Presenter:  Isabel Scherl, Postdoctoral Scholar, Caltech

Understanding and manipulating unsteady fluid flows is foundational to advancing many technologies. Improved modeling and control will increase the efficacy of systems across science and engineering. Due to limited state measurements, observation can be difficult. Many recently developed data-driven methods (i.e. machine learning) can alleviate these challenges. In this talk, Isabel Scherl will explore how data driven techniques can be used to study cross-flow turbine (i.e. vertical-axis) turbine arrays and other fluid systems. By employing online optimizations in the loop with experiments and robust modal decompositions with previously collected experimental data, we are able to better optimize, model, and control these flows. This work will be extended to multi-model ensembles for control and improved sampling of experimental systems using active learning for efficient discovery of new phenomena. She will show how cutting-edge algorithms in data assimilation, uncertainty quantification, and control have the potential to improve modeling and performance of wind and wave energy systems and our understanding of other complex, unsteady fluid flows.

About the speaker: Isabel Scherl is a postdoctoral scholar in mechanical and civil engineering at Caltech in the Computational and Data-Driven Fluid Dynamics group with Tim Colonius. She recently completed her Ph.D. at the University of Washington, advised by Steven Brunton and Brian Polagye in mechanical engineering. Her graduate research augmented experimental fluid mechanics with machine learning. Specifically, she focused on data-driven modeling, control, and optimization of vertical axis turbine arrays. Before UW, she graduated with a Bachelor of Science with honors in mechanical engineering from Brown University. Her continued focus is in applying data-driven methods to pressing challenges in fluid dynamics.