Proceeding of the 10th International Congress on Engineering Applications of Neural Networks
Volume, Number, Page
pp. 210-220
Published date
Aug. 2007
Publisher
Japanese:
English:
Publishing Centre Alexander T.E.I. of Thessaloniki
Conference name
Japanese:
English:
Engineering Applications of Neural Networks (EANN 2007)
Conference site
Japanese:
English:
Thessaloniki, Greece
Abstract
The batch Self Organizing Map, abbreviated to SOM, is adapted to process imaging for dynamic behavior of aerated agitation vessel, and the application methods are investigated in this article. In the application, the direct imaging by CCD video camera and the PIV technology are adopted. As a result of mapping time-series patterns of velocity distributions to two dimensions, it is shown that generated map and clusters could give process engineers useful information about degree of spatial dispersion of bubbles and about determination of design parameter. The next, two approaches for the data processing in the SOM are investigated to enhance efficiency of pattern analysis: phenomenological approach and statistical approach. As to the statistical approach, it is found that adoption of sigmoid transformation enhances the efficiency of separating the minor difference in the nodes representing the “well-dispersion” patterns and that it would give process engineers useful information about transition of process patterns.