/ Research

Data-efficient, physics-aware modelling for aerodynamics.

My research focuses on the development of data-efficient, physics-aware and interpretable modelling strategies for aerodynamics and fluid mechanics. The central question is how to extract useful predictive models from scarce, expensive, heterogeneous, and imperfect data.

/ 01

Aerodynamic surrogate modelling

Developing surrogate models for high-dimensional aerodynamic fields, including surface pressure distributions, transonic wing data, airfoil optimization, propeller aerodynamics, and aircraft components. Methods include neural networks, autoencoders, variational autoencoders, Gaussian processes, diffusion models, neural operators, and latent-space regression.

Surrogate modellingTransonic aerodynamicsSurface pressure predictionNeural operatorsVAEsDiffusion modelsGaussian processesLatent spaces
Related publications · see list
/ figure 01
Replace with aerodynamic pressure field (Cp on a wing surface)
/ 02

Multifidelity and data-efficient learning

Combining low-fidelity and high-fidelity aerodynamic data to reduce the need for expensive simulations or experiments. Includes multifidelity Gaussian-process regression, autoencoder transfer learning, adaptive sampling, active learning, conformal prediction, and uncertainty-aware model assessment.

Multifidelity learningActive samplingData fusionScarce high-fidelity dataConformal predictionUncertainty quantification
Related publications · see list
/ figure 02
Replace with multifidelity sampling diagram
/ 03

Physics-aware and soft-physics AI

Intermediate approaches between purely data-driven models and fully physics-constrained methods. The goal is to incorporate approximate, cheap, or partial physical structure into machine-learning models when full physics-informed neural networks are too expensive or difficult to deploy for realistic aerodynamic fields.

Physics-aware AISoft-physicsInductive biasFluid mechanicsScientific machine learningReduced-order modelling
Related publications · see list
/ figure 03
Replace with reduced-order latent-space visualization
/ 04

Flow control and optimization

Machine learning, genetic algorithms, reinforcement learning concepts, and optimization for the design and control of aerodynamic and thermal-flow systems. Active flow control, pulsed jets, plasma actuators, sweeping jets, convective heat-transfer enhancement, drag reduction, and flow-control optimization.

Flow controlGenetic algorithmsActive controlHeat transferDrag reductionJet actuationOptimization
Related publications · see list
/ figure 04
Replace with flow-control experiment image
/ 05

Experimental fluid mechanics and heat transfer

Experimental characterization of turbulent boundary layers, convective heat transfer, synchronized heat-flux and velocity measurements, infrared thermography, PIV/PTV, wall-bounded flows, and impinging/sweeping jets.

Infrared thermographyPIVPTVHeat transferTurbulent boundary layersWall-bounded flowsExperimental aerodynamics
Related publications · see list
/ figure 05
Replace with PIV / IR thermography snapshot
/ 06

Aerospace design and digital engineering

Application of surrogate models, optimization, and data-driven tools to aircraft design, propellers, drones, hybrid-electric aircraft, manufacturing-aware design, and digital twins.

Aerospace designPropellersDronesDigital twinsHybrid-electric aircraftManufacturing-aware optimization
Related publications · see list
/ figure 06
Replace with propeller / aircraft surrogate visualization