• Citer

A time-lag autoencoder reduced-order model to predict combustion chemical kinetics

par Castellanos, Luisa;Da Silva Machado De Freitas, Rodolfo ;Parente, Alessandro ;Contino, Francesco
Référence European Combustion Meeting(11: 26 - 28 April 2023: Rouen, France)
Publication Publié, 2023-05-08

Publication dans des actes

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Modeling chemical kinetics is a very demanding task with the current simulation techniques, the reason why there is an ongoing effort for the development of accurate reduced models for their accurate representation. In this work, a technique based on a time-lag autoencoder is applied for methane ignition cases; such a study allows for the selection of physical features that allows for a statistical description of the chemical kinetics. The chemical features are used for the construction of ROMs through the application of clustering techniques, which allows the grouping of the obtained chemical features for the posterior application as predictive models. The obtained results are promising for the development of cheap-to-compute models that promise higher compatibility with state-of-the-art CFD techniques allowing to accelerate simulations with detailed chemistry.