Adaptive fault tolerant control with adaptive residual generation

Abstract

This paper presents an adaptive predictive fault tolerant controller and an adaptive residual generator in an integrated scheme applied to a non-linear process. A case study on the steam super-heating subsystem of a boiler, represented by a detailed non-linear physical based model, is described. The faults considered are additive, step like and occur on the actuator. The controller is a GPC, rendered adaptive by a RLS algorithm, and is designed as in the "classical" way, with the addition of the actuator fault estimate information. The adaptive residual generator is based on the parity space method, rendered adaptive through on-line estimation of the model parameters allowing residuals to stay near zero for different amplitudes of the set-point. Simulation results of this integrated scheme are presented highlighting the performance advantages of the adaptive residuals vs. the standard residuals and discussing the proposed modified GPC controller vs. the "classical" GPC.