Automated control loop selection via multistage optimal control formulation and nonlinear programming

In this work, a novel approach based on the multistage optimal control formulation of the control loop selection problem is introduced. Currently, state-of-the-art approaches for controller loop design have been focused on data that yield only the pairings between input-output variables, and are not able to incorporate path and end-point constraints. Thus, they only produce the optimal loops for control purposes, without the simultaneous consideration of their optimal tuning. This formulation overcomes these drawbacks by producing an automated integrated solution for the task of control loop design, which also obviates the need for any form of combinatorial optimisastion to be used. To illustrate the procedure, as well as the advantages of the proposed scheme, different practical case studies are discussed and the results compared with those obtained with standard controller loop selection methods and their tuning. The results of the proposed approach show improved performance over previous methodologies found in the literature. Furthermore, the framework is extended to the selection of the control loops that must obey path and end-point constraints imposed by the underlying dynamical process. This task is usually difficult for classical methods, which violate them or exhibit underdamped response in some cases.& COPY; 2023 Institution of Chemical Engineers. Published by Elsevier Ltd. All rights reserved.

Automated summary

This work introduces a novel approach for the control loop selection problem based on multistage optimal control formulation. Existing approaches for controller loop design have focused on input-output variable pairings and cannot incorporate path and end-point constraints. The proposed approach overcomes these limitations by providing an automated solution for control loop design that considers both optimal loops and their tuning. Case studies demonstrate improved performance compared to standard methods, and the approach is extended to include path and end-point constraints, which are challenging for classical methods.