题目：Physics-Informed Methods for Fluid Measurement: Statistical Particle Tracking and Background-Oriented Schlieren

时间：2022年12月2日 20:30-21:30

腾讯会议室：333779801

报告人：Dr. Samuel Grauer

邀请人：蔡伟伟 教授（叶轮机械研究所）

Biography

Dr. Samuel Grauer is an Assistant Professor of Mechanical Engineering and Faculty Fellow at the Institute for Computational and Data Sciences at the Pennsylvania State University. He completed his Ph.D. in Mechanical and Mechatronics Engineering at the University of Waterloo (Canada) in 2018 and worked as an NSERC Postdoctoral Fellow in the Ben T. Zinn Combustion Laboratory at the Georgia Institute of Technology from 2018 to 2020. Dr. Grauer’s lab at Penn State develops quantitative optical diagnostics and data processing techniques for measuring fluids, especially for high-speed and reacting flows. His group creates methods to increase the accuracy and spatio-temporal resolution of parameter estimates using high-fidelity optics and spectroscopy models, statistical inference, and data assimilation techniques.

Abstract

Optical diagnostics are essential to the study of fluid dynamics because they can achieve the high spatio-temporal resolution needed to resolve the wide range of scales in a complex flow without disrupting its development. As a result, techniques like particle tracking velocimetry (PTV) and schlieren have become ubiquitous. However, significant post-processing is usually required to quantify the variables of interest, such as Eulerian velocity, pressure, and density fields. Many optical diagnostics like PTV and quantitative schlieren involve one or more ill-posed inverse problems when calculating these fields, in which case supplemental information must be added to regularize the solution. This talk outlines two novel, physics-based techniques for enhancing optical measurements of fluid flows. First, a stochastic algorithm for PTV is introduced, which incorporates the physics of particle advection as well as the governing equations to improve the accuracy and resolution of PTV. The approach, called stochastic particle advection velocimetry (SPAV), is demonstrated using digital holograms of turbulent flows from simulations and experiments. Second, a physics-informed workflow for background-oriented schlieren (BOS) is presented. This technique can be used to estimate the density, velocity, and pressure fields in a supersonic flow from as few as two images, as shown through a series of numerical and experimental tests. Both SPAV and physics-informed BOS are conducted using a physics-informed neural network, also known as a PINN, to estimate flow fields that approximately satisfy the governing equations and agree with the measurements.