Numerical Simulations of Low- Velocity Impact on Thick ... · Ağır, Impact response of composite...

1
Numerical Simulations of Low- Velocity Impact on Thick Composites Lucas Patrick Thierry Monnin Introduction Fibre reinforced composites are nowadays widely used in many technical applications, such as aerospace, sport, automotive or energy due to their advantageous properties, such as their excellent stiffness/weight ratio, resistance to fatigue and corrosion, etc. Nevertheless, these materials present relatively poor damage resistance, which can have dramatic consequences. Low-velocity impacts may occur during handling, and the damage they could induce and their consequences have to be well understood, especially because they are often difficult to detect with visual inspection. Experimental investigations allow designers to dimension their parts, but these tests are costly in time and resources. A numerical model able to predict accurately the impact induced damage would be a powerful tool for composite structure design The aim of this project is to develop a finite element model to simulate numerically low- velocity impact on thick composite, based on an experimental paper [1]. This model should be able to predict low-velocity induced damage and dynamic response accurately, in order to use this numerical tool to have a better understanding of the experimentally obtained results. This model could then be used to perform analyses that are difficult or impossible to achieve experimentally. Numerical model Two types of reinforcements were used: quadriaxial [0º/-45°/+45°/90º] E-glass fibre fabric with a chopped strand mat (CSM) backing (a) and 2/2 twill carbon fabric (b), with epoxy as matrix. Different thicknesses, impact energy level and configuration were tested experimentally and numerically. [2-3] The hybrids were composed of 10mm thickness carbon twill with one or two added layers of E-glass. Different damage models were tested, for both inter-laminar damage (delamination) and intra-laminar damage (fibre failure and matrix cracking), depending on the type of composite studied. The different methods tested for the inter-laminar damage model, called cohesive zone models (CZM), are presented here. The final choice for the CZM is the zero-thickness cohesive elements with tie constraints. [4] During the experimental tests, the samples are clamped on a solid area. Only the solid area of the figure below is modelled. The impactor, solid area under the composite laminate and pins are treated as rigid bodies. The degrees of freedom of the impactor and the pins are restricted to vertical displacement (out-of-plane) only, while all the degrees of freedom of the solid area are blocked. The mesh size was determined by the damage models. Each of the model has a maximum element length to avoid instabilities and capture the damage accurately. The use of tie-constraints allows to refine the cohesive mesh compare to the one of laminate plies. The cohesive elements are modelled with COH3D8, while the composite ply are modelled with SC8R (continuum shell elements), less sensitive to aspect ratio than solid elements. Results Conclusions In this work, the numerical results from the different models have been compared to decide which model was the most appropriate, and then compared to experimental data. The numerical model used for the carbon twill specimens proved that it was able to reproduce with accuracy the experimental results, through the different impact energy levels and laminate thicknesses. The cross-sectioned delamination showed good agreements with the experiments, even if it was slightly over-estimated in the simulation. Concerning the E-glass, the numerical results were internally consistent, and the evolution of the outputs with respect of the impact energy was reasonable. However, the obtained results differed from the experimental ones, especially regarding the delamination. Improvements were obtained when modelling the CSM layer separately from the quadriaxial, which may be a good track for future modelling strategies. The hybrids were not predicted accurately compared to experiments; it is attributed to the E-glass layers, which in the experiments seemed to concentrate the damage, but this was not observed in the simulations. Numerically, nothing suggests that hybrid composite could be better than monolithic one of the same thickness. EPFL supervisor: Dr. Joel Cugnoni RMIT supervisor: Dr. Raj Das Master degree Thesis March – August 2018 The contacts between the rigid parts and the composite plies is defined with the general contact algorithm of ABAQUS/Explicit. The normal behaviour is considered as a “hard contact”, and a friction coefficient of 0.3 is applied . Regarding to the geometry of the problem, only the quarter of the system is modelled for computational efficiency. References: [1] Pantelidi, M.D., Material Effects On Low Velocity Impact Of Thick Composite Structures. School of Engineering (SoE), Royal Melbourne Institute of Technology (RMIT), Melbourne, Australia. [2] Bektaş, N.B. and İ. Ağır, Impact response of composite plates manufactured with stitch-bonded non-crimp glass fiber fabrics. Science and Engineering of Composite Materials, 2014. 21(1): p. 111-120 [3] fibermaxcomposites. [01.08.2018]; Available from: http://www.fibermaxcomposites.com/shop/index_files/weavingstylesandpatterns.html [4] González, E., et al., Simulation of drop-weight impact and compression after impact tests on composite laminates. Composite Structures, 2012. 94(11): p. 3364-3378. Critical load and damage evolution This figure shows the evolution of the degradation of the cohesive zone at the critical load. This threshold force does not physically coincide with the initiation of damage, but indicates a sudden stiffness loss caused by instantaneous large delamination due to unstable crack propagation. In the numerical model, this load is reached after approximately 28ms, and one can observe that it corresponds to the moment when the delamination reaches the bottom of the laminate, i.e. the inter- laminar damage has propagated through the whole thickness of the composite, which explains the sudden stiffness degradation and thus the change of slope in the force-time response. Carbon /epoxy specimens The numerical model for carbon specimen agrees well with the observed experimental results, regarding the delamination and the force, as well as the evolution of the different outputs with respect to impact energy and laminate thickness. The drop in the force-time response corresponds to the critical load. 0 5 10 15 20 25 30 35 0 0.0005 0.001 0.0015 0.002 Force (kN) Time (s) Carbon 10mm E-glass/epoxy specimens The results obtained for the E-glass specimens are consistent with each other and with the results of the carbon specimens. However, regarding to the experimental results, the delamination from the numerical model for the E-glass specimen is not satisfying. This composite is more complex than the other one, and the numerical model used is maybe too simple to represent accurately the reality. Hybrids specimens Regarding to the experimental samples, it seems that the added E-glass layers concentrates the damage, which seems not to be the case in the numerical model. The numerical model over-estimated the inter-laminar damage compared to experimental cross-sectioned samples. Furthermore, the critical load was under-estimated in the numerical model. 0 5 10 15 20 25 30 0 20 40 60 Peak force (kN) Impact Energy (J) Carbon 10mm Carbon 6mm Carbon 4mm 0 2 4 6 8 10 12 14 16 18 0 1 2 3 4 Force (kN) Displacement (mm) E-glass 9mm Numerical Experimental Numerical Experimental

