4:10-4:35 p.m. Presentation
“Machine Learning on Detecting Turbulent/Non-turbulent Regions in a Transitional Boundary Layer”
Presented by ZHAO WU (Adviser: Prof. Meneveau)
In the graduate seminar about two years ago, I presented the use of a self-organizing map (SOM), an unsupervised machine learning method, to detect the turbulent/non-turbulent regions in a transitional boundary layer (Wu et al., PRF 2019). We found the SOM can successfully distinguish the near-wall turbulent boundary layer (TBL) region and the non-TBL region, the latter of which includes the laminar portion of the flow and the outer flow. We said, this method works well without the need of setting any user-specified thresholds and the results are qualitatively good compared with the traditional threshold-based method.
However, the above results have some limitations: we only tested the method for a particular flow, the boundary layer, and the data used are not frame invariant. This time, I will report our recent (unsuccessful) trial on developing a more generalized method.
4:35-5:00 p.m. Presentation
“Extraction of Interfacial Force Coefficients for Air Bubbles Interacting with Intense Turbulence”
Presented by ASHWANTH SALIBINDLA (Adviser: Prof. Ni)
In bubbly ship wakes and breaking waves, air bubbles get trapped at the ocean surface and interact with intense oceanic turbulence until they rise back to the surface. Under these conditions, in order to predict the kinematics of these bubbles, it is important to understand the contribution of different forces (buoyancy, lift, drag, pressure and added mass forces) acting on them. However, such a force balance equation entails multiple interfacial force coefficients including the drag, lift and added mass coefficients that need to be modeled for closure. Therefore, we make use of our unique, simultaneous 3D experimental measurements of bubbles and flow around the bubbles that allows us to extract these three coefficients.