Ray Chang | Search arXiv

arXiv2022

Taylor dispersion in arbitrarily shaped axisymmetric channels

Advective dispersion of solutes in long thin axisymmetric channels is important to the analysis and design of a wide range of devices, including chemical separation systems and microfluidic chips. Despite extensive analysis of Taylor dispersion in various scenarios, most studies focused on long-term dispersion behavior and cannot capture the transient evolution of solute zone across the spatial variations in the channel. In the current study, we analyze the Taylor-Aris dispersion for arbitrarily shaped axisymmetric channels. We derive an expression for solute dynamics in terms of two coupled ordinary differential equations (ODEs). These two ODEs allow prediction of the time evolution of the mean location and axial (standard deviation) width of the solute zone as a function of the channel geometry. We compare and benchmark our predictions with Brownian dynamics simulations for a variety of cases including linearly expanding/converging channels and periodic channels. We also present an analytical description of the physical regimes of transient positive versus negative axial growth of solute width. Finally, to further demonstrate the utility of the analysis, we demonstrate a method to engineer channel geometries to achieve desired solute width distributions over space and time. We apply the latter analysis to generate a geometry that results in a constant axial width and a second geometry that results in a sinusoidal axial variance in space.

physics.flu-dyn↗

arXiv2021

Recurrent Neural Network for MoonBoard Climbing Route Classification and Generation

Classifying the difficulties of climbing routes and generating new routes are both challenging. Existing machine learning models not only fail to accurately predict a problem's difficulty, but they are also unable to generate reasonable problems. In this work, we introduced "BetaMove", a new move preprocessing pipeline we developed, in order to mimic a human climber's hand sequence. The preprocessed move sequences were then used to train both a route generator and a grade predictor. By preprocessing a MoonBoard problem into a proper move sequence, the accuracy of our grade predictor reaches near human-level performance, and our route generator produces new routes of much better quality compared to previous work. We demonstrated that with BetaMove, we are able to inject human insights into the machine learning problems, and this can be the foundations for future transfer learning on climbing style classification problems.

cs.LG↗