Thin ply composites: Experimental characterization and modeling ICCM19 Montréal 2013 Robin Amacher, Joël Cugnoni, John Botsis Ecole polytechnique fédérale de Lausanne , Switzerland In partnership with North-TPT, FHNW, RUAG Technology, RUAG space, and Connova

Why Thin Ply composites? Thin ply : below 125 g/m2, down to 15g/m2 today Advantages: Improved delamination resistance, higher onset of damage & ultimate failure Improved fatigue properties and in some case damage tolerance More design degrees of freedom = more optimal laminate Easier to design ply drops / small angle laminates / local reinforcements Can produce “homogeneous” laminates, no more dependency on stacking Can produce mixed “thin & thick” laminates for structures with large shell thickness variations. Thin ply materials are now commercially available Challenges: Characterization and understanding of ply-thickness effect for a wide range of constituents Efficient and accurate models for prediction of thin ply composite performance Flexible manufacturing (assembly of complex preform) & automation to reduce layup time Develop & validate efficient design method to account for the performance benefit without blowing up the degree of complexity of the problem Relevant literature (small subset): [1] S. Sihn, R.Y. Kim, K. Kawabe, S. Tsai, Experimental studies of thin-ply laminated composites, Composites Science and Technology, 67, 2007 [2] M.R. Wisnom, B. Kahn, S.R. Hallet, Size effects in unnotched tensile strength of unidirectional and quasi-isotropic carbon/epoxy composites, Composite Structures, 84, 2008 [3] A. Arteiro, G. Catalanotti, J. Xavier, P.P. Camanho, Notched response of non-crimp fabric thin-ply laminates, Composites Science and Technology, 79, 2013 Intro Lamina level Laminate level Element level Let’s have a look at the main contributions to the subject in the littérature: Wisnom described size effects in a geometrical scaling with cst ply thickness, observing an increase in strength with the number of sublaminate repetitions Tsai showed that thin-ply laminated composites can supress the microcracking and delamination. The comparison was btw thin plies and a ply-block stacking of thin plies Camanho showed the same trends for non-crimp fabrics, with also an improved behaviour in mechanically fastened joints . The comparison was thin plies vs … Simulation Design Conclusion

Objectives & Method Objectives: Understand, characterize and model thin ply composites from the ply level to part level. Experimental study of thin ply size effects Constant specimen thickness UD prepregs of varying ply thickness & number of sub-laminate repetitions Material North TPT Thin ply composites UD prepreg with ATL production of complex laminates Thick = 300 (2x150) g/m2 ~300 microns / ply Intermediate = 100 g/m2 ~100 microns / ply Thin = 30 g/m2 ~30 microns / ply M40JB fiber / NTPT TP80ep (80°C epoxy resin) from the same batches and production on the same machine the same week. Autoclave production, 55% fiber volume fraction Intro Lamina level Laminate level Element level What we have done is to take a design approach keeping the thickness of the speciment constant and varying the ply wheigths and thus the number of repetitions. I will first present you the major experimental results. Here I must make a small parenthesis about our main industrial partner, which is North-TPT: They are able to process UD pregs between 200 and 16g/m2! As it is very difficult to handle such light plies, a robot taper was dev. to lay up complexes. This enables a design philosophy based on the assembly of custom complexes, whit a dedicated sftware to handle it. So this has enabled us to process 3 ply weights to compare. We have also minimized the scatter factors by … Simulation Design Conclusion

ThinPly composites: design performance advantages Click one of the icons below to learn more about Thin Ply composites performance in a specific area Lamina & Laminate Open Hole & Fatigue Bolted joint & Ageing Part Design & Analysis ? Impact Damage Resistance … or just continue reading normally to have a full overview

Experimental characterization of Thin-ply size effects

UD ply level properties Overall no change in intrinsic lamina properties when reducing ply thickness One exception: 0° compression Thick 300 g/m2 Intro Compressive strength (ASTM D5467*) Lamina level Intermediate 100g/m2 Laminate level Element level On the lamina or UD level, which is our baseline, there is as expected no change on tensile and shear properties. There is however a tendency towards higher compressive strength for thin. We believe this is due to a more uniform microstructure Simulation Thin 30g/m2 Design Conclusion Thin Ply : more uniform microstructure and improved 0° compressive strength

