Excitation Versus Control Issues in Closed Loop Identification of Plant Models for a Sugar Cane Crushing Mill 1

Abstract This paper addresses the selection of excitation signals to enhance the identification of a linear model of a plant system from data collected whilst the plant is operating under closed loop control. Theoretical results are developed which characterize the effective frequency weighting induced by the closed loop to the modelling criterion. These results then are used to establish the cause of ‘identification of the controller’ and to study mechanisms for influencing the model fit to focus on the plant system through the selection of exogenous excitation signals introduced to the loop. Experimental results of closed loop identification with and without excitation on the plant inputs are presented. The plant models identified here were used to develop MISO LQG controllers for a Crushing Plant in a Sugar Mill.