Control of a Bleach Plant in a Pulp Factory
The case study presented here was the first application of adaptive predictive control to an industrial production process, which was the bleach plant of the pulp factory of CANFOR Ltd. in Port Mellon in British Columbia, Canada. In this plant, pulp was fed into a mixing tower, where chlorine was added. Bleaching was produced as the result of two chemical reactions: a fast substitution followed by a slow oxidation. The first reaction was completed after a few minutes, while the second one took approximately one hour. The reaction rate was affected by the amount of lignin in the incoming pulp, as well as by the chlorine concentration, the wood species and the stock temperature. Thus the dynamics were complex and time-varying. Moreover, surges of black liquor carryover were frequent and very detrimental as they consumed a large amount of chlorine. This was therefore a difficult control problem in which there was a large economic incentive for good control. Chapter 8 of the book “Adaptive Predictive Control: From the Concepts to Plant Optimization” describes the implementation and results of this application.
The variable to be controlled was the residual chlorine in the tower, which was required to be kept close to zero using the chlorine flow as the manipulated control variable. The residual chlorine was measured using two oxide reduction potential (ORP) sensors. One of them (ORP2) was placed after the tower, supplying the information about the correct operation of the process. However, because of the long time required to perform the oxidation reaction in the tower, the residual chlorine was also measured after the substitution reaction by sensor ORP1, and this measurement was used as the feedback signal for control.
Adaptive predictive control was implemented in FORTRAN on an LSI 11/23 microcomputer under the real time operating system RT--11. The LSI 11/23 communicated with the PDP 11/44 mill process computer through a serial communication line RS-232.