Quasi isotropic laminate, tensile ASTM D3039, constant thickness, sub-laminate scaling: [+45°/90°/-45°/0°]ns with n=1 for thick, n=3 for interm. and n=10 for thin ply. Intro Lamina level Laminate level Applied stress Element level Let’s move to the laminate level. For thick plies, the stacking is the minimal balanced and symetric quasi-iso, with only one repetition. The graph shows on the vertical axis the stress, and on the horizontal axis the cumulative ac en, which is used to monitor the damage. You can see that the onset of d is below half of the ultimate str. If thin plies are grouped in a ply-block fashion, the result is similar Using sublaminate scaling with a repetition of 3 or even 10, the result is completely different. The ult str is higher and the onset of damage is just below, which means above the utlimate strength for thick. Simulation Design Conclusion Damage (Acoustic emission)

QISO tensile properties Ultimate strength: +42% Intro Onset of damage: +227% Lamina level n=10 & n=3 n=3 Laminate level Change of failure mode: Thick: extensive matrix cracking & delamination Thin: brittle rupture by fiber failure (max strain of fiber) Little effect of n n=1 Element level On this graph, the strength is plotted in function of the ply thickness. The solid curve correspond to the failure and the dotted one to the onset of damage. Going from n=1 to n=3, there is a 40% increase on the ult str, and then a plateau. The onset seems to have a quasi linear behavior between n=1 and n=10, with more than 200% increase! For a first-ply failure design approach, this means huge possible gains. This is linked to a radical change in failure mode from multimode with strong delamination for thick to quasi-brittle for thin. This transition is really a key to understanding the possible benefit of thin plies. Simulation Design Conclusion

Open Hole Compression Open Hole Compression [+45°/90°/-45°/0°]ns (ISO 14126 / ASTM D6484) +18% Intro Lamina level Laminate level OHC Strength [Mpa] Element level TODO: insert table here; conclusion: strong improvement of most properties, delamination and damage in off axis plies are nearly suppressed; Much improved hot wet properties as weakening of the matrix is less critical in thin plies (still no delamination) impact tests showed very different features => changing ply thickness allows tayloring which damage modes are active. Checklist Simulation Design

Open Hole Tensile fatigue Open Hole Tensile: static and fatigue (ASTM D5766 & D7615, R=0.1) [+45°/90°/-45°/0°]ns Intro -34% Lamina level Laminate level +31% Thick plies 300g/m2, n=1 @12k cycles, 316MPa Thin plies 30g/m2, n=10 @1M, 316MPa Element level We now move to the element with open hole tensile test. This graph shows the norm far field stress on the vertical axis in function of the number of cycles on the horizontal axis. On the left, we have the results for the quasi-static test. As there is no damage to release the stress concentration around the hole, the ultimate strength for thin plies is 34% lower. However, the onset of damage is still 31% higher, meaning again more strength bearing capacity in a first ply failure design. This is confirmed by fatigue testing. For 316MPa, which correspond to 70% of its ultimate str, the thick fails before 20’000 cycles. Thin plies laminate, for the same load, which correspond to 85% of its ult str, has not failed after 1M cyles Simulation Design Ruin = -10% stiffness Lower static ultimate strength. No damage around hole means no stress concentration relief but better predictability (Wisnom & al) Conclusion Strong improvement in fatigue life (<20k vs >1M cycles)

Bolted joint bearing strength Single lap bearing test, standard and Hot Wet condition (ASTM 5961), fastener type EN-6115 Hot Wet cond. 95%RH/70°C, test 90°C [+45°/90°/-45°/0°]ns Intro sbr_ult = 156 MPa sbr_ult = 476 MPa sbr_ult = 573 MPa sbr_ult = 294 MPa sbr_ult = 584 MPa sbr_ult = 372 MPa Thick Ply 300g/m2, n=2 Intermediate 100g/m2, n=5 Thin Ply 30g/m2, n=18 Hot Wet 90°C As Produced, 20°C Lamina level Laminate level Element level The next logical step is to test bearing strength. The test is single lap, and was performed in standard and hot wet conditions. Thin plies provide better properties in both conditions, especially for hot wet as weakening of the matrix is less critical in thin plies (still no delamination) Simulation As produced, 20°C Design Conclusion Strength improvement for as produced @ 20°C  +18% Strength improvement for Hot Wet @ 90°C  +58%

