Novel PCA-Based Technique for Identification of Dominant Variables for Partial Control

The control structure problem which has been considered the central issue in modern process control has attracted large research attentions in the last three decades. There are various methods which have been developed but only a handful of them can really provide a practical solution to such an open-ended problem. The concept of partial control structure which has long been adopted in process industries has to a certain extent provided the engineers with sound theoretical foundation upon which this issue can be tackled in a systematic way. One major limitation of partial control preventing its effective adoption to solving this problem is due to its heavy reliance on engineering experiences and process knowledge in finding the dominant variables. In this paper, we propose a novel PCA-based technique to identify the dominant variables, which avoids the need for extensive experience and process knowledge. Various criteria and conditions forming the backbone of the technique are also proposed. The effectiveness of the technique is demonstrated based on its application to the continuous extractive alcoholic fermentation system.