Recently, I (with Prof. Bruns) have been interested in exploring how nonlinear time series analysis techniques can be applied to predict flooding in distillation columns. Some preliminary results look promising enough to warrant further data collection followed by the application of multivariate information-theoretic tools with some pattern recognition techniques or specifically developed statistical tests to anticipate flooding. See paper 465d in the 2003 AIChE Annual meeting for details on what we have found so far.

Another area of interest is to use multiple features for decision making processes. In an independent project I wrote a whole toolbox in MATLAB to implement Logical Combinatorial Pattern Recognition LCPR paradigm developed by Prof. Shulcloper and applied it to images. Here's what we found in applying the algorithm to segmented binary images of alphabets.