Modelling of Peroxide-Bleaching of Pulp Using Gaussian Processes

Authors:	Lenz Henning, Siemens AG, Germany Horn Joachim, Helmut-Schmidt-University Hamburg, Germany Runkler Thomas, Siemens AG, Germany Dinkel Markus, Gebr. Lang GmbH Papierfabrik, Germany Schmidt Thomas, Gebr. Lang GmbH Papierfabrik, Germany Sieber Albrecht, Siemens AG, Germany Mickal Volkmar, Siemens AG, Germany
Topic:	6.1 Chemical Process Control
Session:	Advances in Automation in Pulp Industry
Keywords:	Paper Industry, Process Models, Nonlinear Models, Gaussian Processes, Optimization, Quadratic Programming

Abstract

Recycling of paper is typically performed in a chain of processes. An important process is the peroxide bleaching implementing the means to achieve a high brightness quality. As this process step requires consumption of various chemicals, it is quite cost intensive. This paper presents means to optimize the bleaching stage to achieve an (adaptable) trade-off between cost and brightness: First, the time-varying deadtime of the bleaching stage is estimated using a Kalman filter. Second, the bleaching stage is modelled by Gaussian processes aiming to be used for optimization of the bleaching stage by application of a model predictive controller.