Cohesive fracture models are an established tool for the prediction of fracture in concrete. However, the reliable estimation of fracture parameters, specifically traction separation laws, is a necessarysteptowardstheirsuccessfulapplication. Currently,fractureparameterestimationismainly performed by fitting simple analytical models to experimental results. This requires…
Cohesive fracture models are an established tool for the prediction of fracture in concrete. However, the reliable estimation of fracture parameters, specifically traction separation laws, is a necessarysteptowardstheirsuccessfulapplication. Currently,fractureparameterestimationismainly performed by fitting simple analytical models to experimental results. This requires the design and execution of dedicated experiments, for which analytical models are available. Numerical models, along with optimisation and inverse problem solution techniques have the po- tential to lift the limitations inherent in the above process, allowing for instance the estimation of fracture parameters from more general experiments, involving complex geometries and loading con- ditions. However, fracture is computationally demanding process to simulate, while the solution of inverse problems requires multiple model evaluations, which can render whole process infeasible. This work explores the application of a simple technique for accelerating fracture simulations to the identification of fracture parameters of concrete. The technique relies on statically condensing parts of the model that are not affected by fracture, thus substantially reducing the model size, while preserving the accuracy and generality of the original model. Furthermore, it can be combined with different discretisation schemes such as standard or extended finite elements (FEM/XFEM), allowing for increased flexibility. The accelerated models are combined with Bayesian optimisation, allowing to solve the inverse problem in a highly efficient way. The effectiveness of the proposed approach is demonstrated through physical and numerical experiments.