Robust Self-Tuning Control under Probabilistic Uncertainty using Generalized Polynomial Chaos Models

Abstract

A robust self-tuning controller for a chemical process is developed based on a generalized Polynomial Chaos (gPC) model that accounts for probabilistic time-invariant uncertainty. Using this model, it is possible to calculate analytical expressions of the one-step ahead predicted mean and variances of controlled and manipulated variables. The key idea is to consider these predicted values for performing online robust tuning of the controller through a quadratic optimization procedure. The gPC model is also used to identify overlap between consecutive probability density functions (PDFs) of manipulated variables and to find trade-offs between the aggressiveness of the self-tuning controller and robustness to uncertainty based on this overlap. The proposed methodology is illustrated by a continuous stirred tank reactor (CSTR) system with stochastic variations in the inlet concentration. The efficiency of the proposed algorithm is quantified in terms of control performance and robustness.