Modelling procedures for non-linear analysis of metallic structuctrual components

Abstract

This thesis describes the experimental testing and a numerical study of extruded aluminum components subjected to three different loading scenarios; three point bending with and without a notch and acompression test. The components are of the same AA6xxx-T6 alloy and have the same nominal crosssectional measurements. The material is modeled using a metal model that includes fracture modeling, isotropic work hardening and the YLD-2004-18P anisotropic yield surface applicable to full stress states. The numerical study focuses on investigating the capabilities and limitations of large (3-5 mm) reduced integrated shell elements which is the present industry standard for large scale simulations. The performance of the shell element simulations is evaluated by comparing them to simulations using small solid elements. The comparison also reveals which phenomena the shell elements describe poorly.

The calibration of the material model was done using simple material tests cut from the aluminum component. The calibration of the yield surface was validated by the good compliance between the results from a simulation and test data for a plain strain tension test. Material failure was modeled with both the Bressan-Williams-Hill (BWH) instability criterion and the Cockcroft & Latham (CL) criterion. The CL criterion gave reasonable results when used in component simulations. Component simulations employing the BWH criterion strongly indicated that the current implementation of the criterion was unsuited for modeling failure inthe components and load situations in this thesis.

The comparative study of speed optimized shell elements and solid elements revealed several weaknesses for the use of fast shell elements. A general trend was that the shell element simulations predicted a conservative load capacity. This is thought to be related to the assumption of plane stress used in most shell element formulations. Difficulties were also shown for the use of shell elements when modeling components where folding dominated the deformation. The predicted folding pattern when using shell elements was shown to be highly dependent on simulation parameters. This was not observed when using solid elements.There were also shown some difficulties for the description of frictionless sliding parallel to the shell plane.These difficulties seemed to be related to the use of reduced integration, and where not present in fully integrated solid elements.