You are on page 1of 12
Ineraviona journal of Solids an Srustres 106-107 (2017) 36-67 Contents lists available at ScienceDirect iby AND Stacie International Journal of Solids and Structures — Journal homepage: www.elseviercomiocatolisolstr Numerical simulation of the ballistic response of needle-punched nonwoven fabrics Mem F, Martinez-Hergueta®, A. Ridruejo”, C. Gonzalez, J. Llorca®™* ‘ine Moros ste, Ce Kandel 2 26906, Gf, Mad, Sain Dapurent of Matias Scone, Payecnic Unersty of Maan, ETS de Inger de Cris, 28040 - Madr, Sale ARTICLE INFO ABSTRACT Ire Fon Received 12 Tuy 2016, Review 31 October 2016 Frnlable one 29 November 2016 ‘The ballistic performance of @ mechancaly-entangled nonwoven fabric was studied by means of num ‘cal simulations based on a physialy-based constitutive model, which was recently ceveloped and val ‘dated for quasi-static in-plane deformation (Martinez-Hergueta, ct al, 201Gb). The constitutive model Included the main deformation and failure mechanisms experimentally observed (fiber uncuring.re- wont Corientation, non-afine deformation, anisotropic connectivity indice by the entanglement points 28 well Nonwoven fabrics 2 fiber dsentanglement and pull-out from the knots). The numerical simulations captured the balisic Impact, limit of the as-received material as well as of samples which have been pre-formed along diferent Nutiscate moding forientations. Moreover, the anisetropy ofthe nonwoven 35 well 2s the localization of damage around the Impact point were adequately predicted. The simulation results helped to understand the role played by {the diferent deformation and failure mechanisms during impact. nal the mode was used to explore the influence of non-affine deformation and of the anisotropy in the fiber connectivity on the impact sponse. This information is very usful to design navel nonwoven fabrics with improved ballistic Pet- formance. © 2016 Elsevier Lid. All rghts reserved. 1. Introduction Dry fabrics made up of high strength fibers are a lightweight solution for ballistic protection. They can be used in the form of woven or nowoven fabrics. In the former, fibers are bundled to- gether in yarns weaved following a regular pattern, while fibers form a disordered network consolidated by means of thermal, chemical or mechanical bonds in the latter (Russell, 2007). Non- woven fabrics present lovier stiffness and strength but higher de- formability and energy absorption capability than their woven counterparts. The use of woven fabrics in ballistic protection is a conventional solution (Naik et al, 2008; Cunniff, 1992; Tabiei and Nilakantan, 2008) but nonwoven fabrics have also been used in im- pact applications involving damage reduction on rear components (Lin et al, 2009a; 2009) or as cores of sandwich structures (Lin fet al, 2008; Shahdin et al, 2009), Moreover, it was shown that ‘nonwoven fabrics outperform the woven counterparts in the par- ticular case of impact of very small fragments, providing the same ballistic protection with one third of the areal weight (Cheeseman and Bogett, 2003: Ipson and Wittrock, 1966) ‘omesponing autor. Fx: #54915503047 Ema eds: javierneabupmes Uo ae ot orto oet-2NE LOL (0020-76836 2006 Ese Ld, Al igh eeserved A detailed understanding of the deformation and failure mi cromechanisms of dry fabrics duting impact is necessary to im- prove their ballistic performance. Because the experimental char acterization of these micromechanisms during impact is very chal- Jenging, numerical simulations have been used as a complemen- {ary tool. Most of the work on this area has been focussed in ‘mesoscale approaches for woven fabrics, in which every yarn is, explicitly included in the aumerial simulations (Termonis, 2008; CChocron et al, 2010; Ha-Minh et al, 2013: Rao et al, 2009: Talebi et al, 2008; Barauskas and Abraitien, 2007; Nilakantan et al., 2010). Thus, the mechanical interaction of the projectile with the individual yarns and between yarns can be ascertained. Mote de- tailed models within this framework included every fiber within the yatn around the impact region to take into account the indi vidual fiber interactions within the yarn (Nilakantan and Gillespie, 2012: Nilakantan, 2013), Other authors have analyzed the impact behavior of woven fabrics by means of a multiscale approach, in Which a constitutive equation of the fabric is obtained by means of the analysis of a Representative Volume Element (RVE) of the ‘microstructure. This information is then used to predict the me- chanical behavior of the specimen by means of the finite element ‘method (Wang et al, 2010; Tabiei and Ivanov, 2002; Parsons et a, 2013: Zohdi and Powell, 2006 Mons erg enema Jura of Sos ana Sica 106-107 (2017) 56-67 ” ln the case of nonwoven fabrics, recent modeling efforts have been focussed in the