Mechanical Engineering Seminar Series: Class 530.804
“Turbulent Scalar Transport: MRI Experiments, Large-Eddy Simulation, and Machine-Learning Models”
Presented by Professor John K. Eaton
Charles Lee Powell Foundation Professor and Martin Family University Fellow
Department of Mechanical Engineering, Stanford University
Turbulent scalar mixing is the controlling process in many technological and natural systems including boundary-layer heat transfer, volcanic plumes, fuel/oxidizer mixing and discrete-hole film cooling. Direct numerical simulations accurately represent turbulent transport, but lower fidelity models used for practical computations are not predictive. Magnetic Resonance Imaging (MRI) techniques to measure velocity and scalar concentration fields provide new insight into transport in complex flows, particularly for film cooling and urban canopy flows where we have focused our attention. For example, full-field measurements of isolated film-cooling jets reveal important regions with negative scalar diffusivity. Wall resolved large eddy simulations (LES) accurately reproduce the MRI measurements and are used to explore these unexplained phenomena. We have developed new Reynolds-averaged scalar transport models using physics-informed, machine learning (ML) techniques. The goal is accurate representation of a tensorial turbulent diffusivity leading to improved mean scalar field predictions. Our Tensor Basis Neural Net for scalar (TBNN-s) models trained using high fidelity simulations provides major reductions in predictive uncertainty of the mean scalar concentration relative to conventional models in cross validation tests. Interpretation of the ML models leads to new understanding of the flow features that cause failure of conventional models.
Dr. John K. Eaton is the Charles Lee Powell Foundation Professor of Engineering at Stanford University where he has been on the faculty since 1980. He earned all his degrees in Mechanical Engineering at Stanford. He conducts research in turbulence, convective heat transfer, advanced measurement techniques, multiphase flow. Recent emphasis has been on high-fidelity, rapid turnaround experiments in complex flows, measurement and modeling of turbulent mixing, and extreme sensitivity of certain high Reynolds number flows to small perturbations. Much of the work is characterized by close interaction between experiment and simulation and a recent emphasis is on new ways to use large experimental data bases to develop machine-learned models. Professor Eaton has supervised 54 completed Ph.D. dissertations including those of 21 professors. He won the Senior Award from the International Conference on Multiphase Flow, the NSF Presidential Young Investigator Award, and is a Fellow of the American Society of Mechanical Engineers and the American Physical Society. He has won the Tau Beta Pi award as the best engineering teacher at Stanford twice and currently holds the Martin Family University Fellowship in Undergraduate Education.