State estimation and non linear model based control of benchmark processes using unscented kalman filter

Abstract

In this work, Non-linear Model Based Control (NMBC) scheme for newlinethe non-linear system is proposed. The proposed control law is a function of newlinethe model states and the measured output variable. Performance comparison newlineof the proposed NMBC scheme with that of Non-linear Model Predictive newlineControl Scheme (NMPC) is reported. The proposed model based control law newlinehas a single tuning parameter and it can be optimally determined by newlineminimizing a suitable performance measure. From the extensive simulation newlinestudies on benchmark processes such as Continuous Stirred Tank Reactor newline(CSTR), Conical-tank and pH processes, it can be inferred that servo and newlineregulatory performances of the proposed NMBC scheme is found to be newlinematching closely with the NMPC scheme, with the proposed control scheme newlinebeing computational less intensive. Performance comparison of NMBC with newlineNMPC, highlighting offset free servo performance, rejection of unmeasured newlinedisturbance and robustness to model-plant mismatch are reported. newlineIn addition, a Non-linear Model Based Control Scheme using an newlineUnscented Kalman Filter (UKF-NMBC) for the process affected by random newlinedisturbances and random errors in measurements are formulated. The newlineproposed control law is also a function of the model states and the measured newlineoutput variable. An Augmented Unscented Kalman Kilter (AUKF) is used to newlinegenerate unbiased state estimates in the presence of step-like changes in the newlineunmeasured disturbances. From the extensive simulation studies on newlinebenchmark processes such as Continuous Stirred Tank Reactor (CSTR), newlineConical-tank and pH processes, it can be inferred that servo-regulatory newlineperformance of the proposed AUKF-NMBC scheme is also found to be newlinematching closely with the Non-linear Model Predictive Control Scheme using newline newline