Modeling and Prediction of Vibration-Induced Gas Turbine Blade Failure Using Artificial Neural Networks

Gas turbine blades are subjected to significant operational loads, with vibration-induced failures being a primary cause of blade damage. These vibrations, caused by aerodynamic forces, centrifugal forces, and temperature gradients, can result in fatigue and eventual failure. Predicting vibration-induced stresses is critical to increasing turbine dependability and lifetime. This study investigates using Artificial Neural Networks (ANNs) to model and predict the effects of vibration stress on gas turbine blades. Parameters such as rotational speed, pressure, and temperature data of the gas were obtained from Omotosho Phase II power plant in Ondo state and the Transcorp Nigeria Limited Power Plant in Ughelli, Delta state, Nigeria. A modal analysis was carried out using ANNs to examine the body's reaction to vibration as well as the inherent frequencies and mode shapes. With a correlation coefficient (R2) of 0.81474, the ANN simulation's ideal values for temperature, pressure, and speed were 535 0C, 16 bar, and 2782 rpm, respectively. The vibrational stresses and centrifugal force were 23.25MPa and 22384.3N. It was determined from the calculated values that the network has been trained correctly and may be used to forecast turbine operating circumstances to reduce failures and improve turbine performance. Turbine operators can use ANN-based simulations to optimize maintenance schedules, improve blade designs, and reduce failure risk.