Fault Diagnosis Of Diesel Engine Based On Empirical Mode Decomposition And Wavelet Threshold Denoising

Diesel engine is the most widely used heat engine in industry and agriculture,transportation and even the national defense with a high thermal efficiency. Once a part of thediesel engine malfunctions, it will directly or indirectly result in significant economic loss oreven threat to life. Therefore, the research on fault diagnosis of diesel engine has been one ofproblem people always concentrate on. This thesis will focus on the fault diagnosis of dieselengine based on empirical mode decomposition and wavelet threshold denoising.After the in-depth study and research of the type and mechanism of diesel fault as well asthe principle and construct of diesel engine, this thesis determined the collection method ofvibration data and gathered the diesel engine vibration signals under different fault conditionswith experimental apparatus. On the basis of deep understanding of the characteristics ofvibration signals, two kinds of wavelet and Fourier on waveform analysis are contrasted inthis paper. The empirical mode decomposition method, combined with wavelet threshold, isresearched. It also made a comparison on the effect of several types of wavelet denoising,providing the result of empirical mode decomposition and wavelet threshold process onwaveform, as well as the optional contrast in different decomposition levels. The purpose isfully excavating the hidden characteristics from the original signals. Meanwhile, it optimizesthe extracted eigenvalue through variable precision rough set to realize the efficiency of dieselengine diagnosis.These eigenvalues are extracted by the relative energy method. After, studyon the application of BP and RBF network in diesel engine fault identification, an optionalplan is presented as following:First of all, basing on the relative theory of empirical mode decomposition and waveletthreshold denoising, extract informative IMF component of diesel engine vibration signal inthe following five conditions:1800rpm, measure point14, normal, advance angle increasingin off oil, oil spill and oil supply, and air filter clogging; collect the relative eigenvalue afterthe decomposition. Then, the variable precision rough set is applied to solve theincompleteness, redundancy and contradiction of characteristic parameters to optimize the parameters, accomplishing the effective diesel engine fault diagnosis with less property valuein case of no information missing. Finally, obtain the superiority of RBF network in faultdiagnosis.