simulation of the in-plane mechanical be- havior under quasi-static conditions. A first simulation strategy in- cluded the actual microstructure of the fiber network. The inter- action among individual fibets or fiber bundles is accounted for by means of different mechanisms located at the crossovers or en- tanglement points (Ridruejo et al, 2010; Kulachenko and Uesaka, 2012; Isaksson et al, 2004; Farukh et al, 2012; 2013). This ap- proach is computationally prohibitive when the ratio between the specimen dimensions and the average fragment length of the non~ ‘woven becomes large and computational homogenization can be used instead, In this case, the mechanical response of the nonwo- ven is obtained by means of numerical simulation of the behavior of a RVE of the microstructure (Heyden, 2000: Dirrenberger etal 2014). OF course, the size of the RVE has to be large as compared ‘with the length of critical microstructural features (average dis- tance between entanglement points, pull-out length, ec) to obtain results that are independent of the RVE dimensions. Nevertheless this approach can also be very expensive from the computational viewpoint for high density nonwoven fabrics. To overcome these limitations, more recent models were based in multiscale homoge- nization (Ridruejo et al, 2012; Raina and Linder, 2015; Martinez- Hergueta et al, 2016b), which is achieved by the integration of two main ingredients: fiber and network features. The first one Includes the mechanical response of a single ber and the inter actions with its neighborhood while the second takes into aecount the fiber spatial distribution, the fiber orientation distribution and the bond distribution. They are combined to obtain an average re- sponse of the mechanical field variables by adding the contribu- tions of al fibers with different orientation given by the fiber ori- entation distribution. This homogenization scheme is able to deter- mine the local averaged response of an RVE of the nonwoven fab- ric and it is particularly suited to work as a constitutive equation in commercial finite element codes as a material user subroutine. These advanced modeling approaches have not been used -to the authors’ knowledge- ta analyze the mechanical behavior of nonwoven fabrics under impact and the only attempts have been based in phenomenological models (Chocron et al, 2002) that ean- not provide information about the relationship between the mi- rostructure of the fabric and the ballistic performance. In this in vestigation, a physically-based model of the mechanical behavior of mechanically-entangled nonwoven fabrics, recently developed and validated for quasi-static in-plane deformation (Martinez-Hlergueta et al, 2015; 20160) is applied to study the ballistic behavior. The model predictions are compared with a new set of experimen- tal data, in which the influence of the fiber pre-deformation be- fore impact on the ballistic performance was measured (Martinez- Hergueta etal, 2016a). The simulations provide novel insights into the influence of eritical microstructural parameters (fiber orienta- tion, anisotropy of the fiber connectivity) as well as deformation mechanisms (affine v5. non-afine deformation) on the ballistic per- formance, 2, Material and impact experiments, ‘The constitutive model is based on a detailed experimen- tal characterization of the deformation and failure micromecha- nisms under quasi-static and ballistic conditions of a polyethy- lene needle-punched nonwoven fabric trademark Fraglight NW201 (DSM), The main microstructural features and the ballistic perfor- mance of the nonwoven fabric were detailed in Martinez-Hergueta Cal (2015); 2016a) and are briefly recalled here for the sake of ‘completion, ‘The nonwoven fabric was manufactured by the continuous de- Position of single filaments of Dyneema SK75 polyethylene fibers ‘of approximately 60mm in length. The batt was consolidated by needlepunching. producing mechanical entanglements among fibers (Russel, 2007). The nominal areal density and thickness of the fabric were ~ 190-220 gim? and ~ 15 mm, respectively ‘The manufacturing process introduced two principal material di- rections known as machine (MD) and transverse (TD) which fol- lowed the bed displacement and its orthogonal, respectively ‘The fiber orientation distribution function (ODF) was analyzed in detail by means of 3D X-ray computed tomography (XCT) and 2D wide angle X-ray diffraction (WAXD) and it was found to be isotropic. However, the in-plane mechanical properties were highly anisotropic: the stifiess, strength and energy dissipated upon ten- sile deformation in the TD were 2-3 times higher than those along MD, while the strain at maximum load along TD was only one half ‘of that along MD. Micromechanical pull-out tests indicated that the structure of the knots connected more fibers along TD than along MD and the better fiber interconnection led to a larger active fiber skeleton, enhancing the mechanical response along TD. In terms of| affinity, fabrics deformed along TD essentially displayed affine de- formation -ie. most of the macroscopic strain was transferred to the fibers by the surrounding fabric-, while MD-deformed fabrics underwent non-affine deformation, and most of the macroscopic strain was not transferred to the fibers. Further information can be found in Martinez-Hergueta etal. (2015). ‘The ballistic performance of the nonwoven fabrics was charac- terized by means of impact tests using steel spheres of 55.