Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Modelling Damage in Composite Materials

Background

The advantages of using composite materials in a wide range of industrial applications (e.g. aerospace, automotive, offshore structures and shipbuilding) have been recognised for many years. These include: (a) reduced weight; (b) better corrosion resistance; (c) lower whole life cycle costs; (d) no hot work required for retrofitting, and (e) improved thermal, acoustic and vibration properties.

Traditionally, the investigation of damage in composites has been addressed experimentally, requiring extensive testing programmes in order to deal with the inevitable variability of the data and to ensure a robust statistical analysis. In order to reach an optimum balance between cost and lead-time in Programme Development Plans (PDPs), the use of numerical modelling has been encouraged to enable predictive virtual testing.

As part of Core Research Programme (CRP) activities and aligned European-funded projects, TWI has investigated the feasibility of using modelling tools to study damage initiation and propagation in composite materials.

Objectives

  • Review and evaluate existing techniques for damage modelling of composites.
  • Understand the material properties obtained from standardised experimental tests which are needed to implement damage models.
  • Verify and validate a selected modelling approach by comparison with benchmark samples.

Solution

The study addressed two different aspects of modelling damage in composite materials; damage initiation and damage evolution.

The modelling study was applied to an aerospace grade of carbon fibre reinforced polymer, T700-M21. Relevant orthotropic material properties as well as fracture mechanics properties were obtained via experimental tests according to ASTM standards.

During the first phase of the study, damage initiation and evolution were studied using continuum damage mechanics (CDM). Hashin’s and Puck’s criteria were selected as suitable damage criteria, as they provide information about failure modes and the location of damage. These criteria were implemented in models and their performance was assessed against experimental results generated from static four point bending tests in a TWI-led Clean Sky European project, HEGEL (Grant agreement ID: 738130). A sensitivity study was undertaken with respect to the modelling approach for the lay-up sequence, mesh element type and mesh element size. It was observed that stress levels causing failure in the laminate could be predicted with an acceptable margin of error (within 5%) – Figure 1. The models were also successful in representing the main mechanism of failure observed experimentally, i.e. failure of the matrix in tension in the outermost ply (Figure 2). 

Figure 1. Comparison of constant strain rate (CSR) simulated data with experimental CSR data
Figure 1. Comparison of constant strain rate (CSR) simulated data with experimental CSR data
Figure 2. Representation of damage initiation using a CDM-based model: The damage occurs by matrix failure in tension at the outermost 90° ply
Figure 2. Representation of damage initiation using a CDM-based model: The damage occurs by matrix failure in tension at the outermost 90° ply
Figure 3. Inclusion of inherent defects (in red) in a four-point bending specimen model. The specimen under consideration is a 80mm x 13mm laminate, with [0/90/0/90/0]s stacking sequence. A python script was developed capable to introduce defects within the laminate, based on experimental statistical data from computed tomography scan of the defects
Figure 3. Inclusion of inherent defects (in red) in a four-point bending specimen model. The specimen under consideration is a 80mm x 13mm laminate, with [0/90/0/90/0]s stacking sequence. A python script was developed capable to introduce defects within the laminate, based on experimental statistical data from computed tomography scan of the defects

Whilst a CDM approach uses damage variables to represent a typical crack density in a composite, damage initiation conditions will predominantly be controlled by material strength, and lay-up sequence. As a result, it was noted that, if the minimum stress levels required to achieve such conditions were not reached, damage initiation would not occur and, as a result, damage accumulation could not take place. In real materials, damage initiation usually nucleates from pre-existing flaws (e.g. manufacturing defects) within the laminate and, as a result, the stress level required to initiate damage can be much lower compared with that required in ideal laminates. This indicates that capturing the realistic distribution and morphology of pre-existing flaws is crucial for acceptable prediction of damage in composite parts. This challenge was addressed by identifying and implementing a phenomenological approach to damage modelling. This required the development of a modelling strategy able to include “inherent” manufacturing defects on the basis of relevant statistical data from experimental observations, i.e. void volume fraction, morphology, dimensions, location and orientation (Figure 3). Fracture mechanics based models, using the Virtual Crack Closure Technique (VCCT) or extended FEM (XFEM), were subsequently implemented. The validity of the approach was demonstrated against experimental data from flexural four-point bending fatigue testing generated during the HEGEL project (Figure 4). The use of the proposed phenomenological approach enabled a parametric study of damage behaviour in composites with respect to size and percentage of defects within the part.

Conclusion

Predictive modelling approaches for damage initiation and damage evolution were evaluated. The study has shown the promising value of using virtual testing to support the analysis of composite parts, either during the design phase or during the integrity assessment of in-service structures. In addition to conventional approaches using readily available commercial finite element software capabilities, TWI has developed a phenomenological approach to the study of damage in composites. This involves a framework of experimental and modelling activities, with the main aim of achieving a closer representation of the behaviour of real materials.

Figure 4. Example of fatigue master curves generated as a function of void volume fraction and applied stress. The experimental data points for composite laminates with a void fraction ranging between 1% and 2% are observed to be in between the simulated master curves obtained from FEA models containing 1% and 3% void fraction
Figure 4. Example of fatigue master curves generated as a function of void volume fraction and applied stress. The experimental data points for composite laminates with a void fraction ranging between 1% and 2% are observed to be in between the simulated master curves obtained from FEA models containing 1% and 3% void fraction
Avatar Damaso De Bono Principal Project Leader, Numerical Modelling and Optimisation

Damaso has worked at TWI since 2010. As Principal Project Leader in the Numerical Modelling and Optimisation section, he manages projects for TWI’s members in finite element analysis (FEA) and numerical modelling for fitness-for-service (FFS) and engineering critical assessments (ECAs). Damaso also has a background in inverse analysis methodologies using finite element methods and experimental characterisation of engineering surfaces. He obtained a Master’s degree in Aerospace Engineering and completed an Engineering Doctorate with the University of Surrey, in the field of material structure influence on inverse analysis of nanoindentation testing; correlating nano-scale to macro-scale properties of materials.

}