End-to-end modeling of laminar-turbulent transition
Traditional methods of transition prediction cannot be easily extended to flow configurations where the transition process depends on a large set of parameters. Proposed model uses convolutional neural network and sequence-to-sequence mapping to predict transition in a physically-consistent manner, for a variety of instability mechanisms.
M. I. Zafar, M. M. Choudhari, P. Paredes, and H. Xiao. Recurrent neural network for end-to-end modeling of laminar-turbulent transition. Data-Centric Engineering 2 (2021).
M. I. Zafar, H. Xiao, M. M. Choudhari, et al. Convolutional neural network for transition modeling based on linear stability theory. Physical Review Fluids 5, 113903 (2020).
P. Paredes, B. Venkatachari, M. M. Choudhari, F. Li, C.-L. Chang, M. I. Zafar, and H. Xiao. Toward a Practical Method for Hypersonic Transition Prediction Based on Stability Correlations. AIAA Journal 58, 4475–4484 (2020).