Transcript of Numerical Simulations of Low- Velocity Impact on Thick ... · Ağır, Impact response of composite...

Page 1: Numerical Simulations of Low- Velocity Impact on Thick ... · Ağır, Impact response of composite plates manufactured with stitch -bonded non -crimp glass fiber fabrics. Science

Numerical Simulations of Low-Velocity Impact on Thick Composites

Lucas Patrick Thierry Monnin

IntroductionFibre reinforced composites are nowadays widely used in many technical applications,such as aerospace, sport, automotive or energy due to their advantageous properties, suchas their excellent stiffness/weight ratio, resistance to fatigue and corrosion, etc.Nevertheless, these materials present relatively poor damage resistance, which can havedramatic consequences.Low-velocity impacts may occur during handling, and the damage they could induce andtheir consequences have to be well understood, especially because they are often difficultto detect with visual inspection.Experimental investigations allow designers to dimension their parts, but these tests arecostly in time and resources. A numerical model able to predict accurately the impactinduced damage would be a powerful tool for composite structure design

The aim of this project is to develop a finite element model to simulate numerically low-velocity impact on thick composite, based on an experimental paper [1]. This modelshould be able to predict low-velocity induced damage and dynamic response accurately,in order to use this numerical tool to have a better understanding of the experimentallyobtained results. This model could then be used to perform analyses that are difficult orimpossible to achieve experimentally.

Numerical modelTwo types of reinforcements were used: quadriaxial [0º/-45°/+45°/90º] E-glass fibrefabric with a chopped strand mat (CSM) backing (a) and 2/2 twill carbon fabric (b),with epoxy as matrix. Different thicknesses, impact energy level and configurationwere tested experimentally and numerically.

[2-3]

The hybrids were composed of 10mm thickness carbon twill with one or two addedlayers of E-glass. Different damage models were tested, for both inter-laminardamage (delamination) and intra-laminar damage (fibre failure and matrix cracking),depending on the type of composite studied. The different methods tested for theinter-laminar damage model, called cohesive zone models (CZM), are presented here.The final choice for the CZM is the zero-thickness cohesive elements with tieconstraints.

[4]

During the experimental tests, the samples are clamped on a solid area. Only the solidarea of the figure below is modelled. The impactor, solid area under the compositelaminate and pins are treated as rigid bodies. The degrees of freedom of the impactorand the pins are restricted to vertical displacement (out-of-plane) only, while all thedegrees of freedom of the solid area are blocked.

