||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.