Numerical investigations of collision experiments considering weld joints

• The results of two collision experiments are shown.
• In numerical analyses the experimental results are reproduced.
• Filled welds are considered through an analogous model in finite element simulations.
• The impact velocity in numerical analyses has a minor influence on the results.

The design of double hull structures is investigated through two collision experiments with different large-scale specimens. The experiments were performed with a design of a typical double hull structure and a design with an alternative stiffening system (English: plate strengthened stiffeners, German: Plattenverstärkte Profilsteifen – PVPS) invented by Röhr and Heyer (2007) [1]. In contrast to the typical double hull design the alternative stiffening system is characterized by an additional trapezoidal sheet metal which is welded on the bulb profiles of the outer shell. The focus of this investigation is to reproduce the experimental results in terms of the force–displacement curves and failure mechanisms in numerical analyses. For this purpose, true stress–strain relations are determined from tension and compression tests of the different steel batches and the weld metal of the large-scale specimens. The RTCL and Lou–Kõrgesaar–Romanoff failure criteria are calibrated using tension tests and applied to the numerical analyses of the experiments. Both failure criteria lead to nearly identical force–displacement curves as well as failure mechanisms and they are capable of reproducing the experimental results. In the finite element model, an analogous model of fillet welds is implemented by assigning a fictitious thickness and material behavior of the weld metal shell elements at the location of weld joints. It is necessary to consider weld joints with the analogous model in the numerical analyses to predict the maximum reaction forces in the force–displacement curves. Furthermore, the impact velocity in the numerical analyses is investigated and it is concluded that an impact velocity of 10 m/s is appropriate to balance the accuracy and the simulation time.