Classical and dynamic matrix control of Kamyr digesters - A comparative study

In Kamyr digesters, used for the production of paper pulp from wood chips, there are strong interactions between the many process variables, long time delays, and frequent disturbances. As a result, steady-state operation is difficult to attain and process control is a challenging problem. There are, however, major economic incentives to stabilize the operation of the digester. This paper investigates and compares, via extensive simulation using MATLAB, the performances of both classical and dynamic matrix controllers on Kamyr digesters, using the actual Purdue model as well as a transfer function matrix that is generated from the Purdue model and reported earlier. In classical control, the performances of both single-input-single-output (SISO) and multiple-input-multiple-output (MIMO) controllers were tested. A dynamic matrix controller that is built from four controlled variables, five manipulated variables, and two measured disturbances was designed and simulated using a discrete step-response model. Both constrained and unconstrained cases were addressed. It has been shown that a SISO feedback controller pairing the blowline Kappa number with the lower-heater outlet temperature, while treating all other major inputs as disturbances, performs much better than a decentralized MIMO controller based on a 4×4 transfer function matrix. The dynamic matrix controller (constrained and unconstrained) was remarkably superior over all conventional control structures.