Investigation of unknown input observer for sensor fault diagnosis for a CSTR process

Abstract

Abstract The main objective of this work is to determine sensor fault in a Continuous Stirred Tank Reactor (CSTR) process using Unknown Input Observer (UIO). For any practical systems, there will be unavoidable disturbances and uncertainties, which leads to deviations in the plant output. Therefore it is essential in the design of any fault diagnosis system to consider these effects. CSTR equipped with a mixing device, and it runs at a steady state with a continuous flow of reactants and products. In CSTR, the outlet concentration is controlled by changing the volumetric flow rate. Initially, the data of the process were collected and preprocessed. Finally, the state space model was obtained using system identification tool in MATLAB. Pole placement technique is used for designing UIO for the system affected by input disturbances. The designed unknown input observer is used for generating a residual signal, and it is a deviation of plant output and observer output. The residual signal is zero during the fault-free condition. When a fault present, there is a deviation in the residual signal. In this work bias fault, drift fault was detected. Output observer also designed in this work to highlight the feature of the unknown input observer. Finally, it is observed that the unknown input observer has eliminated the input disturbances than that of output observer and faults was also detected effectively.