Wavelet denoising using different mother wavelets for fault diagnosis of engine spark plug

Abstract

This paper deals with vibration-fault diagnosis of spark plug of an internal combustion engine using wavelet analysis and support vector machine. In order to reduce the noises of the vibration signals, wavelet denoising technique was used. A performance comparison was made between different mother wavelets as well as different levels of decomposition in order to find the best cases for the system under study. The results showed that the maximum classification accuracies were obtained by 13 different wavelets, namely, db1_4, db1_5, db2_4, db3_4, coif1_4, coif1_5, coif2_4, coif3_3, coif3_4, coif3_5, dmey_2, dmey_4 and bior3.7_6. It was also demonstrated that db1, coif1, coif3 and dmey were valuable mother wavelets for this study. Moreover, the results indicated that the proposed approach can reliably be used for spark plug fault diagnosis.

Structural damage detection and localization using a hybrid method and artificial intelligence techniques

“…Some of the wavelet families are Symlets, Daubechies, Fejer-Korovkin, Morlet, Mexican Hat and Coiflet wavelets. 20,49 The most proper choice of mother wavelet is case-dependent and in theory, there is no single best approach. 49 The selection of these mother wavelets is in line with Ovanesova and Sua´rez.…”

“…20,49 The most proper choice of mother wavelet is case-dependent and in theory, there is no single best approach. 49 The selection of these mother wavelets is in line with Ovanesova and Sua´rez. 30 The Shannon, Gaussian, Morlet and Mexican Hat wavelets are not the appropriate selection for DWT due to the explicit expression of wavelet function and lack of the scaling functions, causing the unavailability for discrete reconstruction.…”

Engine Working State Recognition Based on Optimized Variational Mode Decomposition and Expectation Maximization Algorithm

“…A number of signal processing algorithms have been researched to overcome this problem such as wavelet transform and empirical mode decomposition (EMD). Moosavian et al analyzed the performances of different mother wavelets and used wavelet transform to denoise vibration signals [7], Ma et al mixed wavelet transform and EMD to analyze vibration signals in order to diagnose abnormal combustion in engine [8], Li et al detected abnormal clearance between contacting components using EMD [9]. However, the wavelet transform used for signal decomposition is generally binary discrete wavelet transform (DWT), which just decomposes signals in Fourier spectrum mechanically.…”

Analysis of failure in concrete and reinforced-concrete beams for the smart aggregate–based monitoring system

“…30 However, the key factor in discrete wavelet analysis is an appropriate selection of mother wavelet function, as the coefficients are computed through the dot product of mother wavelet and original signal. 31 Several popular wavelet families are Symlets, Daubechies, Fejer-Korovkin, Morlet, Mexican Hat and Coiflet wavelets. 31,32 The most proper choice of mother wavelet is case dependent, and in theory, there is no single best approach.…”

“…31 Several popular wavelet families are Symlets, Daubechies, Fejer-Korovkin, Morlet, Mexican Hat and Coiflet wavelets. 31,32 The most proper choice of mother wavelet is case dependent, and in theory, there is no single best approach. 31 Therefore, the second goal of this study is to perform DWT-based signal de-noising and compare the performances between different mother wavelets to find the best one for potential noise reduction from the SA-based data.…”