The vast majority of large-scale chemical energy conversion systems into power or thrust are based on gas and steam turbines. The efficiency of the energy conversion mechanism is limited by both thermodynamics and aerodynamics constraints. The latter, which has a strong impact on the former especially on the turbine side, requires accurate aero-thermodynamic analysis of the fluid flow. This is normally accomplished by CFD (Computational Fluid Dynamics) which tries to account for both large scale unsteadiness and turbulence. The impact of turbulence on the mean flow is generally modeled by Reynolds-averaging (URANS) or space averaging (LES), as the resolution of all time and length scales at the high-Reynolds-number of real applications by DNS will call for prohibitive computational resources. URANS and LES have some known fundamental limitations that, while compromising the prediction accuracy, do not allow further design improvements. The limitations c an only be partly resolved by using experimental testing and advanced measurements that indeed help and had provided so far the ground for any design tool improvement. Still, testing cannot be the only source of data in a design process, where the accurate prediction of heat and fluid flow is of paramount importance. Here, GPUs are an enabler as they might allow very large scale simulations with computational time and cost able to meet industry needs of deep fundamental physics understanding to improve energy conversion systems efficiency.
|Source||Hours||Start Date||End Date|