Our team undertakes research in the following areas.
supervised machine learning
design of experiment
Relevant papers may be found below.
Embedded ridge approximations
Chun Yui Wong, Pranay Seshadri, Geoffrey Parks, Mark Girolami
Constructing ridge approximations for vector-valued quantities, and scalar-valued functions that are integrals of vector-valued fields. For example, the lift, drag and pressure coefficients of an airfoil.
Computer Methods in Applied Mechanics
Paper | Preprint | Blog
Blade envelopes parts I & II
Chun Yui Wong, Pranay Seshadri, Ashley Scillitoe, Andrew Duncan, Geoffrey Parks
Interpreting the design of a blade (airfoil) as a Gaussian distribution with known mean (nominal) but unknown covariance. Estimating this covariance matrix by generating samples from the inactive subspace for loss in part I. In part II we extend the approach in part I to multiple objectives (flow capacity, loss and peak Isentropic Mach number) and demonstrating how this approach can be used for inverse design.
ASME Journal of Turbomachinery (under review)
Preprint I | Blog I | Preprint II | Blog II | Slides
Design Space Exploration of Stagnation Temperature Probes via Dimension Reduction
Ashley Scillitoe, Bryn Noel Ubald, Pranay Seshadri, Shahrokh Shahpar
Using polynomial variable projection to explore the design space of a stagnation temperature probe representative of those used in gas-turbine aero engines. Dimension reducing subspaces are used to find more accurate designs with less sensitivity to manufacturing uncertainties.
ASME Turbo Expo
Paper | Blog | Code
Turbomachinery active subspace performance maps
Pranay Seshadri, Shahrokh Shahpar, Paul Constantine, Geoffrey Parks, Mike Adams
On generating 2D contour plots of the entire 25D design space of a modern fan blade from a jet-engine.
ASME Journal of Turbomachinery
Preprint | Paper
Supporting multi-point fan design with dimension reduction
Pranay Seshadri, Shaowu Yuchi, Shahrokh Shahpar, Geoffrey Parks
Polynomial variable projection applied to three distinct fan blades at different operating conditions.
The Aeronautical Journal
Effectively subsampled quadratures
Pranay Seshadri, Akil Narayan, Sankaran Mahadevan
Subsampling quadrature points from a tensorial grid via QR with column pivoting on a weighted Vandermonde-type matrix.
SIAM/ASA Journal on Uncertainty Quantification
Paper | Preprint
Quadrature strategies for constructing polynomial approximations
Pranay Seshadri, Gianluca Iaccarino, Tiziano Ghisu
A review of recent approaches for identifying quadrature rules in hypercubes, followed by a template-type characterisation of how one arrive at new quadrature rules.
Uncertainty Modeling for Engineering Applications, Springer.
Coming soon! But in the interim check out the slides below:
Bayesian polynomial chaos
Extremum sensitivity analysis with polynomial Monte Carlo filtering
Chun Yui Wong, Pranay Seshadri, Geoffrey Parks
Deriving Sobol’ indices for polynomial ridge approximations, and developing methods for extremeum sensitivity analysis using Monte Carlo filtering and correlated polynomial approximations.
Reliability Engineering and System Safety (under review)
Preprint | Blog
Sensitivity analysis of a coupled hydrodynamic-vegetation model using the effectively subsampled quadratures method
The application of effectively subsampled quadratures for computing Sobol’ indices for a coupled hydrodynamic-vegetation model.
Geoscientific Model Development Discussions
Paper | Blog
Polynomial ridge flowfield estimation
Ashley Scillitoe, Pranay Seshadri, Chun Yui Wong, Andrew Duncan
Flowfield estimation for a new geometry or boundary condition given a relevant training repository of the same test case.
Physics of Fluids (under review)
Preprint | App