques in the diagnosis of the motorcycle engineknock. The specific works are summarized as follows:Introduction, in the beginning of the paper, clarifies the meaning of the topics,outlines the status of the engine knock diagnosis and common problems, sums up theapplication and development of the wavelet filtering technology in the field of faultdiagnosis, and the advantage of wavelet filtering techniques in dealing with the transientpulse signal.Starting from the experimental design, chooses the measurement point and testconditions, which can best reveal the motorcycle engine operatingconditions.Introduces the basic mechanism of Fourier transform, and point out thedeficiencies in the handling of transient and singular signal. Time-frequency analysismethods was introduced, and introduced windowed Fourier transform, continuouswavelet transform, discrete wavelet transform, wavelet packet transform and the basicprinciples of wavelet packet decomposition by turn, analyze and compare theadvantages of the transformation methods, sums up the fundamental differencesbetween the Fourier transform with the wavelet transform, and illustrates the advantagesof wavelet transform, provides the theoretical basis for the wavelet filter and waveletpacket filtering algorithm below.Points out the limitations of the conventional FFT filtering noise reduction method,introduce time-frequency filtering method, compares advantages and disadvantages ofthe conventional filtering method based on Fourier transform and the filtering methodbased on wavelet transform, and describes the current situation and development ofwavelet time-frequency analysis methods; introduces the definition of the wavelet filter,model and filtering noise criteria. The paper uses conventional filtering method andwavelet filtering method to analyze the simulated signal, the signal to noise ratio（SNR）,mean square error（MSE） and waveform similarity（NCC） three mathematicalparameters were used to evaluate the filter effect, verifies the advantages of the wavelet filtering. Finally, discusses the key issues in the wavelet filtering algorithm, significantcoefficient, the selection of wavelet basis and decomposition level, these parameterswill affect the wavelet filtering effect, need to be focused and discussed.Take110cc engine to research in this subject as an example, for the abnormalsound type gearbox and the right cover different sound signals, used wavelet filteringand wavelet packet to analysis, and clear features could be extracted. Combined withwavelet filtering algorithm, a subjective auscultation method was proposed-waveletinteractive filtering playback, wavelet filtering algorithm and the experience of expertsare blended in with in a certain extent, the playback information of the transient pulse isa breakthrough in a sense.