Rambod Mojgani, University of Illinois-Urbana-Champaign
Pedram Hassanzadeh, Rice University
Although computational resources are becoming increasingly more powerful, accurate and many-query simulations of partial differential equations remain prohibitive in many of the problems that involve multiple spatio-temporal scales, stochastic nonlinear dynamics, and/or multiple physical processes. Fortunately, data, either from high-fidelity and yet expensive simulations or sparse measurements in the physical world, are becoming more abundant, and consequently, modern approaches seek to leverage the emerging potential of this abundance of the data to construct purely data-driven or hybrid (physics-based + data-driven) models enabling predictive simulations with a fraction of the cost of the high-fidelity, first principle-based models. Fluid dynamics, with a broad range of applications in engineering, biological, and geophysical systems, and their rich dynamics can benefit from this shift in paradigm. These approaches also aim to provide us with opportunities to expand our knowledge of fundamentals and underlying mechanisms such as flow instabilities and transition to turbulence, development of turbulence, and heat/mass transfer.
This symposium seeks to discuss data-driven models, hybrid models, reduced order models, and machine-learning efforts aiming towards problems in fluid dynamics, including but not limited to turbulence closure, multi-scale modeling, extreme events, and data assimilation.