In order to assess the safety against fracture of nuclear shipping casks, drop tests are simulated using the finite element method. The simulations are based on a global model without defects. From the results positions are derived at which a crack initiation from a defect would be most likely. At each of these positions a submodel with a crack is inserted. The insertion of the crack changes the stiffness and therefore the stress of this region. Depending on the submodel size the calculated fracture mechanics parameters deviate from the reference value found without using the submodel technique.
The objective is to determine this deviation only from one submodel. A practicable method for the determination of the required submodel size is developed calculating the difference of the traction vector between the submodel with opened and with closed crack. Its scalar product with the weight function results in the contribution of the submodel boundary to the deviation. Integrated over the boundary it yields the deviation of the submodel. The shape of the submodel can be optimized by expanding sections of the submodel boundary with a high contribution.