- This event has passed.
Applied Math Graduate Student Seminar: Harley Hanes, NC State, Talk 1 Low-cost Quantification of Fluid Flow Parameter Sensitivity using Reduced-order Modeling, Talk 2 Convergent Uncertainty Quantification in Fluid Dynamics using Reduced-order Models and Machine Learning
November 28, 2022 | 3:00 pm - 4:00 pm EST
Talk 1: Low-cost Quantification of Fluid Flow Parameter Sensitivity using Reduced-order Modeling
– Abstract 1: Sensitivity analysis for computational fluid dynamics (CFD) simulations is a complicated procedure, which still relies, in many cases, on engineering judgment and factors of safety. This is, in part, because the computational cost of quantifying the simulation’s sensitivity to all meaningful parameters (e.g., body surface roughness) and hyperparameters (e.g., subiteration convergence criterion) is intractable for even a single simulation. Reduced-order modeling dramatically lowers this computational cost of simulating fluid flows, but usually only where similar data is already available. In this work, fluid reduced-order models are utilized to quantify a flow’s sensitivity to certain physical parameters for the purposes of improved sensitivity analysis. Characteristic observability and sensitivity are both explored. The resulting sensitivity results enable more informed CFD frameworks and more rigorous uncertainty bounds on the resulting data.
Talk 2: Convergent Uncertainty Quantification in Fluid Dynamics using Reduced-order Models and Machine Learning
– Abstract 2: Constructing computational fluid dynamics (CFD) simulations from experimental data is a critical process in aerospace engineering design, but sparsity and error in local sensors limits the ability to condition CFD results with experimental data. These limitations often lead to different measurements of critical quantities in experimental and computational results. Neural networks can quantify nonlinear relationships between sparse or integrated sensor data and the corresponding flow-field, potentially increasing the accuracy of other CFD methods by conditioning them on initial estimates of flow-fields produced by neural networks. However, accuracy of this method requires investigating optimal placement of sensors and uncertainty quantification of network reconstructions. In this work, we investigate the sensitivity of a shallow encoder network CFD reconstruction to the location and readings of sensors using Morris screening and network sensitivity methods. The resulting analysis allows identification of flow-field regions where increased precision in sensor readings or placements most effects reconstruction accuracy.
SAS 4201 or through Zoom, with link https://go.ncsu.edu/amgss-zoom and passcode amgss