Comparison of the performance of a reduced-order dynamic PLS soft sensor with different updating schemes for digester control

In this work, the previously developed reduced-order dynamic PLS (RO-DPLS) soft sensor is extended to its adaptive version to address frequent process changes in a pulp digester. The properties of four model update schemes and the corresponding data scaling methods are investigated through one simulated case study and two industrial case studies of Kamyr digesters. Our findings obtained through extensive experiments are presented, which are expected to provide useful information and some guidance to practitioners. Finally, the effectiveness of the recursive RO-DPLS soft sensor is demonstrated through a digester closed-loop control case study, which shows that the closed-loop control performance can be significantly improved if the soft sensor prediction is fed back to a PID controller.