Detection of decoupled input/output pairs in multivariable systems

Abstract

A method is proposed to detect if there is no coupling between an input and an output in systems operating in open-loop, that is, without a supervisory controller. The proposed technique is applicable to multiple input multiple output (MIMO) systems, whose intent is to detect no-model input/output (IO) combinations in a transfer matrix. Traditional approaches for selecting IO pairs are usually performed after the plant model is identified. The presented approach is applied during the pre-identification stage and is based on IO cross-correlation, signal filtering and fuzzy logic analysis. A case study involving the identification of a 7×6 simulated Fluid Catalytic Cracking (FCC) is discussed, as well as an influence analysis of detecting no-model IO pairs in the identification process and in the performance index of a Model Predictive Controller (MPC) applied to a 2×2 simulated distillation column. Finally, the method is tested with a real dataset obtained from an FCC unit of a petrol refinery.

Keywords: Cross-correlation analysis; Fuzzy logic; Input/output pairs detection; Multivariable identification; No models in the transfer matrix; Process identification.