A Probabilistic Approach to Process Identification and Control A Case Study in Pulp Bleaching Focused on the Stationary Probability Density Function

This paper uses the operation of a pulp bleaching reactor as a platform for the illustration and discussion of several ideas involving use of the stationary probability density function in process control. A computer simulation of the partial differential equations, derived from first-principles, describing the bleaching reactor is used to represent the true process. Empirical, discrete-time models of the reactor are then constructed using sampled input and output time series. In the model building phase, the influence of data distribution on final model quality is emphasized. Once the discrete-time model has been derived, two control laws are synthesized. One law is based on established design techniques while the other takes advantage of less recognized probabilistic concepts. The controllers are then applied to the true process model the relative performances discussed.