-mm in diameter and 0.706 g of mass. Special attention was paid to assess the influence of the fiber ODF on the ballistic performance. To this ‘end, nonwoven specimens were subjected to quasi-static deforma- tion along TD oF MD prior to the impact tests to modify system- tically the initial fber ODE. Rectangular samples of 500mm in width and different heights were deformed up to a fina length of |, =500 mam, so the final dimensions of the samples for the bal- listie rests were 500 x 500mm The initial height of the sample, Ip, was selected according to fp = Ip/(1-+) and the samples were ‘deformed up to an engineering strain ¢ of 10%, 20%, 40%, 60%, 80%, 100% or 120% along MD and up to 10%, 20% or 40% along TD. As- received and pre-deformed fabrics were fully clamped to a metal frame by screws and subsequently subjected to impact tess. The impact velocities were in the range 270 to 400. mjs, leading to impact energies between 25 and $5. (Martinez-Hergueta etal 20160), ‘The deformation and failure micromechanisms were equivalent for quasi-static and impact conditions (Martines-Herguets et al. 2015; 20160). Fibers were initially curved and load was trans- ferred within the fabric through the random and isotropic net- ‘work of knots created by needlepunciing, which led to the for- mation of an active fiber network, Uncurling and rotation of the active fibers was followed by fber sliding and pull-out from the ‘entanglement points. The main difference in the deformation mi- ‘romechanisms between in-plane and ballistic tests was the av- ‘erage pull-out length in the failure region, tye, which increased from 35 mm under quasi-static deformation up to the whole fiber length, Lyage = 60 mm, under impact (Fig. 1), More information can be found in Martinez-Hergueta et a. (20163). 3. Const ive model ‘The constitutive model of the needle-punched nonwoven fabric ‘was presented in Martinez-Hergueta et al. (20160) and validated ‘against the in-plane mechanical tests under quasi-static conditions. ‘The model provides the mechanical response of a mesodomain of the nonwoven fabric. which corresponds to a finite element in the numerical simulation, and it is briefly recalled here for the sake ‘of completion. The model is divided in two blocks which deal ‘with the network and the fiber response. The network model es= lablishes the relationship between the macroscopic deformation ss Marine Herguta eh latemaronl ura of Ss an Struares 106-107 (207) 56-51 Fig. Deformation of the omen fic pre-deformed 60% along MD during impact at 322_ ms. 8) € Iiiges su the extraction othe woe er tom te fa dng penetration, Fig. 2 Schematic of the network meson formed by dierent se of cl es gradient F and the microscopic response obtained by integrating the esponse of the fibers in the mesadomain. whichis formed by 2 planar square region ofarbitary size containing a random net work of tng, cy, noninteracting bes Each ber schracterie inthe unstressed configuration by 3 unit vettr Whose orientation i given By the two extremes of the ber, fy ahd Oy scerding to (Fs 2) FoQ0 MPoQoI shore and Qy stand forthe positions ofthe wo extremes of the Abe nthe ina, nsessed Eongration, which taken a ret erence. fl forms an angle Bp with respect to an arbitrary, privileged sirection (eg. the tases erection TD of nonwoven fare In the itil configuration ithe mesodomain i subjected ta cetain imposed deforms ton, given bythe deformation gradient F, the equivalent ber ex tension = AER wil be given by the projection ofthe deforma- ton gradient Fon the uit vector A accoring to Jena @) where C = F'F is the right Cauchy-Green strain tensor, assuming tis deformation mechani forthe fibers within the network the mechanical response ofthe mesodomain can be ex- pressed in terms of the second Pil Kichat nominal tres enor, Si (force per unit width) a6 GED ap, @) 5 ie ‘V(Bo)4,)5(Bo) FE where W(Bo) is the fiber ODF in the reference configuration, 047) the stress cartied by the fiber as a function of the stretch % //o) a 150 ps, b) 1 = 225 ns and (6) = 3505 The sucessive the active fiber length engaged in the deformation