Statistical process monitoring of continuous catalytic reforming heat exchangers using canonical variate analysis

Continuous petrochemical processes may have a large number of variables which are always driven by random noise and disturbances; and they are usually operated under closed-loop control yielding large amounts of process measurements that are auto-correlated and cross-correlated. Canonical variate analysis (CVA) is a dimensionality reduction technique through maximizing the correlation degree between two sets of variables; and the obtained canonical variates (CV) can maximally explain the information of variable sets. In this paper a statistical process monitoring model based on CVA is proposed. The proposed model combined with control charts is applied to an existing industrial continuous catalytic reforming (CCR) unit to monitor two parallel vertical shell-and-tube heat exchangers. The calculation results show that the proposed model is an efficient statistical process monitoring technique to detect error states and has potential for application in continuous petrochemical processes.