부산대학교 도서관

Data-driven modelling in fluid mechanics is a promising alternative given the continuous increase of computational power and data-storage capabilities. Highly non-linear flows which include turbulence and reaction are challenging to model, and accurate algebraic closures for the unresolved terms in large eddy simulations of such flows are difficult to obtain. In this study, an artificial neural network is developed in order to directly model an important unclosed term namely the unresolved stress tensor. The performance of this approach is evaluated a priori using direct numerical simulation data of a highly demanding flow configuration, a turbulent premixed V-flame, and compared against the predictions of eight other classic models in the literature which include both static and dynamic formulations.