Gas turbine operating temperatures are projected to continue to increase and this leads to drawing more cooling air to keep the metals below their operational temperatures. This cooling air is chargeable as it has gone through several stages of compressor work. In this paper a surrogate based design optimization approach is used to reduce cooling mass flow on combustor tiles to attain pre-defined maximum metal surface temperatures dictated by different service life requirements.
A series of Kriging based surrogate models are constructed using an efficient GPU based particle swarm algorithm. Various mechanical and manufacturing constraints such as hole ligament size, encroachment of holes onto other features like side rails, pedestals, dilution ports and retention pins etc. are built into the models and these models are trained using a number of high fidelity simulations. Furthermore these simulations employ the proprietary Rolls-Royce Finite Element Analysis (FEA) package SCO3 to run thermal analysis predicting surface heat transfer coefficients, fluid temperatures and finally metal surface temperatures.
These temperature predictions are compared against the pre-defined surface temperature limits for a given service life and fed back to the surrogate model to run for new hole configuration. This way the loop continues until an optimized hole configuration is attained. Results demonstrate the potential of this optimization technique to improve the life of combustor tile by reducing tile temperature and also to reduce the amount of cooling air required.