< Back to previous page

Publication

Material Variability across the Scales in Unidirectional Composites - Virtual Material Characterisation under Longitudinal Tension

Book - Dissertation

Their high stiffness, strength and low density make composite materials an excellent choice for lightweight structural components. However, the high performance of these materials comes at the cost of complex mechanical characterisation, certification and design. For this reason, composite components are often designed applying large safety factors, limiting their lightweight potential. Virtual testing is a cheaper and faster alternative to physical testing, and therefore it can be used to predict material properties and their variability inexpensively. Nevertheless, virtual testing presents several challenges. The complex mechanisms at the microscale, such as the fibre break development occurring during failure, need to be simulated (e.g. by means of fibre break models) to reproduce failure correctly. Moreover, computational time becomes critical when modelling large volumes of material while considering the variability of the microstructure. Fibre break development and material variability can be considered at the same time by incorporating fibre break models into a multiscale framework. In this thesis, a virtual testing methodology (based on multiscale modelling) is developed to predict longitudinal tensile strength in unidirectional carbon fibre composites considering material variability. Three sources of variability are considered: fibre strength variability, local fibre volume fraction variability and local fibre misalignment. In the first part of the work, a finite element model of a tensile coupon (representative of a unidirectional ply) is developed to predict longitudinal tensile strength considering material variability. To accelerate the simulation process, a computationally efficient strategy is developed. A linear regression model, inspired by machine learning, is used to predict the stress field in the coupon, based on the experience built with a small number of FE simulations. This strategy proves very effective in reaching the same results of the full FE model, while reducing the computational time by a factor of 60. The virtual testing method is extended to a multidirectional composite ring loaded with internal pressure, representative of the cylindrical section of a composite pressure vessel. Material variability parameters used in the simulations were measured from a real composite pressure vessel, produced by filament winding. Including material variability in the model lowered the burst strength predictions, showing the importance of variability on the performance of the component.
Publication year:2020
Accessibility:Closed