fault detection

Degradation due to the long time operation at the gas turbine may cause a fall in the combustor and compressor efficiency and characteristics, which makes a decrease in total efficiency and, an increase in fuel consumption with growth in the production of the pollutants. These phenomena are not distinguishable by conventional diagnostic systems. Although instrumental protection of gas turbine may be profited by the control system but, it depends to components damage result revealation. So, in this research, a special method using the Multiple Model approach, based on multi-operating points is developed to continuously estimation these kinds of faults, to sustain the major defects in an industrial gas turbine. The object is done by defining the main components health indicator variables. First nonlinear thermodynamic static and dynamic models are constructed and linearized. By a combination of those in a dynamic convex set the alternate adaptive model is developed that, could cover the dynamic of the gas turbine. By decoupling the adaptive filter residuals, estimation of the faults and operating point determination are executed. Finally, the efficiency of the proposed approach is demonstrated in a simulation environment.

The design of accurate fault diagnosing systems aides process safety and also improves product quality. Therefore، in recent years، many researchers have conducted work on this topic and suggested various techniques for fault detection and diagnosis. In this paper، the fuzzy system based on knowledge of the process has been proposed for fault detection and diagnosis. In order to improve the performance of the diagnoser system، the genetic algorithm has been used to optimize membership function parameters of fuzzy diagnosis system. To test the performance of the proposed diagnose system، toluene production plant was used. The simulation results indicate that the proposed algorithm is capable of diagnosing various combinations (from two to four simultaneous faults) of simultaneous faults.