Scientific Computing · CWI Amsterdam
Towards truly predictive computational science
I develop methods for making predictions under uncertainty in physical systems, combining physical modelling with data-driven techniques — above all for problems in fluid dynamics. I lead the Scientific Computing group at CWI in Amsterdam, and am part-time associate professor at Eindhoven University of Technology.
The future of computational science lies in the combination of physical modelling and data-driven techniques. In my group we develop new methods and algorithms that enable the transition to truly predictive models, working at the intersection of uncertainty quantification, partial differential equations and discretization methods, computational fluid dynamics, machine learning, reduced-order modelling, and Bayesian inference. The models we develop are applied to sloshing of liquefied natural gas in tanker ships, aeroelastic predictions for wind turbines, and transport of multiphase flow in pipelines.
More about me · Research · Publications
ERC Consolidator Grant
I have been awarded an ERC Consolidator Grant for the project SYMBIOSIS, on “Entropy-driven model reduction for fluid flows” — turbulent flows with applications in the energy sector. Read more about SYMBIOSIS.
What I work on
- Stochastic differential equations
Combining physics, machine learning, and data assimilation - Reduced order models
Structure-preserving methods with enhanced stability and accuracy. - Structure-preserving neural networks and machine learning
- Closure modelling for multiscale problems
Such as turbulence, using “discretize first, filter next”. - Uncertainty quantification
Surrogate modelling for parametric PDEs, efficient solvers for Bayesian inverse problems, statistical learning for closure models, novel quadrature rules. - Computational fluid dynamics
Time integration for differential-algebraic equations, and structure-preserving discretization of the incompressible single- and multi-phase Navier-Stokes equations.
More detail on the Research page.
Open positions
There are several open positions in my group, as PhD candidate or postdoc — see vacancy 1305098 and vacancy 1304811 on the CWI website. These vacancies have closed. Further positions funded by SYMBIOSIS will be advertised on the CWI vacancies page; for frequently asked questions see ERC Consolidator positions: FAQ.
Research team
Robin Klein
PhD candidate
Henrik Rosenberger
POSTDOC
Syver Agdestein
Postdoc
Pardeep Kumar
PhD candidate
Recent publications
- Preprint A Unified Framework for Bayesian Data Assimilation with Generative Models and Observation Interpolants
- Preprint A differentiable software suite for accelerated simulation of turbulent flows
- Preprint Approaching the optimal closure: equivariance, inductive bias, and Reynolds-number generalization in data-driven LES
- Journal paper Energy-conserving neural network closure model for long-time accurate and stable 2D LES
- Journal paper Exact Operator Inference with Minimal Data