With the increasing closure of white water circuits in paper mills, breaks in the sheet of paper have become a systemwide disturbance. Upon recognizing that such breaks can be modelled as a Markov chain type of process which when interacting with the continuous mill dynamics yields a jump Markov model, linear quadratic jump regulation theory can be used to derive white water and broke recirculation strategies which minimize process variability. Reduced process variability comes at the expense of relatively large swings in white water and broke tanks levels. Since the linear design does not specifically account for constraints on the state-space, under the control law damaging events of tanks overflow or emptiness can occur. A methodology mainly founded on the first passage-time theory of stochastic processes is proposed to choose the performance measure design parameters to limit process variability while maintaining mean time between incidents of fluid in broke and white water tanks either overflowing, or reaching dangerously low levels, sufficiently long. The corresponding approximation technique involved in evaluating mean first passage-times of the controlled stochastic processes appears to have an applicability which largely exceeds the problem area it was designed for.
I -INTRODUCTIONA paper machine is a very complex unit operation. It has several types of tanks that allow the recuperation, the storage and the recirculation of white water (water coming from the pulp and containing small fibres which are responsible for its white colouration) and broke (pulp coming from the paper breaks). Paper breaks could occur in different parts of a paper machine and are recuperated in generally distinct broke tanks. White water coming from the paper machine is also recuperated in a white water tank. Whenever necessary, the brokes are diluted with white water until the mix reaches an appropriate level of fluidity to be sent back to the paper machine. Also the thermomechanical pulp mill (TMP) which provides the pulp to the paper machine, allows the recuperation of a significant amount of white water.In [4] a method to optimally control the broke and white water tank levels and the variation of their inputs and outputs is proposed. It is founded on jump linear regulator theory. The solution to the problem emerges as a piecewise constant gain feedback law, for a given choice of the performance measure. However, this solution does not account for the fact that tanks can overflow or become empty. Therefore an iterative method is suggested to adjust parameters of the performance measure so as to achieve minimal recirculation flows variability while keeping mean times to overflow or emptiness of tanks at sufficiently high levels. The implementation of the tuning methodology for performance measure parameters requires in general repeated, both lengthy and computationaly expensive Monte-Carlo types of simulations. The aim of this paper is to use the theory of stochastic processes and their first passagetimes to develop approximat...