Sensor-based Temporal Superresolution: Application to turbulent separated flow over a three-dimensional Gaussian hill
Date
2023-09-14
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The high Reynolds-number turbulent separated flow over a Gaussian speed-bump benchmark geometry presents challenges for predicting smooth-body flow separation. The lack of time-resolved experimental data further hampers the understanding of the three-dimensional unsteady dynamics. This thesis addresses these issues in two parts. First, a data-driven technique using high-rate surface-pressure sensors and long short-term memory (LSTM) neural networks is proposed to estimate aliased velocity dynamics from undersampled particle image velocimetry (PIV) data, revealing low and medium-frequency modes. Second, the three-dimensional unsteady wake dynamics is characterized using additional surface-pressure measurements and two-component PIV. Four dominant frequencies are identified, with a very low-frequency spanwise oscillation of the recirculating zone, two low frequencies associated with the primary separation front motion, and a higher frequency from shear layer vortex shedding. Proper orthogonal decomposition analysis highlights interactions between these modes. The instantaneous vortex topology is conceptualized to infer physical mechanisms that give rise to these frequencies.
Description
Keywords
Turbulent separated flows, Machine learning, Sensor fusion, Superresolution
Citation
Manohar, K. H. (2023). Sensor-based temporal superresolution: application to turbulent separated flow over a three-dimensional Gaussian hill (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.