Unknown input reconstruction: A comparison of system inversion and sliding mode observer based techniques

The problem of unknown input reconstruction for nonlinear dynamic system is addressed in this paper. A system inversion based method as well as a combination method which integrates sliding mode observer with system inversion is implemented for achieving this purpose on a Hex reactor system. The second order sliding mode observer is considered to exactly estimate the successive derivatives of the output, aims at replacing these information used in the differential algebraic polynomial obtained by system inverse. These two methods are compared in terms of convergence and response time to state initialization, as well as noise situation during operation. Simulation results show that sliding mode observer based approach is more suitable in the presence of noise compared to state-of-the-art of direct system inversion based approach, while convergence rates is obviously faster through system inversion method in ideal environment.