(665b) On-Line Determination of Appropriate Control Loop Configuration Using Directed Spectral Decomposition of Process Data

An important part of designing effective decentralized multivariable control schemes is determining control loop configurations that minimize control loop interactions. Traditionally, the most appropriate control loop configuration is often determined using the relative gain array of a process obtained at a single operating condition, with the configuration unchanged for the duration of subsequent process operation. For systems with significant non-linear characteristics the appropriate control loop configuration may change with operating conditions, leading to poor control system performance and potential closed loop instability. For such systems, good control system performance can be maintained and closed loop instability avoided over a variety of operating conditions if the control loops can be reconfigured appropriately on-line using informative data collected during process operation.

Our goal is to develop and evaluate a procedure for on-line data-driven determination of the most suitable control loop configuration using directed spectral decomposition (DSD)—a spectral analysis technique that uses informative process data to determine connectivity strength between process variables. The procedure involves two steps: 1) quantifying connectivity strength between process variables using DSD; and 2) using connectivity strengths to ascertain the control loop configuration that minimizes control loop interactions at the current operating conditions. The procedure has been tested on a simulated example 3-input, 3-output non-linear stirred mixing tank under varying operating conditions. This presentation will focus on the development of the proposed procedure and the specific results obtained by applying the procedure to the stirred mixing tank.