Modelling of Continuous Damage in Composite Structural Components
Keywords
Composite Material Fatigue Strength Stress Cycle Stiffness Reduction Local StiffnessThe use of composite materials for structural components in vehicles imposes new requirements on testing, production and simulation. Combining various materials, fibre topologies and manufacturing methods results in numerous influences that need to be considered in simulation. Siemens PLM Software demonstrates how to produce lightweight components with optimum properties.
The need for lightweight construction methods in vehicle engineering has led to the fact that composite materials are also used for loadbearing structural parts. These are exposed to various strains while in use on the road and often subject to continuous damage. This raises then the issue of their fatigue strength.
Previously, damage modelling for composite materials tended to concentrate on aircraft construction, where long test series are possible and where only a limited number of selected materials and simplified load scenarios need to be considered. The vehicle industry, however, adds additional challenges in the form of complex stresses as well as production processes.
Simulation of Composite Materials
Composite materials impose new demands on testing, production and simulation. The combination of various materials, various fibre topologies and manufacturing methods results in numerous influences that simulations must be taken into account to manufacture components with optimum properties. At the same time, the vehicle industry requires high production volumes ’ a challenge which new methods and tools must meet. [1] provides insight into this.

▸ On a material (microscale) level: the damage mechanism can be, for example, a matrix crack, fibrematrix debonding, fibre breakage, or delamination, Figure 1

▸ At the load level, for example, static load, shock (impact) and fatigue loading.
Various damage mechanisms occur in composite materials. Although such damage may emerge early on in the useful life of the component, complete failure only happens much later. Nevertheless, changes in local stiffness and stress redistribution can develop, which, in turn, may affect the behaviour during the lifetime. Accordingly, different approaches to simulation are required.
Approach for Block Loads ’ the njump Algorithm
Given that local stiffness changes with load, all the stresses and strains occurring for each load step must be recalculated from scratch. However, this is impractical for fatigue loads with up to millions of stress cycles.
Early damage with substantial loss of stiffness is typically followed by an extended phase of stability prior to the failure phase.
One proposed solution was presented by Paepegem [2] in 2001. It is based on a homogenised approach to continuous damage, described in terms of the resulting reduction in stiffness.
Continuous Damage Model
This decrease in stiffness is described phenomenologically; using a continuous damage model and with fatigue damage behaviour in three stages used as the basis for the description.
In the continuous damage model (CDM), damage influences are homogenised at the layer level in the directions dictated by fibre alignment, In other words damage is considered parallel and transverse to the fibre direction as well as shear damage. In principle, energybased approaches are pursued for this purpose. One example is described in [3].
The advantage of these approaches is the ability to describe damage based on different load types (static, shock, fatigue) using the same or compatible models and therefore combined. This also means that many material parameters for the different types of load only need to be determined once.
A further crucial advantage of this modelling approach is the fact that material parameters are used at the layer level and thus need only be determined once. No further tests are necessary when the layup design is changed.
Fatigue Behaviour in Three Stages
Good results have been achieved over the last 15 years for composite materials with woven fibres using this method.
These curves could also be replaced by virtual tests instead of using graphed stiffness reduction measurements. These would calculate the stiffness reduction curves based on pure material characteristics and a precise production simulation. Research projects investigating this issue are underway [4].
Hybrid approaches ’ measuring master data that are then adjusted based on local properties such as voids, moisture etc. ’ are also key options within this process. For shortfibre reinforced components, see [5] as an example.
Computational Process for Recurring Loads
In principle, the effect of any change in stiffness on local stresses must be calculated throughout the model. While this is, in principle, numerically possible, the pure number of stress cycles make it practically infeasible.
It is more logical to take the global effect of stiffness changes into consideration when significant changes occur, while always taking local effects into consideration. A closer look at the special case of block loads (i.e. the repetition of the same load stress cycle) allows the local effect of the stress cycle to be first calculated, then extrapolated using the stiffness falloff curve. This extrapolation also allows an estimate of when a recalculation (a socalled njump) is required at the global level.
Good results have been achieved over the last 15 years for composite materials with woven fibres using this method.
Complex Loads
As already mentioned in the introduction, the load collectives to be investigated in the vehicle industry are complex, with varying amplitude and multiple, nonproportional loads. A simple approach with extrapolation no longer possible here, which is why a typical estimate is often made using SNcurvebased methods and all the aforementioned disadvantages (layupdependent, high test scattering).
In this case there is now an extension to the njump method for complex loads that uses cumulative damage based on hysteresis operator theory [6, 7] for local damage accumulation (local stiffness falloff calculation). These methods can accurately map typical nonlinear accumulation when calculating stiffness reduction, while considering correctly the effects on other damage modes.
