Experimental Characterization and Numerical Modelling of the Energy Absorption Capacity of UD-NCF Carbon Fibre/Epoxy Composite Channels

Abstract

While the potential of carbon fibre-reinforced plastic (CFRP) composites for energy-absorbing structures has been widely acknowledged, there has been limited adoption in the automotive industry due to high manufacturing costs and challenges associated with accurately predicting their response during impact loading. Components manufactured via high-pressure resin transfer (HP-RTM) with highly reactive resins offer reduced cycle times, while unidirectional non-crimp fabric (UD-NCF) reinforcements provide a reduction in manufacturing costs, high in-plane mechanical properties, and design flexibility. However, the energy absorption capacity (EAC) of UD-NCF CFRP composites has not been widely studied, with much of the available research focused on the quasi-static axial crushing or dynamic drop-testing of unidirectional tape or woven CFRP composite tubes. Additionally, since predictive modelling is extensively used in the automotive industry, additional research on accurately modelling the EAC of CFRP composite components is required to expand their implementation in vehicle structures. Thus, the goal of this study is to support the development of a high-fidelity impact simulation model for predicting the EAC of UD-NCF carbon fibre-reinforced epoxy composite structures manufactured via HP-RTM. Firstly, the EAC and failure modes of [0/±45/90]s, [0/90/±45]s, and [±45/02]s tapered hat channels subject to dynamic axial compressive loading were experimentally characterized. Distinct modes of failure were observed for the tested channels, with the highest energy absorption coinciding with a splaying failure mode for the [±45/02]s channels (24% to 29% higher than other layups). These results were compared to results from previously performed tests on single corrugated and hat channels comprising the same material system and tested under the same conditions. While the stacking sequence is influential for a given component geometry, manufacturing-induced defects for components with complex geometry can cause significant reductions in EAC. For all geometries, the [±45/02]s channels performed best, while the tapered hat channel exhibited up to 25% lower EAC compared to the hat and corrugated channels for this stacking sequence. Next, simulation models were developed using the finite element software LS-DYNA to predict EAC of the tapered, hat, and corrugated CFRP composite channels with different stacking sequences. Various loading conditions were simulated, including dynamic and quasi-static axial compressive loading and three-point bending. An available material model for the progressive failure of composite laminates, MAT_054, was calibrated for a baseline case, namely the [0/±45/90]s hat channel under dynamic axial compressive loading. It was concluded that while the EAC of channels with different geometries from the baseline condition was accurately predicted when using the calibrated material model, the predictive capability of the simulation models was limited for other stacking sequences and loading rates and for the three-point bending cases. The calibrated material model was unable to capture the strain rate-dependent response of the UD-NCF composite material, which resulted in underprediction of EAC at loading rates different from the baseline condition (i.e. 33% for the [0/±45/90]s hat channel subject to quasi-static axial loading). Additionally, the single-shell part representation used in the study prevented capture of delamination, which resulted in underpredictions for energy absorption when delamination was prominent (i.e. 24% for [±45/02]s hat channels subjected to dynamic axial loading). Lastly, the EAC was significantly overpredicted for the three-point bending cases (from 13% for the [±45/02]s corrugated channels up to 60% for [±45/02]s hat channels subject to dynamic three-point bending) owing to the fact that some of the material model parameters were non-physical and thus tailored for the axial compression loading case. In general, changes in stacking sequence and loading condition required recalibration of the non-physical material model parameters; most notably, the mode-dependent failure strains. Lastly, the energy-absorption capacity and failure behaviour of adhesively bonded double hat channels under axial compressive loading were experimentally characterized and corresponding simulation models were developed with the calibrated material model. The average total energy absorption was higher for the [±45/02]s channels when compared to the [0/±45/90]s channels with a 7.5% and 32% increase for quasi-static and dynamic cases, respectively. However, the average specific energy absorption of all double-hat channels tested was less than that measured for the single-hat channel counterpart of the same stacking sequence for both loading rates. To represent the bond numerically, the elastoplastic rate-dependent material model, MAT_240, was calibrated in LS_DYNA to experimental data and applied to cohesive elements. Predictions of EAC were more agreeable with experimental data for the dynamic axial loading cases (deviations of +11% and -21% for the [0/±45/90]s and [±45/02]s channels respectively) than the quasi-static axial loadings (deviations of -33% and -35% for the [0/±45/90]s and [±45/02]s channels respectively). The study yielded critical performance data for UD-NCF CFRP channels and an improved understanding of the associated influence of stacking sequence, component geometry, and loading rate. The simulation results serve as a benchmark to which more physically based models (i.e., considering interlaminar interactions, strain rate dependencies, etc.) can be confidently compared.