When: Apr 26 2019 @ 4:00 PM
Where: 132 Gilman Hall
132 Gilman Hall

4:10 pm Presentation
“Displacement Thickness-Based Recycling Inflow Generation Method For Spatially Developing Turbulent Boundary Layer Simulations”
Presented by SAMVIT KUMAR (Advisers: Profs. Meneveau & Mittal)
An improved method for generation of turbulent inflow for simulations of developing boundary layers is presented. The approach is based on prior recycling methods for flow over smooth (Lund et al., 1998) and rough (Yang and Meneveau, 2015) surfaces. Both these methods rely on obtaining δ99 from the mean velocity profiles based on a velocity threshold. Since this value is heavily dependent on the shape of the profile, it can be very noisy and can suffer from large undesirable fluctuations, even when the profiles are time averaged. A profile-integrated quantity, such as the displacement thickness, can be used instead of δ99. In the recycling method, mean and fluctuation velocities on a sample plane are rescaled, combined and recycled back to the inlet, as the inflow velocity. A roughness-length related scale is chosen for rescaling of the inner layer, depending on the surface geometry and the displacement thickness is chosen instead of δ99 as the length scale to rescale the outer layer. The blending function, dependent on both the inner and the outer length scales, is used to combine the two profiles, to obtain the inflow velocity. Since the displacement thickness depends on the profile shape, an iterative scheme is implemented. This cushions the effect which an unusual mean velocity profile at the sampling plane may have on the value of the outer length scale and hence, on the rescaled velocity profile. Some applications and test cases are presented.

4:35 pm Presentation
“Data-Driven Analysis of Aeroelastic Flutter”
Presented by KARTHIK MENON (Adviser: Prof. Rajat Mittal)
Data-driven methods to analyze fluid flows have been recently gaining popularity in many subdomains of fluid dynamics – from turbulence to reduced-order aerodynamic models. This has been primarily driven by our improved ability to generate large, high-quality data sets, and efforts to extract patterns from large amounts of data in an efficient manner. This talk will describe our initial work to understand the dynamics of aeroelastic flutter from one such data set consisting of over 500 simulations of a pitching airfoil under different conditions. In particular, the focus will be on our efforts to characterize all the possible vortex wakes behind the airfoil using a machine learning-inspired clustering method. We will also discuss our development of a Dynamic Mode Decomposition formulation that allows accurate decomposition of the flow around a moving boundary, which has until now remained a challenge for large amplitude boundary motion.