Computational aspects of inverse analyses for determining softening curves of concrete

Inverse analyses of laboratory experiments for determining strain softening curves of cementitious materials are an important tool of applied fracture mechanics and provide the basis for computational simulations of failure processes in civil structures. In these analyses, experimental results are approximated by repeated numerical simulations of the tests. A new optimisation method for this fitting process is proposed. It is based on an evolutionary algorithm including local neighbourhood attraction for convergence improvement. Several aspects of applicability and performance of the method as well as the objectivity of the results are discussed and applications are presented.