Assessment of the multi-material ALE formulation with FSI
This paper presents LS-DYNA 970.5434a simulations for the multi-material arbitrary Lagrangian Eulerian (ALE or MMALE) formulation (solid ELFORM=11). The objective is to shed light on the dependence of the results on the count of advection cycles that has been found in former investigations of fluid-structure interaction (FSI) with soft, highly deformable structures [1,2]. The advection cycle count is determined by the NADV parameter on *CONTROL_ALE, the TSSFAC on *CONTROL_TIMESTEP and the penalty stiffness of the *CONSTRAINED_LAGRANGE_IN_SOLID interface. For an airbag example a strong random spread of the results on a dual CPU system is found. This spread is estimated to be inevitable. Other, more academic FSI examples are assessed then. The difference of the total energy and sliding interface energy between choosing TSSFAC=0.4 and TSSFAC default is usually bigger than that between the advection schemes alternatives: the now default van Leer scheme METH=2 and the first order donor cell scheme METH=1. Therefore, the dissipative properties of the schemes do not suffice to explain this difference. Furthermore, the influence of explicitly scaling down the default hourglass control in the MMALE elements for fluid and gas materials is smaller than the found difference. The strong influence of the parameters on the *CONSTRAINED_LAGRANGE_IN_SOLID card in conjunction with the advection count on the creation of non-physical interface energy proves to be responsible for non-robustness of the solution in some situations. The paper highlights some of the difficulties encountered and the solutions found. Input decks may be downloaded from www.rudolf-boetticher.de.
https://www.dynamore.de/de/download/papers/forum04/new-methods/assessment-of-the-multi-material-ale-formulation/view
https://www.dynamore.de/@@site-logo/DYNAmore_Logo_Ansys.svg
Assessment of the multi-material ALE formulation with FSI
This paper presents LS-DYNA 970.5434a simulations for the multi-material arbitrary Lagrangian Eulerian (ALE or MMALE) formulation (solid ELFORM=11). The objective is to shed light on the dependence of the results on the count of advection cycles that has been found in former investigations of fluid-structure interaction (FSI) with soft, highly deformable structures [1,2]. The advection cycle count is determined by the NADV parameter on *CONTROL_ALE, the TSSFAC on *CONTROL_TIMESTEP and the penalty stiffness of the *CONSTRAINED_LAGRANGE_IN_SOLID interface. For an airbag example a strong random spread of the results on a dual CPU system is found. This spread is estimated to be inevitable. Other, more academic FSI examples are assessed then. The difference of the total energy and sliding interface energy between choosing TSSFAC=0.4 and TSSFAC default is usually bigger than that between the advection schemes alternatives: the now default van Leer scheme METH=2 and the first order donor cell scheme METH=1. Therefore, the dissipative properties of the schemes do not suffice to explain this difference. Furthermore, the influence of explicitly scaling down the default hourglass control in the MMALE elements for fluid and gas materials is smaller than the found difference. The strong influence of the parameters on the *CONSTRAINED_LAGRANGE_IN_SOLID card in conjunction with the advection count on the creation of non-physical interface energy proves to be responsible for non-robustness of the solution in some situations. The paper highlights some of the difficulties encountered and the solutions found. Input decks may be downloaded from www.rudolf-boetticher.de.
J-I-23.pdf
— 472.2 KB