No. 1134
Wavelet and Knowledge Discovery Technology
Joint Research Committee on Wavelet and Knowledge Discovery Technology
Keyword : wavelet, knowledge discovery, data mining, equipment diagnosis
For a complex or large scale system to be controlled, there may exist a variety of beneficial knowledge in a large amount of data acquired and stored in the system. For example, knowledge about equipment condition, operation, control, maintenance, and process improvement. Unfortunately, such information is buried under the massive data and difficult to obtain. Therefore, comprehensive data processing is desired for knowledge discovery and translation into a human understandable expression. In general, such a process consists of five steps: 1) data selection, 2) pre-processing, 3) data transformation, 4) data mining, 5) interpretation, understanding, and evaluation. Wavelet analysis is expected to be widely applicable to each process with various ways because it is one of the alternative technologies to human visual and auditory processing, and can present variation of frequency properties as time passes. This technical report is organized systematically from the viewpoint of "wavelet analysis for knowledge discovery". It has various contents from fundamental theories to actual applications: industrial applications utilizing time-frequency analysis by wavelet (for example, equipment diagnosis), case studies that bring new knowledge and ideas, and RI-Spline wavelet, which solves essential problems when applying discrete wavelet transform (DWT) to industrial area, and its application. The report also presents dead time estimation and blind signal separation as application examples combining wavelet transform with other methods.

©2007. The Institute of Electrical Engineers of Japan