Comparative study of noise reduction approaches with engineering applications in time domain and frequency domain

Abstract

Noise effects are unavoidable in engineering data acquisition systems. Noise reduction is necessary for numerous practical problems in either time domain or frequency domain, where measurements and observations are contaminated by diverse sources of noise. Advanced noise reduction algorithms should be applied to minimize the impact of data corruption in problem solving. Thus, a comparative study is made on a basis of several denoising methods, such as the adaptive recursive least square (RLS) filter, multi-layer neural network training via backpropagation algorithm and discrete wavelet denoising. Two case studies are conducted: the laser Doppler spectral filtering in frequency domain and rapid compression machine pressure profile signal denoising in time domain, in order to evaluate these typical noise removal approaches.