One special aspect is the ability to implement any linear and nonlinear damage accumulation methods using this method. An open interface allows research institutes and research departments to add their own extensions.
This methodology was developed alongside extensions to stiffness reduction rules and new methods for determining material curves. Details of the methodology and corresponding test method insights are presented in [7].
Material Tests
As stated above, stiffness reduction curves are required for the various damage modes and directions in the case of remaining stiffness methods. [8] presents special tests for this, based on the bending tests used and described for DUbased composite materials. It should be noted that tests to identify materials are very challenging due to the extreme anisotropy of typical test specimens.
The fibre orientation changes during draping when fibre mats have to adapt to the component geometry.
Damage between the Layers
CDM can also be used for accurate predictions when calculating delamination between layers and taking damage within the layers into account [9, 10].
The complex calculations involved in delamination simulation induce that the focus for fatigue calculation is not on the delamination process itself but on the statement questioning whether delamination occurs. The first tools measuring whether an existing short crack between the layers continues to grow during an operating load collective are undergoing tests. They take into account the damage, or stiffness in the layers, that prompts a change to the intermediate layer load. Further investigations into delamination from the border and the behaviour of larger cracks are what follow, to round off the computational tool.
Integration with Process Simulation
As already indicated in the introduction, knowing the component properties after manufacture is important. The fibre orientation changes during draping when fibre mats have to adapt to the component geometry. This impacts directly on producibility as well as local stiffness. A simulation must also take these processes into consideration.
The production process itself, which does not always elicit perfect composite materials, changes local properties. Here too, simulation can help shape the production process such that better properties can be achieved at critical points than those subject to lower structural loads. This also allows the production process to be improved.
Much remains to be discovered in research and development. This is what makes openness in the process chain so crucial here for a simulation platform [1].
Conclusion and Outlook
Switching from metal to composite materials is far more complex than switching from metal to another material. Retaining a testing and simulation level equivalent to that for metals and treating a composite material as a “black metal” would result in restricting the effect of many of the positive composite material characteristics. The goal of lightweight design would remain elusive. This is why the damage simulation methods and process chains need to be adapted. Openness allowing extensions to the various simulation scales is key, given the considerable potential for development here.
Firstly, the right simulation tools mean many tests can be performed more efficiently. Secondly, production processes and designs can be simulated from an early stage. This allows design and production studies to achieve better and also lighter vehicles in the simulation phase.
References
 [1]Küssner, M. E. N.: Simcenter for the predictive validation of composite materials. Simulation von Composites’ Bereit für Industrie 4.0? Hamburg, NAFEMS DACH, 2016Google Scholar
 [2]v. Paepegem, W.: Fatigue Damage Modelling with the phenomenological residual stiffnes approach. In: Vassilopoulos, A.: Fatigue Life Prediction of Composites and Composite Structures. 2010, pp. 102–138CrossRefGoogle Scholar
 [3]Ladevèze, E.; Le Dantec, P.;: Damage modelling of the elementary ply for laminated composites. Composite Science and Technology, 43,1992, pp. 252–267CrossRefGoogle Scholar
 [4]M3, S., 2014, M3  SIMFlanders. Retrieved from SIMFladers: www.simflanders.be/researchprogram/m3
 [5]Jain, A.; v. Paepegem, W.; Verpoest, I.; Lomov, S.: A feasibility study of the Master SN curve approach for short fiber reinforced composites. I Journal of Fatigue, 2016, pp. 264–274Google Scholar
 [6]Brokate, M.; Sprekels, J.: Hysteresis operators. 1996, New York, SpringerCrossRefGoogle Scholar
 [7]CarrellaPayan, D.; Magneville, B.; Hack, M.; Naito, T.; Urushiyama, Y.; Yamazaki, T.; v. Paepegem, W.: Implementation of fatigue model for unidirectional laminate based on finite element analysis: theory and practice. In: Frattura ed Integrita? Strutturale 38, 2016, pp. 184–190Google Scholar
 [8]v. Paepegem, W.: Development and finite element implementation of a damage. PhD Thesis, 2002, U GentGoogle Scholar
 [9]Bruyneel, M.; Delsemme, P.; Groupil, A.; Jetteur, J.; Lequesne, C.; Naito, T.; Urushiyama, Y.: Damage modeling of laminated composites: validation of the interlaminar damage law in SAMCEF at the coupon level for UD plies. World Congress of Comp Mechanics, 2014, BarcelonaGoogle Scholar
 [10]Bruyneel, M.; Delsemme, P.; Groupil, A.; Jetteur, J.; Lequesne, C.; Naito, T.; Urushiyama, Y.: Damage modeling of laminated composites: validation of the intralaminar damage law in SAMCEF at the coupon level for UD plies. European Conference on Composite Material ECCM16, 2014, SevillaGoogle Scholar