Modelling and Optimization of Peroxide Pulp Bleaching Process

Abstract: The wood pulp industry has been around for a long time, but new higher quality pulp require more advanced solutions to old processes. One of these processes is the peroxide (PO) bleaching process, which is the last of a whole chain of bleaching processes at the Mörrum pulp processing plant. The aim of this thesis was to develop and study a model for the peroxide pulp bleaching process, and thereafter optimize the process with the model. The PO-stage is a multivariable, non-linear process with a variable retention time of a few hours. The models tested was a kinetic reaction model, a Gaussian process regression (GPR) model and hybrid models of the two. Before the models could be tested a retention time estimation model was made to compensate for the variable retention time. The bleaching simulations showed the kinetic model could not accurately model the brightness output. The kinetic lacked variability in the maximum brightness parameter C¥. However, the combination of estimating C¥ for the kinetic model with an GPR model proved to be a good performing model. Prediction on slow brightness changes was accurate, but fast changes was harder and large error could occur. The optimization of the hybrid model showed that chemical dosages could be lowered while achieving a smoother and more precise brightness. Further studies on robustness of the brightness model and the optimization model are needed before implementation on the real process can be done.