A reliable judgement of the functionality, safety and reliability of structures requires forecasting models, which have to be judged with regard to their application and the quality of their evidence. At present, no scientifically established methods exist. Until now a quality evaluation of models in civil engineering is made, with very few specialized exceptions, based on the phenomenological experience and knowledge of the model user.

The overall goal in the second phase of the GRK will be to investigate the interaction between experimental models and their numerical counterparts and to adopt the methodology from the first phase for quantitative evaluation of model qualities taking into account all aspects of uncertainty within models, input data and coupling effects between partial experimental models. Similar to the definition for simulation models, also experimental models are seen, as partial models and coupled with theoretical/numerical models as hybrid coupled partial models. The task is to define measures of quality for these hybrid models, enabling the design engineer to assess their predictive capability.

Experimental models can be applied to the validation and verification of theoretical/numerical models and to parameter estimation. To define an experimental model, simplifying assumptions are necessary, e.g. with respect to experimental specimen, initial- and boundary conditions, applied measurement systems. Additional assumptions to define a hybrid experimental theoretical/numerical model, are introduced if the experimental data is used for system and model identification purposes. Therewith the question arises how these assumptions are influencing the quality of the predictive model results.

As outcome of the GRK we expect consistent procedures for the evaluation and assessment of numerical and experimental models in structural engineering and especially for coupling of both model worlds.