ABSTRACT: Recent advancements in data science and machine learning have accelerated the development of data-driven models for complex flow phenomena. By leveraging innovative machine-learning and statistical tools, these models significantly reduce the computational time associated with traditional numerical methods. In this talk, I will present two examples of our recent progress in data-driven modeling using Gaussian Processes (GP) for diverse flow problems. In the first example, we explore efficient modeling of complex two-phase suspension flows. We demonstrate how GP can effectively bridge microscopic and macroscopic computations, enabling accurate and efficient modeling of bulk behaviors in two phase flows. In the second example, we integrate GP with Spectral Proper Orthogonal Decomposition (SPOD) to emulate the spatio-temporal evolution of turbulent flows. This model successfully captures the fundamental frequencies of turbulent wakes behind a square cylinder across varying Reynolds numbers. The talk will conclude with a summary of findings and future research directions.

BIOSKETCH: Yi-Ju Chou obtained his PhD degree from the Civil and Environmental Engineering Dept. at Stanford University in 2009, majoring in environmental fluid mechanics with a minor in computational and mathematical engineering. He served as a postdoctoral scholar at Stanford from 2009 to 2011. He then joined the Institute of Applied Mechanics at National Taiwan University, where he has been a full professor since 2020. Dr. Chou’s research group focuses on advancing analytical and computational methodologies to study complex flow phenomena in both industrial and environmental contexts. Recently, the group has been actively engaged in developing rapid flow prediction techniques across multiple disciplines using various data driven approaches.