Modeling and Simulation visions towards 2020

1. Modeling and Simulationvisions towards 2020 Timo Tiihonen

2. Ingredients of the vision • Trends in research environment • Idea of a research vision • Own expertise • Recognized contributors and stakeholders • First steps on the road map

3. Trends in the environment • Computation • Computational capacity will be increasingly distributed and heterogeneous • Multi core architectures, grid/cloud computing in multitasking environment • High capacity will be available for non specialists • Modeling • Multi disciplinary/multi-scale applications will come to main stream (to be used by non specialists) • Lot of (isolated) single model tools/solutions will be available • Users can not be assumed to be specialists in model selection and coupling

4. Research goal • How to cope with complex, multi disciplinary, multi scale computational models • How to (automatically) divide complex systems to interacting subsystems • How to select most appropriate modeling paradigm for each subsystem and assess its accuracy and complexity • How to map subsystems and tasks automatically to existing heterogeneous resources • How to encapsulate methods for sub-models to enable interaction and resource planning • How to maintain variety of methods optimized for different environments

5. Needed assets • Modeling and mathematics • Domain decomposition (art/science of coupling sub-models together so that overall model is accurate and numerical methods converge) • A posteriori error analysis (tool to assess the accuracy of local sub-models) • Software engineering • Reformulation of existing software to coordinated abilities to solve sub-models (model as a service) • Autonomous coordination of solution of sub-models using ad hoc heterogeneous resources

6. Personal position • Background • Basis in mathematical modeling • Need to shift focus towards IT on longer term (to be compatible with the faculty) • Awareness of autonomic computing (agents etc) • Possible contribution • Models with different accuracy and complexity • Use of model hierarchies during solution phase (preconditioning) • Coupling of sub-models over interfaces • Sensitivity of results w.r.t geometry of (sub)systems

7. Possiblecontributors • VaganTerziyan • Model/method as a service, automated coordination of subtasks • JaquesPeriaux (FiDiPro) • Scientific Cloud Computing (JYU, INRIA/Grenoble etc), database (of models and methods), multiscience • SergiyRepin • Reliable computing (a posteriori analysis and adaptive model selection) • RainoMäkinen • Fluid-Structure and sensitivity analysis w.r.t model coupling, automatic differentiation

8. Possible collaborators II • TuomoRossi • Solvers, domain decomposition, parallel computing, processor architectures • Kaisa Miettinen, FerranteNeri • (Multi-criteria) optimization as a service, optimization and hierarchical models, metamodels • Numerola • Multi disciplinary problems, model coupling, software tools, modeling languages

9. First (feasible) steps • Toy problems for a posteriori analysis with model hierarchies (like non-linear vslinearized model) • Coupling of continuum and mesoscale models (Lattice Bolzmann and Navier-Stokes) • Language to describe modeling and solution variants (vehicle routing) • Fast PDE-solvers for GPU clusters