Uncertainty Quantification for Multi-fidelity Simulations
Swapnil Kumar
Imperial College London · Master Thesis
High- and low-fidelity aerodynamic data from Nektar++ and XFOIL are fused through co-kriging, adaptive sampling and multi-fidelity deep neural networks. The study evaluates lift and drag prediction, uncertainty propagation and benchmark functions extending to 100 dimensions while reducing reliance on costly high-fidelity runs.
- [ Multi-fidelity ]
- [ Uncertainty quantification ]
- [ Nektar++ ]
- [ Surrogate modelling ]
Uncertainty Quantification for GPU-Accelerated High-Fidelity DNS Investigation of Supersonic Retro Propulsion Flow During Martian Descent Using High-Order Flux Reconstruction in PyFR using Surrogate Modelling
Swapnil Kumar · Debdoot Ghosh
Imperial College London · Research presentation
A GPU-accelerated direct numerical investigation of supersonic retro-propulsion during Martian descent using high-order flux reconstruction in PyFR. The work studies shock interaction, shear-layer instability and recirculation while using co-kriging, adaptive sampling and machine-learning surrogates to reduce the cost of uncertainty quantification.
- [ Supersonic retro-propulsion ]
- [ PyFR ]
- [ GPU DNS ]
- [ Mars descent ]
Fluid Topology Optimization of Film Cooling: a Design Exploration
Francesco Montomoli · Thomas W. Rees · Stefano Furino · Ruben Tomlin
ToffeeX · Imperial College London · ETC2025-293
Fluid topology optimisation of film-cooling holes for additive manufacturing, starting from the ITS fan-shaped hole and a wider design domain. Optimised designs show branching ducts, multi-hole exits, upstream steps and coolant pockets, with effectiveness gains that depend on free-stream turbulence and pressure-loss reductions up to 30 %. Reproduced in XLAB Studio as the Film cooling FTO study.
- [ Film cooling ]
- [ Fluid topology optimisation ]
- [ Gas turbines ]
- [ Additive manufacturing ]
No machine-drafted papers have passed governance yet. They appear here once a Studio run is written up and approved.