Generative AI Designs Hydrogen Gas Turbine Combustors

Researchers have utilized invertible neural networks (INNs) to create designs for gas turbine combustors that can operate on 100% hydrogen in premix mode while maintaining low NOx emissions. Transitioning to hydrogen fuel necessitates a comprehensive redesign of combustion systems for all engine sizes, ranging from 4 MW to 600 MW, to avoid flashback issues. The team trained the INN using a database of combustor designs, which were geometrically parameterized and labeled with simulated performance data. By reversing the INN, they produced several design options that fulfilled specific performance criteria. This method aims to minimize the extensive engineering work required and facilitate knowledge transfer across different engine types. The findings were published on arXiv, classified under artificial intelligence.