Empirical approach to robust gramian-based analysis of process interactions in control structure selection

This paper deals with the estimation of a gramian-based interaction measure from logged process data, and thereby removing the need of creating parametric models prior to the selection of the significant input-output interconnections. Moreover, the resulting confidence regions of the estimates can be used to perform a robust control structure selection. The considered interaction measure is the Participation Matrix. Based on previous results, a new unbiased statistic is proposed, and confidence bounds for the estimate are derived. Examples and a case study are used to illustrate how the method can be applied.