ROBUST SMOOTH MODEL-FREE CONTROL METHODOLOGIES FOR INDUSTRIAL APPLICATIONS

Abstract

Model free control methodologies are popular in industry due to their easy implemen- tation. Minor tuning of controller gains yields satisfactory performance from a dy- namical system. The main drawback of the techniques is their lack of robustness. On the other hand, robust control techniques e.g. sliding mode control require mathemati- cal model of the system and their aggressive control effort is the main barrier in their implementation for mechanical systems. The proposed robust smooth control tech- niques with robust state-disturbance observer in the closed loop are the solution to the problem. The proposed state-disturbance observer is model free and relies on input and output of the system only; consequently it estimates states as well as drift term of the system. The estimated drift term is used to cancel out internal and external distur- bances of the system and this cancellation transforms the system into an nth order in- tegrator system. The observed states are used to design any modern or classical state- space control technique e.g. pole placement, Linear Quadratic Regulator (LQR) or Linear Matrix Inequality (LMI) methods etc. The finite time stability analysis of ro- bust state-disturbance observer is given in noisy and noise free environments. In this thesis, two novel control methodologies i.e. robust smooth real twisting second order sliding mode and robust feedback linearization are also proposed. The finite time sta- bility analysis of the robust smooth real twisting control is proven using Lyapunov method along with homogeneity concepts. The stability analyses of overall closed loop systems are given using separation principle. Simulations as well as experimental results with academic bench mark DC motor validate the ideas. The proposed tech- niques are also compared with robust LMI based polytopic controller on an industrial stabilized platform to verify their usage for industry.