In English

Homogenisation of embossed aluminium sheets using virtual material testing

Sofia Alexandersson ; Filip Ljungström
Göteborg : Chalmers tekniska högskola, 2019. 71 s.
[Examensarbete på avancerad nivå]

As the demand for lightweight constructions increases, thin structured sheet metals become more and more popular in industrial applications. As a part of Volvo Cars' strive to lighten their vehicles, embossing is used to add sti ness to aluminium heat shields while keeping the mass unchanged. When analysing an embossed structural component using the nite element method, a very ne mesh is required to capture the geometry, meaning that a simulation of a fully resolved heat shield is computationally expensive. To enable analysis of the embossed structure, a virtual material model is developed in this thesis using homogenisation. This model describes the averaged macroscopic behaviour and enables the embossed sheet to be modelled as a at structure. The applied homogenisation method used to characterise the e ective tensile and bending behaviours was virtual material testing. A representative volume element was established to model the embossed structure in nite element analyses with both Dirichlet and periodic boundary conditions. The virtual material model was veri ed using physical test data and simulation outputs from fully resolved structures. The results showed that embossed sheets are sti er in bending and weaker in tension, compared to at structures with the same sheet thickness. Geometrical factors that in uenced the behaviour were sheet thickness, pattern orientation, wavelength and amplitude. Comparisons between the virtual material model on at sheets and fully resolved embossed models revealed that the homogenisation procedure predicts good tensile representation but a slightly too exible bending behaviour. However, when validating the model with physical test data, the results showed a too sti response from the virtual material model. Since the model indicated good resemblance with the fully resolved nite element model, it was concluded that these inaccuracies probably were due to the neglect of some sheet characteristics in the modelling of the structure. It was realised that the real thicknesses of the embossed sheets were lower than the nominal thicknesses speci ed by the manufacturers. By adjusting these values, the virtual material model showed better resemblance in bending, although it still predicted slightly too high natural frequencies. Possible error sources could for instance be residual stresses in the structure from the manufacturing process and a varying thickness throughout the pattern.

Nyckelord: Homogenisation, virtual material testing, embossed sheet, nite element method, representative volume element, orthotrophy, periodic boundary conditions, structured sheet

Publikationen registrerades 2019-01-08.

CPL ID: 256434

Detta är en tjänst från Chalmers bibliotek