process as a function ofthe fiber orientation in the reference configuration. fo, with respect to the privileged direction (1D), C= FTE the right Cauchy-Green strain tensor, and Ais a unit vector wich indicates the end-to-end fiber orientation in the undeformed configuration. X-ray analysis (XCT and WAXD) of the nonwoven fabric indi cated that the intial ber ODF was isotropic and Wf) = 1 However, the active fraction of fibers (per unit area) that eon- twibuted to the deformation, /4n), depended of the orientation, and it was given by S(Po) = bn (Po) ~ 2 (4) where a = 010836 and 6 fulfill che condition 223 are two parameters selected t0 o_fi ) B= fi sore 6) Which implies that the sum of the active fiber length in all orienta- tions is equal 0 the total fiber volume per unit area of the fabric py. where p is the areal density of the fabric and py the fiber density (Martinez-Hergueta et al, 20160), The fiber model takes into account the deformation features ex perimentally found for each set of fibers. including fiber uncuri- ing and re-orieatation, non-affine deformation, pull-out and dis- entanglement (Martinez-Hergueta et al, 2015). These mechanisms were introduced in a phenomenological model that characterizes the stress (o,) - stretch (2) behavior of the fiber according to oy = K[n(BoX%— DP On <0; . ro 6 of = 1 AK In PoI.— Df Y Our > Ope Where K is the curled fiber stiffness within the fiber network (which includes the effect of the interaction with other fibers), Oo the pull-out strength, and oy isa trial stress computed from the fiber stretch assuming no damage. n(o) is the affinity factor, Which depends on the fiber orientation, as shown in Fig. 3, and in- dicates that all the deformation is transmitted to the fibers along, the TD (Bp = 0} but this is not the case along other orientations. Which show non-affine deformation, Once the fiber stress has attained the pull-out strength, oo the pull-out process begins and the load cartied by the fiber de- creases progressively until it is completely disengaged from the fabric. The stress carried by the fibers during this stage is deter ‘mined by means of a continuum damage model, and it is a func tion of the damage parameter d and of the fracture energy per unit fiber cross-section dissipated dung pull-out, G. This fracture energy was obtained from the pulley theory, and it is given by Martinez-Hergueta et al. (2016), ple 1m ) G = SEE (1 emmy 2) Mons erg enema Jura of Sos ana Sica 106-107 (2017) 56-67 ss 02 ° wa w2 By ig 3. Evolution of he aft ator» with the It Aer eiemaion A. By 0 corfespands to the TD and f= t0 MD. The magnitude Is syrmmetne for te sora [52.0} Tbe 1 Model parameters Fiber densi ( 70 kim ber leh jr) © mm ‘Areal density (9) 02 ui? Fiber sites) 900 Gr Ina Me curate, (0) 2 Pall-ot sera (0) Ins, wri Gre Fer ult lena ye) 0 mn Contr eng Le) 20 mm {tcl vale ofthe host strain Oye) 035 Friction cee 4) ot ‘where Le is the fiber length between entanglements, jc the friction coefficient, the fiber curvature and ize the number of mechani ‘al entanglements involved in the pull-out process, which depends ‘on the local deformation according to (8) where Oya = (On, + Or2)/2 and Gu =Un if Un<1 or Ou=1 otherwise Un if Un=1 or On=1 otherwise ® where Up, and Usp stand for the components of the right stretch tensor U: Oy is the critical value of Og at which the contribution ofthe knots to the pull-out process Becomes negligible. Thus, en- ray disipated by tition during pull-out decreases if one of the Principal stretches i compressive and the reduction increases with the compressive stain, This effect of the local stain state reflects the fact that some entanglement points are lose (and, thus, do not contribute the pull-out stress) if they are subjected to compressive strains, More detalls about the model ean be found in Martiner- Herueta et al, (20160). ‘The model parameters were obtained from independent mi- cromechanical tests or chosen to reproduce the experimental 1 Sults of in-plane quasi-static deformation in Martinez-Hergueta 3, (20169) and they are shown in Tobe I, Te only diferences With the parameters in Martines-Herguta ct al, (20166) are the fiber pull-out length, Ly, which was increased up t0 Lyper because ‘of the experimental observations in the ballistic tests (F. 1), and the pull-out strength, apo, which was in the range [03 ~ 17] GPa for the ballistic simulations, slightly higher than the range [03 ~ 1.0] GPa chosen to simulate the quasi-static deformation. Never- theless, both ranges are within the experimental range measured in the fiber pull-out tests (Martine2-Hergueta etal, 2015), 4, Numerical simulations ‘The in-plane deformation as well as the behavior of the ballistic impact of the nonwoven fabric were simulated using the finite ele- ment method, To this end, the