The workshop will present how machine learning methods can support simulations
performed by complex calculation codes: development of fast surrogate models, optimisation,
uncertainty propagation, code calibration, multi-scale/multi-physics
simulation, hybridisation of physics-based and data-driven models…
It will present the application of the methods to a large variety of physical phenomena
involving their solution of partial differential equations in space and time:
thermal-hydraulics, fluid and solid mechanics, thermodynamics, risk analyses…
A few examples will be taken out of the nuclear industry.
All steps of the development process will be addressed,
including data generation, development of machine-learning models, their
validation and their integration into the final solution.