Robust Fault Isolation of Gas Turbines via Nonlinear Intelligent Observer and Takagi-Sugeno Fuzzy Inference System

Effective fault diagnosis approach in gas turbines is crucial for ensuring reliable and efficient operations, minimizing downtime, and mitigating safety risks. In this article, a robust fault detection and isolation (FDI) method for an industrial gas turbine is developed. First, the system is modeled through a nonlinear feedforward network, which consists of finite impulse response (FIR) filters, and radial basis function neural networks. Following that, appropriate thresholds are estimated for the residuals. Finally, a Takagi-Sugeno fuzzy inference system is designed for isolating the faults. The fuzzy rules are developed not only for healthy and faulty conditions but also for the condition that is supposedly healthy, but the residuals exceed their thresholds due to some reasons such as severe environmental conditions. The main advantage of this work is that the proposed nonlinear structure efficiently deals with nonlinearity and uncertainty. Another contribution is to enhance isolation phase using the proposed fuzzy inference system. Experimental test results indicate that the proposed structure is very accurate and leads to less miss and false alarm rates (FARs).