High fidElity numeRical siMulations of strongly coupled processes for rEpository syStems and design optimisation with physical models and machine learning (HERMES)
Objectives
This WP aims at the development of high-fidelity numerical models for simulations of strongly coupled THMC processes in repository nearfield, repository design optimisation and interpretation of mock up experiments using a combination of physics based models and accelerated computing assisted with machine learning and artificial intelligence.
Description of the WP
Process-based numerical simulations are the basis for in-depth system understanding, analysis of experimental observations and their upscaling. Despite the continuous growth of the computational resources, the realism of the models applied in the simulations of repository systems remains severely limited in terms of dimensions, time-space resolution and process couplings. Interpretation of experimental data, safety and cost driven design optimisation, model uncertainty analysis belong to the class of inverse problems. Numerical solution of inverse problems, imply iterative forward modelling until the solution converges to the optimal parameter set (e.g. satisfactory description experimental data, cost-safety design optimisation of repository optimisation, uncertainty analysis).
For both forward and inverse problems, some orders of magnitude improvement in the computational efficiency can be obtained by replacing the physical based solvers or its components with high fidelitysurrogate models. Particularly promising are the surrogate models based on machine learning for specific aspects of THMC coupled models, data exchange between models at different scales, reduction of big data and extraction of constitutive relations from big numerical, experimental and monitoring datasets.
Recent developments in the field of data sciences and computational efficiency of surrogate models on modern computer infrastructure opens the way for realisation of efficient coupled numerical models (Digital Twins) for real time numerical analysis of laboratory and field experiments, repository design, components optimisation and comprehensive safety analysis. Such numerical tools are essential for repository conceptualisation and the repository design optimisation in both advanced- and early-stage waste disposal programs
The advancement in the THMC coupled code and the surrogate modelling will be a game changer in repository design and optimisation as well as PA/SA ant the stage of conceptualizations. This project will play a central role in facilitating collaboration between WPs focused on experimental studies and application of the results to SA/PA. The details of these interactions should be defined in the next stage of WPs preparation.
Outcomes
- Family of high fidelity/high computation throughput coupled models for multiscale simulations of THMC processes in repository nearfield and host rocks offered via an open access EURAD model-hub (SRA-> Scientific Insight)
- Improved understanding of in situ repository evolution controlled by coupled phenomena (SRA-> Scientific Insight)
- Surrogate models for inverse modelling and large scale simulations for sophisticated optimisation of repository design with respect to THMC FEPs (SRA-> Innovation for Optimisation)
- Dedicated proxy (simplified) models for optimisation, PA/SA and repository design. Integration of the models into DT definition conducted in WP17. Adjustment of the modelling framework according to the output of WP17 on the DT formulations. (SRA-> Innovation for Optimisation)
- Unified collaborative platform and protocols for surrogate model development. Open access database with experimental datasets (including numerical data) for model testing (SRA-> Knowledge Management)