Low energy impact Rectangular specimen clamped on the short sides; bending is dominant QI [0°/+45°/90°/-45°]ns, 300 x 140 x 2.4 mm Thick (300 g/m2) n=1, Intermediate (100g/m2) n=3, Thin (30g/m2) n=10 Energy: 11.5 J & 18J Intro Backside Lamina level Laminate level Element level THICK, n=1 INTERMEDIATE, n=3 THIN, n=10 impact tests showed very different features => changing ply thickness allows tayloring which damage modes are active. Summary Transition of failure mode from delamination to fiber failure An optimal ply thickness can be found to achieve the smallest damage area Thin-Ply technology allows tailoring the material properties wrt impact induced damage Simulation Design mostly fiber failure delamination & fiber failure delamination

Modeling of Thin-ply size effects

Modeling thin-ply size effects Ply thickness effects: reduction of interlaminar shear stresses at the free edges & delayed free edge delamination Constrained intra laminar transverse cracks => apparent 90° and in-plane shear strength increase => in-situ ply strength model (*) Matrix cracking induced delamination also delayed by constraining plies Other mechanisms (bridging?, plasticity of matrix?) Intro Lamina level 2D generalized plane strain or 3D continuum model Laminate level Element level What we have done is to take a design approach keeping the thickness of the speciment constant and varying the ply wheigths and thus the number of repetitions. I will first present you the major experimental results. Here I must make a small parenthesis about our main industrial partner, which is North-TPT: They are able to process UD pregs between 200 and 16g/m2! As it is very difficult to handle such light plies, a robot taper was dev. to lay up complexes. This enables a design philosophy based on the assembly of custom complexes, whit a dedicated sftware to handle it. So this has enabled us to process 3 ply weights to compare. We have also minimized the scatter factors by … In-situ shear / normal strength (*) + coarse 3D model Or high fidelity 3D continuum model + cohesive zone Simulation Design Conclusion High fidelity 3D continuum model + cohesive zone Fracture mechanics : onset ~ 1/sqrt(t) (*) Camanho et al., Composites Part A, 37, 2006

Simulation of ‘thin ply’ effects Goal: capture the transition in dominant failure mode in order to understand and predict ply size effects with interacting damage modes Hypotheses: no change in intrinsic properties of ply and interface wrt ply thickness Intro Lamina level 4th ply: +45° 3rd ply: 90° 2nd ply: -45° First ply: 0° (symetry) Cohesive elements > lateral cracking User material with fiber failure (subroutine) Laminate level Element level We now move to the simulation, which is a work in progress. Our goal is to… The hypothsis we make is that there is… What we do is a 3D explicit modeling of the QI tensile test in Abaqus Each layer is modeled with a central zone taking the fiber damage into account, side zones without fiber failure to avoid failure at the grips, and cohesive elements along the fiber direction to account for transverse cracking. There are also cohesive surfaces btw the layers to account for delamination. The loading is force controlled Simulation Between the layers: cohesive surfaces > delamination UD mx. without fiber failure Simulation: force controlled (sigmoid ramp, quasi-static) Design Conclusion High fidelity 3D modeling of quasi isotropic unnotched tensile test in Abaqus Explicit, [45°/90°/-45°/0°]ns laminate. All material properties coming from tests, no fitting parameters.