constitutive model presented above ‘was implemented as a VUMAT subroutine in Abaqus/Explicit. The rhumerical simulations were carried out within the framework of large displacements and rotations with the initial unstressed state taken as reference, Square specimens of 300 mm width and vari- able length, lg (from 228 mm up to 500 mm) were discretized ‘with MBD4R- membrane elements, with reduced integration (1 Gauss point per element), enhanced hourglass control and sec~ ‘ond order accuracy. Out-of-plane shear stresses were neglected, as in the Euler-Bernoulli beam formulation. This geometry was large ‘enough to accurately simulate the impacts on the clamped fabric with a free area of 350 x 350 mm?. The mesh was finer around the impact zone (finite elements of 1_ mm?) and the element size increased progressively with the distance from the impact zone to reduce the computation time. ‘The simulations were cartied out in two successive steps. The samples were initially stretched along either TD or MD to reach the final length of 500mm, The displacements of the upper and lower boundaries of the samples were fully constrained, while the lateral boundaries were stress-free. The tensile test was simulated by imposing a vertical velocity of 1 mjs to the upper boundary of the square. This velocity was much higher than the experimental ‘one to teduce the computing time but it was able to reproduce the quasi-static deformation because dynamic and inertia effects ‘were negligible (Wartinez-Hergueta et al 2016, The second step ‘of the simulation was the impact of the fabric with a steel sphere ‘of 55. mm in diameter and a density equal to 785. g/cm”. all the boundaries of the specimen were fully constrained during the im- Pact simulation. The steel sphere was modelled asa rigid solid and the friction coefficient between the sphere and the fabric was 01. ‘The right stretch tensor U was used to compute the stretch 2. of ‘each set of fibers. Each mesodomain ofthe fiber network (that co incides with a finite element with ane Gauss point) was described by 65 sets of fibers with different orientation in the range (=. 4) and the second Piola-Kirchoff nominal stress tensor, $, was eb- tained by integrating Eq, (3) along the different orientation. ‘The fiber stretch 7 was used to compute the fiber curvature according to 6 n= Tsing (10) and the fiber orientation fF was implemented under affine assump- Un +Un 8 2a an) ay where Ur, Ura and Un» are the components of the right stretch tensor UThe component Un. ofthe ght stretch tensor U is par allel to TD. “The fiber stretch 2 for each fiber set was wsed to compute the tial stress rin the absence of damage, which was com> pared withthe corresponding pull-out stenath Oye. The pull-out Strength foreach fiber see was chosen atthe beginning ofthe sim ulation using a Monte Carlo lottery in the range [0.3 - 1.7] GPa (Table 1). Ifa = po. Orr = 7 according to Ea. (6). tan ( 0 Marine Herguta eh latemaronl ura of Ss an Struares 106-107 (207) 56-51 350 ‘Simulation 300 250 a0 200 cas 150 . 100 Residual velocity (m/s) 50) 002034060 Initial velocity (m/s) 400 Fig. 4 Biperenta ane simulation eeu of the lise est 09 te as tected If og > Ope. the actual fiber stress oy was computed with a phenomenclogical damage variable d provided by the continuum damage model in Martinez-Hergueta et al, (2016b) which ensures that the energy dissipated is always equal to G and independent of the finite element discretization, The onset and propagation of damage was controlled by a damage activation function and the evolution of damage was controlled by the consistency condition ‘The evolution of damage was inked to the damage variable d, con- trolled by a numerical parameter A, function of the characteristic length of the finite element, fy (equal to the diagonal length in the case of square elements), The effectiveness of the mesh regulariza- tion through the parameter A was checked in a previous publica- tion (Martinez-Hergueta et al, 20160) by carrying out simulations along both directions with different finite element discretizations [A damage variable D was defined for each finite element (Causs point) as the average damage of all fiber sets in the element. If D > 0.98, the element was removed from the simulations to avoid excessive element distortion. 