The Flow Less Traveled: Documenting Independent Original Research on Fluid Flow Interactions in the Laurentian Great Lakes and Immediate Surroundings

File(s)
Date
2021-05-01Author
Hansen, Thomas F
Department
Freshwater Sciences
Advisor(s)
John Janssen
Metadata
Show full item recordAbstract
This work is a compilation of several research projects undertaken by the author. Each research effort identifies a problem that has been addressed traditionally using methods that are significantly costly, to such an extent that, in general, funding, convenience, and practicality are primary limiting factors to their effective implementation. In each case, the author has been able to either build upon existing, less expensive alternatives, or even invent novel approaches. The fundamental recurring research question is, can creative, even novel, computational approaches make more efficient use of resources to interpret or present data in such a way as to make what was previously impractical, inconvenient, or simply unachievable, now well within reach of even those with modest budgets? The first detailed is a research program which examines applications of digital in-line holographic microscopy, which operates by effectively replaces physical lenses with computational methods. The next uses digital storage and visualization to create real-time updating bathymetry maps using single-beam sonar already installed on most aquatic vessels. Also reviewed here is an interactive real-time computational fluid dynamics (CFD) model in which the user can simply reach out their hand and see the simulated fluid flow respond with swirling vortices and other fluid phenomena, a powerful tool for live demonstrations using equipment already installed in most classrooms. Finally, machine learning is applied to the problem of measuring fluid flow, in which a simple subsurface float can be used to measure current speed nearly as well as an expensive ADCP using a trained neural network running on an ordinary PC. Each are examples of replacing or augmenting existing, relatively expensive methods with far less costly alternatives, bringing the science of freshwater to a potentially much wider participatory scale than previously possible.
Subject
deep machine learning
freshwater
instrumentation
neural networks
Permanent Link
http://digital.library.wisc.edu/1793/92635Type
dissertation