Simulation of ‘thin ply’ effects Goal: capture the transition in dominant failure mode in order to understand and predict ply size effects with interacting damage modes Hypotheses: no change in intrinsic properties of ply and interface wrt ply thickness Intro Lamina level 2nd ply: -45° 3rd ply: 90° Cohesive elements => Intralaminar matrix cracking First ply: 0° (symetry) User material with fiber failure (subroutine) Cohesive elements > lateral cracking User material with fiber failure (subroutine) Laminate level Element level 45° We now move to the simulation, which is a work in progress. Our goal is to… The hypothsis we make is that there is… What we do is a 3D explicit modeling of the QI tensile test in Abaqus Each layer is modeled with a central zone taking the fiber damage into account, side zones without fiber failure to avoid failure at the grips, and cohesive elements along the fiber direction to account for transverse cracking. There are also cohesive surfaces btw the layers to account for delamination. The loading is force controlled 90° Simulation -45° Symmetry BC 0° Between the layers: cohesive surfaces => delamination UD mx. without fiber failure Simulation: force controlled (sigmoid ramp, quasi-static) Design Conclusion High fidelity 3D modeling of quasi isotropic unnotched tensile test in Abaqus Explicit, [45°/90°/-45°/0°]ns laminate. All material properties coming from tests, no fitting parameters.

Simulation of ‘thin ply’ effects Damage models: cohesive interfaces between plies, cohesive elements for transverse cracking, continuum damage model for fiber failure Mesh convergence study: 6 linear hex. element C3D8R per ply thickness, 0.5mm in plane elem. size, up to 650k elements, 2.1M dofs Mass scaling & time step convergence study: dt = ~5e-6 s, ~500’000 time steps Intro Lamina level Laminate level Element level We now move to the simulation, which is a work in progress. Our goal is to… The hypothsis we make is that there is… What we do is a 3D explicit modeling of the QI tensile test in Abaqus Each layer is modeled with a central zone taking the fiber damage into account, side zones without fiber failure to avoid failure at the grips, and cohesive elements along the fiber direction to account for transverse cracking. There are also cohesive surfaces btw the layers to account for delamination. The loading is force controlled Simulation Design Conclusion

Thick ply (300 g/m2) QISO tensile Intro Lamina level Laminate level Element level I will show you a quick movie of the results for thick. … description We observe that normal stress within the 90° and then +/-45° plies damage the matrix and open cracks starting from the edges along with delamination. Eventually, the 0° fibers fail. Simulation Design Conclusion Intra laminar cracking (matrix failure) Delamination damage Fiber failure

Thick ply (300 g/m2) QISO tensile Intro Lamina level Laminate level Element level I will show you a quick movie of the results for thick. … description We observe that normal stress within the 90° and then +/-45° plies damage the matrix and open cracks starting from the edges along with delamination. Eventually, the 0° fibers fail. Simulation Design Damage sequence: cracking of 90° ply, then cracking of 45° plies & delamination from edges Final failure after extensive delamination / damage of all off-axis plies Ultimate strength: fiber failure in 0° plies with all other plies broken Conclusion

Intermediate ply (100 g/m2), n=3 Intro Lamina level Laminate level Element level I will show you a quick movie of the results for thick. … description We observe that normal stress within the 90° and then +/-45° plies damage the matrix and open cracks starting from the edges along with delamination. Eventually, the 0° fibers fail. Simulation Design Damage sequence: cracking of 90° plies then outer 45° ply then final failure (localization of damage starting from the free edges and fiber failure in 0° plies Transverse normal and shear cracking delayed, Delamination nearly suppressed Conclusion

Simulation of ‘thin ply’ effects Intro Lamina level Laminate level Element level Let’scompare this to what is observed experimentally: on the left we have the c-scans done at two stages, and in the middle the numeric results for thick: the pattern is much alike. The intermediate on the right shows some localized damages inherent to the dynamic nature of the loading, but these don’t grow. The failure happen at a localized stress concentration at the boundary btw the zones with and without fiber failure and we have still to work on it, but the trend is here. Simulation Design Conclusion

Simulation of ‘thin ply’ effects FE Analysis Intermediate 100 g/m2 Thick 300 g/m2 Here are plots of the stress against an expression of the damage: on the left the experimental data with the cum ac en, on the right the numerical data with damage dissipation energy. The trends are the same, and we have a good agreement for both onset and ultimate strength for thick. Challenge: deal with the computation time increase for thinner plies. Intermediate = 2w -> thin = ? When this is done, we will turn to other cases with a parametric study Good predictions for thick & intermediate ply thickness for both onset of damage and ultimate strength. High fidelity FE model can capture the change of damage mode sequence wrt ply thickness (with constant intrinsic properties) Coarse 3D FE Models (1 elem/ply) with in-situ strength (scaled by 1/sqrt(t)) provide comparable results, could be extended towards shell modeling (work in progress)