5. Results and discussion 5. Impact behavior of as-received fabrics The constitutive model presented above was used to simulate the impact behavior of the as-received nonwoven fabric at differ ent impact velocities. The results of the simulations, in terms of the residual velocity, Vis. as a function ofthe initial velocity, Vig, are plotted in Fig. 4. I" should be noted that several simulations Were carried out for the same initial velocity because the model predictions presented some scatter due to the changes in the fiber pull-out strength of each fiber. The experimental results presented in Martinez-Hergueta et al (2016a) are also plotted in Fig. 4. The comparison between both shows that the numerical simulations were able to capture the ballistic limit as well as the increase in the residual velocity just above the ballistic limit, Nevertheless, the ‘model underestimated the residual velocity for impact at velocities above the ballistic limit. It should be noted that the final penetca- tion of the target in the tests carried out above the ballistic limit was accomplished by tearing as the fibers were pulled out from the entanglement points or broken neat to the impact point due to thermal softening (WMartinee-Hergueta et al, 2016a). This latter ‘mechanism of fiber fracture was not considered in the constitutive ‘model and may account for the differences with the experiments above the ballistic limit. The numerical simulations were useful to understand deforma- ‘ion mechanisms around the impact point. The in-plane stresses were transmitted by a longitudinal wave and the nonwoven fab- ric was stretched towards the impact point, generating. complex stress and strain fields (Fig. 5). The regions affected by the lon- gitudinal waves and by the (slower) transverse waves were elip- soidal with the longest axis parallel to TD because the wave speed slong TD was higher than along MD as a result of the mechanical anisotropy of the fabric. The in-plane stresses were rather similar along TD and MD far away from the impact point. The in-plane stresses around the impact point along TD were, however, higher than those along MD, Fig. 5(b),(d) and (f). These differences be- ‘ween TD and MD were more marked in the strain flelds, Fig. 5(a), () and (e). The strains along MD were much higher because the deformation of the fabric in this orientation was noa-affine. Thus, conly a fraction of the strain was actually transferred to the fab- ric in this orientation and much higher strains were necessary t0 reach the sttess level at the impact point. These differences have been experimentally observed during impact on anisotropic plates (Geidt et al, 2013), The transverse wave led to the global deflection of the fab- ric (Fig. 6), and generated a cone of deformed material with an elliptical cross-section. The ratio of the major to the minor axis fof the cross-section in the simulations was 16 as a resule of the anisotropic mechanical properties induced by the entanglement points. This rato is close to the experimental value of 2 reported Jn Martinez-Hergucta etal, (2016a) and the differences can be at- tributed to the limitations of the continuum model to accurately reproduce the wave propagation along a mechanically-entangled fiber network. Finally, both experiments and simulation showed that penetration took place by the complete disentanglement of the fiber network at the impact point. The constitutive equation also provided information about the energy dissipation processes in the fabric during impact. Outside of the impact point, the fabric was mainly stretched in tension along, ‘TD and MD (Fig. 5) and the deformation micromechanisms were similar 9 those found under in-plane tensile deformation: fiber tuncurling and rotation was followed by fiber sliding and pull-out from the entanglement points, although these latter mechanisms were only active around the impact point. t should be noted that the extracted fiber length (80 mm) was much larger than the length of the damage zone (of the order of the diameter of the projectile, 5.5 mm), indicating that the strain gradient perpendic- Uular to the fabric around the impact region was very large. Never- theless, energy was not only dissipated in this region because the behavior of the fabric was pseudo-plastic in the whole deforma- tion range (elastic, recoverable deformations were always negli ble (Martinez-Hergueta et al, 2015). Thus, the whole volume of deformed fabric contributed to absorb the energy of the projectile ‘More details about the microstructure evolution ean be eb- tained by the reconstruction of the fiber ODF during the impact test. The orientation of each fiber set, 8, was computed according to £4, (1) a8 a function ef the deformation gradient and the fiber curvature, 6, was determined according to Eg, (10) as a function of the fiber stretch 2. Ths information was used to build the fiber ODE. W. following the same methodology employed in Martinez~ Horgueta et al. (2015) for the treatment of the XCT data, The po- lar representation of the fiber ODF at two points X and ¥ located at 15mm from the impact point along TD and MD (Fig. 7) are plotted in Fig. 8, 125 js after the impact at 300 mjs. The origi nal, isotropic ODFS are also plotted for comparison. The maximum strain