Neural Network-Driven Predictive Maintenance for Gas Turbines

The thermophysical characteristics of flue gas are significantly altered when hydrogen fuel is used in gas turbines. More steam content and greater specific heat capacity indicate that hydrogen-fuelled gas turbines may experience major health problems, such as erosion and corrosion in the hot gas channel caused by steam. Therefore, it is vitally important to diagnose problems in these hydrogen-fuelled turbines. There aren't many studies on defect detection for retrofitted gas turbines that take hydrogen as a feasible fuel choice, as far as the authors are aware. This study fills that gap by using MATLAB to create an artificial neural network (ANN)-based defect diagnosis model. The results demonstrated that hydrogen-induced corrosion defects could cause false alarms, resulting in inaccurate fault identifications throughout the detection and isolation procedure, regardless of whether they manifested as a single problem or in conjunction with fouling. It's interesting to note that the fault identification module performed somewhat worse in the hydrogen fuel scenario than in the natural gas scenario. This was mostly because the hydrogen-induced corrosion was assumed to create smaller fault magnitudes.