Towards part level modeling and design High fidelity 3D FE modeling All damage modes with interactions Intro Computation time Delamination follows intra laminar cracking (no complex interaction) Lamina level Coarse 3D FE modeling with in-situ strength Separate damage models for transverse cracking and delamination Laminate level Element level Ply thickness small enough to avoid delamination So we may capture ply size effects in detailed 3D modeling, but how do we translate it into the design? We will have to use the detailed models to feed simpler models, for example shell models, but for that we need to develop a strategy, particularly regarding free-edge crack stability. An interesting thing is that if we run simple calculations according to classical laminate theory with and without taking the damage into account, we find that thick plies correspond well to the model with damage, and that thin plies correspond well to the model without damage, even not bothering about first ply failure. This means a much simpler approach! So we would benefit of better properties but also of a simpler design process. For the next steps, we will perform test on tubes to study the ply size effect without free edges, and then scale-up our analysis into practical design rules and models Simulation Shell models or CLT with in-situ strength Damage models for transverse cracking Delamination Design Thin Ply composites : closer to classical laminate theory as no delamination!! Conclusion Model First ply failure 0° ply failure Experiment Damage Ult strength CLT with damage 287 MPa 609 MPa Thick ply 300g/m2 248 MPa 595 MPa CLT no damage 819 MPa Thin ply 30g/m2 821 Mpa 847 MPa Model First ply failure 0° ply failure Experiment Damage Ult strength CLT with damage 287 MPa 609 MPa Thick ply 300g/m2 248 MPa 595 MPa

Why Thin Ply composites? Thin ply : below 125 g/m2, down to 15g/m2 today Advantages: Improved delamination resistance, higher onset of damage & ultimate failure Improved fatigue properties and in some case damage tolerance More design degrees of freedom = more optimal laminate Easier to design ply drops / small angle laminates / local reinforcements Can produce “homogeneous” laminates, no more dependency on stacking Can produce mixed “thin & thick” laminates for structures with large shell thickness variations. Thin ply materials are now commercially available Challenges: Characterization and understanding of ply-thickness effect for a wide range of constituents Efficient and accurate models for prediction of thin ply composite performance Flexible manufacturing (assembly of complex preform) & automation to reduce layup time Develop & validate efficient design method to account for the performance benefit without blowing up the degree of complexity of the problem Intro Lamina level Laminate level Element level Let’s have a look at the main contributions to the subject in the littérature: Wisnom described size effects in a geometrical scaling with cst ply thickness, observing an increase in strength with the number of sublaminate repetitions Tsai showed that thin-ply laminated composites can supress the microcracking and delamination. The comparison was btw thin plies and a ply-block stacking of thin plies Camanho showed the same trends for non-crimp fabrics, with also an improved behaviour in mechanically fastened joints . The comparison was thin plies vs … Simulation Design Conclusion

‘Thin ply’ effects at different scales At the lamina level More uniform microstructure, lower local variation of fiber Vf and finer voids Improved longitudinal compressive strength At the laminate level Delay or suppression of delamination as damage / failure mechanism Delayed transverse cracking => apparent increase of transverse tensile strength Increased QI laminate onset of damage and ultimate strength Increased laminate fatigue strength At the structural component level Notched tests: improved onset of damage and fatigue life increase; more brittle response but better predictability Bolted joint strength: strong strength improvement in hot-wet conditions Impact effects: from delamination to fiber failure, optimal ply thickness is « intermediate ». Microstructure Intro Lamina level Laminate level Element level There are quite a few conclusions for this experimental part, but the thing to retain is that using thinner plies/having more repetitions leads to strong improvement of most properties as delamination and damage in off axis plies are nearly suppressed . No damage means a better predictability and durability. Summary Change of failure mode (delayed delamination) Simulation Design