From the summary: ``Designing a model for any process such as a distillation column, stirred tank reactor, or power plant boiler is highly complex, involving higher-order dynamics. We present an approach that uses canonical variate analysis, a Markov process and Akaike information criteria for system identification and reduction in order of the model. The optimal model structure of a distillation column is found using Akaike information criteria. In the method proposed, the statistically significant values are first obtained by using the CVA approach, thus providing a mathematical basis for the model. The open loop response data are collected from the 40m tall, 4.5m diameter distillation tower of Chennai Petroleum Corporation Ltd., India. Ten manipulated variables are reduced to two, which in turn are used for refined model selection.''