Run-to-run optimization methodologies exploit the repetitive nature of batc
h processes to determine the optimal operating policy in the presence of un
certainty. In this paper, a parsimonious parameterization of the inputs is
used and the decision variables of the parameterization are updated on a ru
n-to-run basis using a feedback control scheme which tracks signals that ar
e invariant under uncertainty. In this run-to-run framework, terminal const
raints of the optimization problem and cost sensitivities constitute the in
variant signals. The methodology is conceived to improve the cost function
from batch-to-batch without constraint violation. (C) 2001 Elsevier Science
Ltd. All rights reserved.