This work focuses on robustness of model-predictive control with respect to
satisfaction of process output constraints. A method of improving such rob
ustness is presented. The method relies on formulating output constraints a
s chance constraints using the uncertainty description of the process model
. The resulting on-line optimization problem is convex. The proposed approa
ch is illustrated through a simulation case study on a high-purity distilla
tion column. Suggestions for further improvements are made.