Sensor Fault estimation for a Nonlinear CSTR using Super-twisting and High-gain Observers

This paper presents the design of a super-twisting with a high-gain observer-based state and sensor faults estimations for a continuous stirred tank reactor (CSTR). The proposed schemes transform the original system into two subsystems, the first subsystem includes the nominal plant where a high-gain observer is proposed for state estimation, whereas the second subsystem has only sensor faults where an adaptive-gain super-twisting is derived to handle sensor faults and hence can be reconstructed from the equivalent output error injection signal. The proposed observer by means of the higher order sliding mode guarantees the reconstruction of faults without chattering effect caused by the high frequency switching of the traditional sliding mode. Using the Lyapunov theory, the global observer convergence is proved. Simulation results show that the designed observer yield robust stability and guaranteed good performance estimation and reconstruction.