Transfer Learning for Wall Temperature Prediction in Regeneratively Cooled 22N Thrust Chambers

Kurzfassung

ural network originally trained for hot gaswall temperature prediction in regeneratively cooledLOX/Methane rocket engines is adapted via transferlearning to predict hot gas wall temperatures in muchsmaller engines operated with nitrous oxide/ethane,which show different thermal behavior. The source netwas trained with synthetic data generated by CFD simula-tions, whereas the transfer learning was done with a lim-ited number of experimental data. The adapted networkshows good prediction of the hot gas wall temperature inthe test data set.