Contributed talk

Speaker: Jade Buzinski (Arizona State University)

Location and Time: SAS 221A, Saturday, 10:40–11:10 AM

Title: Multiscale Analysis of Laboratory-Generated Turbulence

Abstract: The ocean is responsible for a huge amount of heat capacity in the global climate system, owed in large part to the energy transfer driven by internal waves below the surface. These complex interactions are caused by less-dense fluid sinking through more-dense fluid at a variety of spatial and temporal scales. Due to the high cost of oceanic data collection, internal waves are frequently studied via numerical simulation or laboratory-scale physical experiments. One such experiment is the Stratified Inclined Duct (SID) setup, with which Cambridge University researchers have obtained over 50,000 frames of shadowgraph data, illustrating various flows. This work is part of an ongoing effort to automate regime classification of flows in real time, enabling research to prioritize relevant data and minimize storage. The authors have performed data analysis using the curvelet transform, which utilizes a multiscale approach and returns coefficients that may be used as input for an already-developed convolutional neural net (CNN). Work is also being conducted on temporalizing the CNN, both with a naive temporal labeling system, and with a convolutional long short-term memory network (LSTM),.