Abstract:
Gas turbines (GT) are, and will continue to be, the backbone of aircraft
propulsion, as well as power generation and mechanical drive. A small GT
performance improvement is expected to have a fuel-spend advantage of
the billion-dollar order, together with a significant CO2 emission
benefit. Part of the possible performance improvements can be enabled by
continuously advancing the understanding of the GT flow physics and thus
further reducing the inaccuracy of current design tools based on
computational fluid dynamics (CFD). By exploiting the capability of our
high-fidelity CFD solver on leadership GPU-accelerated supercomputers,
we have been able to perform state-of-the-art high fidelity simulations
of turbomachinery flows. The generated data, therefore, can shed light
on the detailed fundamental flow physics, in particularly the behavior
of transitional and turbulent boundary layers affected by large-scale
violent freestream turbulence, under strong pressure gradient and
curvature. Furthermore, machine learning methods are applied to the
high-fidelity data to develop low order models readily applicable to GT
designs.
Language: English
|
