Optimization with uncertainty - A case study for controller recalibration

Abstract Although Advanced Process Control (APC) applications have a very attractive economic return, they are often turned off after a year or so because changes in the process or operating conditions create a mismatch between the process and the original models and constraints. Periodic recalibration of the models can restore function, and substantially extend the useful life of an APC project. Through optimization of a stochastic simulation, this work shows that the recalibration schedule is important, and as well, relatively insensitive to a precise date. This case study also reinforces a general methodology for optimizing stochastic processes.