Intelligent fault detection and diagnosis based o optimized fuzzy model for process control rig

Abstract

This thesis focuses on the application of artificial intelligent techniques in fault detection and diagnosis. Fault detection and diagnosis scheme is a technique used in supervisory systems. The function of the supervisory system is to indicate unnecessary process states and to take the most appropriate actions to maintain continuous operation and to avoid damages. There are two main methods in fault detection and diagnosis: model free and model-based. In this thesis, model-based fault detection and diagnosis is used. One of the research challenges in model-based fault detection and diagnosis of a system is to find the accurate models. The objective of this thesis is to detect and diagnose the faults to a process control rig. A technique for the modeling of nonlinear control processes using fuzzy modeling approach based on the Takagi–Sugeno fuzzy model with a combination of genetic algorithm and recursive least square is proposed. This thesis discusses the identification of the parameters at the antecedent and consequent parts of the fuzzy model. For the antecedent fuzzy parameters, genetic algorithm is used to tune them while at the consequent part, recursive least squares approach is used to identify the system parameters. The proposed method is used to develop fault model and to detect the fault where this task is performed by using residual signals. When the residual signal is zero or nearly zero, the system is in normal condition, and when the fault occurs, residual signals should distinctively diverge from zero. Meanwhile, neural network is used for fault classification where this task is performed by identifying the fault in the system. This approach is applied to a process control rig with three subsystems: a heating element, a heat exchanger and a compartment tank. Experimental results show that the proposed approach provides better modeling when compared with Takagi Sugeno fuzzy modeling technique and the linear modeling approach. The overall accuracy for classification results also shows the best performance of around 93%.