was oriented along TD in point X and along MD in point Y 075 20 08 16 0.45 2 03 8 os ° Re & Conor poof the stan and ses els forte aseceedtonwoven fc impacted a 300. ms 2 safe the np (2) Mamam pric gai lines inthe Figures stand forthe boundary ofthe transverse we Ds (Stress along TD, ay (6 nganthmic stain slng ME. eu along MD, eu. Te Brae Mo 10 é x () 08 06 04 mn Fg. 7. Locatos of points X and with espec 10 the impatpoit 02 0 D cs ° o Fe. 6. Later view ofthe asec bi ding the mgt at 360 mis eading (a) 8S (wssy Iisa (€) 250s (D) Cantour pak of the mage arable a 100. ee Fay eseatange avi (or aterpetaion ofthe relereces 2 Marine Herguta eh latemaronl ura of Ss an Struares 106-107 (207) 56-51 400 - 400 Predeformed| ” ] ) Predeformed * e)) 350 20% along TD aso. 40% along TD a Experimental] © Experimental a 200, » Simulation : mol | © Simulation E a0 E aso 2 . 2 8 20 . 3 200 dj Bo. : 10 J 3 3 2 100) & 100 50) : J | 9 tees owe. 00240288" 320360400 200 “240 "280320360400 Initial velocity (m/s) Initial velocity (m/s) Fig 8, Experimental an simulon results ofthe Dali ess in he specimens pre-deformed akong TD. (2) 20% pee deformation (b) 40% pre eemaion ‘and thus fibers within the fabric rotated towards different diree- tion in both points. Fibers in point X tended to be oriented along. ‘TD (B = 0) while those in point ¥ were aligned towards MD (B. 12/2). The degree of orientation depended on two factors: the vol lume fiaction of fibers involved in the deformation process (which \was larger along TD than MD due to the anisotropic connectivity of the entangled fiber network) and the applied strain (which was larger along MD due to the non-affine deformation in this orien- tation). Both contributions led to a slightly higher re-orientation along MD in point ¥ than along TD in point X. but the differences were not very large. 52, Pre-deformed nonwovens Previous experimental results (Martinez-Hergueta et al, 2016a), have shown that pre-deformation of the nonwoven fabric along TD and MD led to a reduction in the ballistic limit. They ate compared With the results of the numerical simulations in Figs. 9 and 10 for the specimens pre-deformed along 1D and MD, respectively. The solid lines in these figures stand for the fitting of the simulations results to the Lambert formula, Ves = (Vins ~ Vio)" (12) Where Vp) and Ves ate the initial and residual velocity of the pro- jectile, respectively, The ballistic limit, Vso, and the exponent 1 were obtained by the least squares fiting ofthe simulation results ‘The simulations provided very accurate predictions of the reduc tion in the ballistic limit with pre-deformation in both orientations and of the ballistic curves along MD up to a pre-deformation of 0S, The predicted values of the residual velocity above the balls tic limit for the specimens pre-deformed along 1D and for those pre-deformed along MD more than 100% were slightly lower than the experimental data, very likely for the same reasons indicated for the as-received specimens. The experimental and the predicted variations of the ballistic limit with the pte-deformation strain along TD and MD ae plot- ted in Fig. 11. Experimental results were obtained from Mastinez~ Hergueta etal (2016a) while Vsq for the simulations were obtained from the ballistic limits given by the Lambert formula. The agree- ‘ment between numerical and experimental results is very good and both show an approximately linear reduction of the ballistic limit with the applied pre-strain, However, deformation along TD Jed to a much higher reduction inthe ballistic limit than along MD. The information provided by the numerical model can be very useful to understand these large differences in the ballistic perfor- ‘mance as a funetion of the pre-deformation along TD and MD. The contour plots of the maximum principal logarithmic strain ej, and of the Von Mises stress, oye ate plotted in Fig. 12 for the spec- Jimens pre-deformed by 40% along TD and MD 125 1s after the impact at 300. ms. Pre-deformation along TD by 40% led to the preferential orientation of the fibers along TD, as shown in the fiber ODF in Fig, 13(2) and (b), increasing the anisotropy of the fabric. This is shown in the shape of the pseudo-ellipse that de- lineates the boundary ofthe transverse wave in Fig. 12(a) and (b). However, the strain and stress fields in the fabric were rather ho- ‘mogeneous and higher stresses were only found very close to the impact region, Moreover, further re-orientation of the bers along, ‘TD of MD due to the impact stresses was very limited, These data Indicate that pre-deformation along TD consumed a large fraction fof the energy dissipation capacity of the fabric and reduced the efficiency of the fabric against ballistic impact. In addition, the ex- perimental observations in Martine2-Hergueta