Thursday, April 16, 2020, 16:00
The first part of this talk focuses on the development of methods for integrating process operational safety and process economics within model predictive control system designs. To accomplish these critical control objectives, various model predictive control (MPC) schemes that maintain the process state within a safety region in state-space while optimizing process economics are considered for the first time. The second part of this talk proposes an integrated framework that combines a Neural Network (NN) algorithm with an MPC scheme that can guarantee closed-loop stability in the presence of deception cyberattacks (e.g., min-max cyberattack). While the aforementioned MPC formulations explicitly handle process safety, cybersecurity and economics considerations, they are centralized in nature and may lead to control action calculations that exceed the allowable sampling period. To address this potential practical limitation of the centralized MPC designs, the third part of this talk addresses the development of distributed model predictive control architectures. Nonlinear process examples will be used throughout the talk to demonstrate the applicability and effectiveness of the proposed control methods.