Fault Detection and Fault Isolation in Chemical Process System

Abstract

Fault detection and isolation are critical for industrial applications, especially, for chem- ical and petrochemical industries to maintain a continuous production cycle and also to

reduce maintenance time. The demand of the fault detection techniques has been in- creased in a recent decade for improving the operational safety, reliability and availability

of the overall system. This research work is focused on improving the aforementioned parameters of the overall system by using the fault detection and isolation schemes, specifically by using the parity equations. In this context, firstly, the mathematical models of three-tanks and four-tanks systems have been developed to investigate the behaviour of a system with time under influence of external disturbances. To study the dynamic behaviour of the system, the nonlinear mathematical models are linearized by using the Taylor series followed by the Jacobian matrix transformation to obtain a linearized state-space model.

Direct and temporal parity equation methods are derived in this work as a fault de- tection strategy to detect the faults in the sensors and actuator. The derived methods

are then tested on the proposed three and four tanks systems to determine whether the output data in the form of residuals indicate a fault in the system. Fault isolation technique of structured residual has been derived and tailored in this research work to

isolate the faults. The faults code table is generated for a better understanding and rep- resentation of the structured residuals. The proposed parity equation method is applied

on the three and four tank systems in MATLAB by purposely injecting the faults in the system. The results show that the proposed methods successfully detect and isolate the injected fault from both three and four tanks systems.