Carbon fibre reinforced plastic energy absorbing structures under crash loads : Numerical simulations validated with experimental tests

Sammanfattning: The development of a numerical modelling approach for carbon bre reinforced plastic energy absorbing structures designed for crash events using Abaqus/CZone is described within this master thesis. Several crash tube series have been designed, manufactured, and tested with a rather unconventional cross-sectional geometry. The squared cross-sectional geometry consisted of "curved" at sections, and double flanges for adhering two halves together. The crash tube halves contain carbon bre epoxy UD laminates, and are manufactured using a hot-press machine. Adjustments for experimental tests were made in the geometry, laminate denition, and impact velocity. Numerical simulations were focused on geometry, laminate definition, impact velocity, flange geometry, material model, laminate thickness, and crush properties. The numerical model consisted of two parts, namely a 3D discrete rigid planar shell as impactor plate without material properties, and a 3D shell crash tube with an imported geometry from Dassault Systems CATIA V5. Material, and cohesive properties were assigned to the crash tube using the Abaqus Ply Fabric material model, and CZone. Mesh seed length was smaller than the critical mesh seed length. An initial clearance between the impactor and crash tube has been implemented within the assembly. Step size was set to 0.05 s, and dynamic explicit step type was selected. General contact was defined using default settings. A qualitative good agreement between numerical and experimental test results is achieved for V100, V500, V600, and V700 series with c.o.v. values for stroke length of 2.2%, 7.0%, 5.3%, and 4.1%,respectively. V500-V700 series are only tested once, whilst the V100 series has been tested five times. The V200 and V300 series with modified geometries achieved c.o.v. values within a complete different order of magnitude for stroke length, namely 14% for V200 series, and 15% for V300 series. Although only three specimens have been tested for the V200, and V300 series, it is not the main reason for this mismatch. The starting deceleration at the linear increasing segment is much lower than observed in experiments, and causes this major difference. Elevated impact velocities for the V1502, and V1503 series tend to have a positive influence on the numerical results, and ensure a higher initial starting deceleration at the linear increasing segment. Numerical results show that increased taper ratios result in decreased decelerations. A flange geometry study showed that implementing a double flange suppresses debonding, and increases the deceleration, as the geometrical stiffness is increased. Not having a flange results in lower decelerations, and larger stroke lengths, as the geometrical cross-section provides less stiffness. A bonded single flange will be debonded, independent of bonding properties, as even extremely high bonding properties have shown direct debonding under impact load. Stiffer laminates result in higher decelerations and shorter stroke lengths. This statement is tested in three different ways, namely by changing the laminate denitions, laminate thickness's, and the crush properties. Three different material models have been tested. Abaqus Ply Fabric has proven to be easiest in usage, and showing a qualitative good agreement with the experimental results. Abaqus Ply Fabric does over predict the stroke length, whereas Hashin, and Tsai-Wu achieve a more accurate stroke length prediction. However, it is safer to over predict the stroke length. No material model is capable of capturing the initial peek decelerations.