Hierarchical modelling in benchmarking, analysis and code development for coupled geo‐processes

Abstract

Geotechnical safety analysis relies on numerical simulations capturing interacting physical phenomena. Systematic model development and selection are vital for interpretable results without black‐box aspects. Interpretability implies that models can be checked for plausibility and consistency as a whole and in terms of individual components by the handling engineers. This contribution discusses the following three related aspects: (1) confidence‐building through embedding process models in a complexity hierarchy that ensures physical consistency among its members and provides an avenue for software testing and confidence building; (2) combining probabilistic engineering analyses with such model hierarchies to allow selection of the most appropriate process model given a set of experimental data while avoiding overfitting and (3) separating distinct model aspects both technically and logically for reproducibility, efficiency and quality assurance of simulation software. The latter aspect is illustrated based on the example of a material model database that can be linked to different process codes.