Wavelet based de-noising in manufacturing and in business

Last two decades a powerful new tool has appeared in the quality assurance and quality control: the data mining. It enables discovering the hidden connections and confluences in big data systems, collected by on-line measuring and control systems in manufacturing processes. The main difficulties appear in distinguishing the valuable signal and the noises in measurements, as well as the time dependent, relatively slow changes indicating the inherent tendency of changes inside the process. It has been shown ([1], [2] etc.) that the Self Organizing Maps can be used effectively in analyzing the manufacturing process (e.g. hot rolling) in order of predicting the expected quality. Source of the analysis is the big amount of data collected by the hierarchically structured control system of the hot rolling mill. Since the sampling frequencies in different measurement devices were different, the problem of the possible noise became vital for the reliability of quality prediction. Recent paper intends to illustrate the efficiency of the wavelet analysis for noise removal in different areas of data mining, including the thermal processes in hot rolling, and in another area, very far from it in financial analysis. Technically, the wavelet is based on the usage of orthogonal functions. Unlike the similar and mathematically closest Fourier analysis - wavelets conserve the entirely time dependent information, where the temporal identification of the signals frequency content is also important. (C) 2013 The Authors. Published by Elsevier B.V.