What's inside?
A multi-objective, multi-point optimisation framework was applied to a novel combined inducer-impeller LH2 turbopump using TURBOdesign1's machine learning-based RRS with CAE module. Trained on a small dataset, the system improved stage efficiency across 0.85–1.15 Qd, significantly enhanced cavitation resistance, and met head requirements, outperforming the conventional separate inducer-impeller configuration.
An 18% reduction in NPSHr, achieved not by redesigning from scratch, but by optimising blade loading and throat area using a machine learning model trained on fewer than 80 CFD simulations.
In this publication, you will:
- Discover how combining inducer and impeller blades into a single compact configuration can match, and in some conditions exceed, the efficiency of a conventional separate design
- Learn how a Kriging-based Machine Learning system, trained on fewer than 80 CFD simulations, simultaneously optimises efficiency across multiple operating points and enhances cavitation resistance.
- See how targeted blade loading and throat area modifications deliver an 18% reduction in NPSHr, without compromising head performance or best efficiency point location.
