Physics-aware reduced-order modeling of transonic flow via <b><i>β</i></b>-variational autoencoder

Autoencoder-based reduced-order modeling (ROM) has recently attracted significant attention, owing to its ability capture underlying nonlinear features. However, two critical drawbacks severely undermine scalability various physical applications: entangled and therefore uninterpretable latent variables (LVs) the blindfold determination of space dimension. In this regard, study proposes physics-aware ROM using only interpretable information-intensive LVs extracted by $\beta$-variational autoencoder, which are referred as throughout paper. To extract these LVs, their independence information intensity quantitatively scrutinized in a two-dimensional transonic flow benchmark problem. Then, meanings thoroughly investigated we confirmed that with appropriate hyperparameter $\beta$, they actually correspond generating factors training dataset, Mach number angle attack. best authors' knowledge, our work is first practically confirm autoencoder can automatically field applied physics. Finally, ROM, utilizes compared conventional ROMs, validity efficiency successfully verified.