The mesh size was determined by the damage models. Each of the model has amaximum element length to avoid instabilities and capture the damage accurately.The use of tie-constraints allows to refine the cohesive mesh compare to the one oflaminate plies. The cohesive elements are modelled with COH3D8, while thecomposite ply are modelled with SC8R (continuum shell elements), less sensitive toaspect ratio than solid elements.

Results

ConclusionsIn this work, the numerical results from the different models have been compared to decidewhich model was the most appropriate, and then compared to experimental data.The numerical model used for the carbon twill specimens proved that it was able toreproduce with accuracy the experimental results, through the different impact energylevels and laminate thicknesses. The cross-sectioned delamination showed goodagreements with the experiments, even if it was slightly over-estimated in the simulation.Concerning the E-glass, the numerical results were internally consistent, and the evolutionof the outputs with respect of the impact energy was reasonable. However, the obtainedresults differed from the experimental ones, especially regarding the delamination.Improvements were obtained when modelling the CSM layer separately from thequadriaxial, which may be a good track for future modelling strategies.The hybrids were not predicted accurately compared to experiments; it is attributed to theE-glass layers, which in the experiments seemed to concentrate the damage, but this wasnot observed in the simulations. Numerically, nothing suggests that hybrid composite couldbe better than monolithic one of the same thickness.

EPFL supervisor:Dr. Joel Cugnoni

RMIT supervisor:Dr. Raj Das

Master degree ThesisMarch – August 2018

The contacts between the rigid parts and thecomposite plies is defined with the general contactalgorithm of ABAQUS/Explicit. The normalbehaviour is considered as a “hard contact”, and afriction coefficient of 0.3 is applied . Regarding tothe geometry of the problem, only the quarter ofthe system is modelled for computationalefficiency.

References:[1] Pantelidi, M.D., Material Effects On Low Velocity Impact Of Thick Composite Structures. School of Engineering (SoE), Royal Melbourne Institute of Technology

(RMIT), Melbourne, Australia. [2] Bektaş, N.B. and İ. Ağır, Impact response of composite plates manufactured with stitch-bonded non-crimp glass fiber fabrics. Science and Engineering of

Composite Materials, 2014. 21(1): p. 111-120[3] fibermaxcomposites. [01.08.2018]; Available from: http://www.fibermaxcomposites.com/shop/index_files/weavingstylesandpatterns.html[4] González, E., et al., Simulation of drop-weight impact and compression after impact tests on composite laminates. Composite Structures, 2012. 94(11): p. 3364-3378.

Critical load and damage evolutionThis figure shows the evolution of the degradationof the cohesive zone at the critical load. Thisthreshold force does not physically coincide withthe initiation of damage, but indicates a suddenstiffness loss caused by instantaneous largedelamination due to unstable crack propagation.In the numerical model, this load is reached afterapproximately 28ms, and one can observe that itcorresponds to the moment when the delaminationreaches the bottom of the laminate, i.e. the inter-laminar damage has propagated through the wholethickness of the composite, which explains thesudden stiffness degradation and thus the changeof slope in the force-time response.

Carbon /epoxy specimensThe numerical model for carbon specimenagrees well with the observed experimentalresults, regarding the delamination and theforce, as well as the evolution of thedifferent outputs with respect to impactenergy and laminate thickness. The drop inthe force-time response corresponds to thecritical load.

0

5

10

15

20

25

30

35

0 0.0005 0.001 0.0015 0.002

Forc

e (k

N)

Time (s)

Carbon 10mm

E-glass/epoxy specimensThe results obtained for the E-glassspecimens are consistent with each otherand with the results of the carbonspecimens. However, regarding to theexperimental results, the delaminationfrom the numerical model for the E-glassspecimen is not satisfying. This compositeis more complex than the other one, andthe numerical model used is maybe toosimple to represent accurately the reality.

Hybrids specimensRegarding to the experimental samples, it seems that the added E-glass layers concentratesthe damage, which seems not to be the case in the numerical model. The numerical modelover-estimated the inter-laminar damage compared to experimental cross-sectionedsamples. Furthermore, the critical load was under-estimated in the numerical model.

0

5

10

15

20

25

30

0 20 40 60

Pea

k fo

rce

(kN

)

Impact Energy (J)

Carbon 10mmCarbon 6mmCarbon 4mm

0

2

4

6

8

10

12

14

16

18

0 1 2 3 4

Forc

e (k

N)

Displacement (mm)

E-glass 9mm

Numerical Experimental

Numerical Experimental