et al. (2016a) also pointed out that the large anisotropy in the fiber orientation in- duced by the pre-deformation along TD facilitated the penetration fof the projectile through the fabric: very few fibers remained ori- ented along MD and the projectile was able to slip through the fibers because the fabric resistance to deformation along MD was negligible. Pre-deformation by 40% along MD led to a preferential fiber orientation along MD, Fig. 13(¢) and (d). This promoted a mare ‘sotropic mechanical response, as shown by the shape of the boundary that delineated the propagation of the transverse wave, which was close to a circle Fig. 12(€) and (dl). Nevertheless, pre- deformation by 40% along MD did not exhaust the energy absorp- tion capacity of the fabric, as shown by the strain and stress con- centrations coming from the impact point 125 j¢s after the impact, Fig, 12(¢) and (d). as well as by the re-orientation of the fibers along TD (point X) and MD (point ¥) as a result of the impact. ‘Thus, the reduction in the energy absorption capacity of the non- woven fabric pre-deformed along MD was smaller. The experimen- tal observations in Martinez-Hergueta et al, (2016a) indicated that Mons erg enema Jura of Sos ana Sica 106-107 (2017) 56-67 a 400, ' ' 0 400 - : : : 7 Predeformed @) Predeformed (0) 50, 40% along MD 50 80% along MD © Experimental A Experimental 300-| Simulation + 2 900;| 6 simulation 250) ° E 250 A : 2 200 , 8 200 150 & 10 . 7 100 3 100 4 & 50 | 50 8 8 200 "240 280" 32036000 200 "200 “350 20-360 a0 Initial velocity (im/s) Initial velocity (m/s) 400) ' 1 Predeformed © 350, 120% along MD © Experimental A = oso 3 200 5 10 3 100 e 7280 320360400 Initial velocity (m/s) Fig. 10. Experimental an simulation rests ofthe baltic tests inthe specimens pre-formed along MD. (a) 40% preefomation. (b) 80% pre-dformation, () 120% Pre-eformation, slippage of the projectile between the fibers was more dificult in this ease (because of the isotropic mechanical response). Moreover, interaction of the projectile with the fiber network, which led to more extensive fiber pull-out during penetration, enhanced the en ergy dissipation capacity ofthe fabric 15.3, Influence of affinity and connectivity on the impact respanse ‘The constitutive model presented above can be very useful to understand the influence of non-affine deformation and of the anisotropic fiber connectivity on the ballistic performance of this material, To this end, the ballistic curve and the ballistic limit were ‘computed for four different virtual fabries with different affinity and/or fiber connectivity. The first one was an affine, isotropic non- ‘woven fabric, characterized by 7) = 1 in Eq, (6) and isotropic fiber connectivity J; = £2. The second material presented a non-afine, isotopic behavior with » = 05 in Eq () and fy = ff. The third nonvoven shovied affine deformation and anisotropic connectivity sen by 1) 1 inf. (0) and fin. (4) computed with the val ues of mPa) shown in 3, while the fourth material presented non-affine deformation with = 0 in Ea (6) and anisotropic con necivity ‘given by fn Ea (4) “The bale performance of the four nonwoven fabrics is plot ted in Fig. 6, These results cleaty show that noafine defor mation leads to lage improvement inthe balistic performance, ‘while the fiber connectivity (either isotopic or anisotropic) had 2 eligible influence. The origin of these diflerences can be ana- iyzed inthe contour plots 125 afer the impact ofthe maximum principal logarithmic stan, e1 and ofthe Von Mises stess, ora in the viral fabrics impacted a 300 mis, i. 15, Wave props sation as well asthe strain and stessflds presented radial sym- tmety ftom the impact point in the fabrics with isotropic connec- vi. Fis. 15(a) and (b) and Figs. 15) and (while anisotropic 360 ) 340 320 300 200 280 240 Ballistic limit, V,, (mm/s) 220 200. 4 MD MD MD MD MD MD AR TD TD 120 100 80 60 40 20 20 40 Pre-deformation Fig. 1 Experimental ane simulation rests ofthe ballistic it Vo the on fac fonction of the pre-deformation slong MD an TO, Sola ym. Ftng ofthe experimental Blistici connectivity led to a faster wave propagation along TD, Figs. 15(c and (4) and_15(g) and (h) Overall, the effect of the isotropic ot anisotropic fiber connectivity on the energy absorption capability an be assessed by comparing the contour plots in the first and the second row (for affine deformation) or in the third and fourth row non-affine deformation). In both cases, the differences are mini- ‘mum although fabries with an anisotropic fiber connectivity (sec- ‘ond and fourth rovr) seemed to present slightly higher stres strains than those with isotropic fiber connectivity (First and third sow) and thus provide a marginal increase in energy dissipation These results are in agreement with the ballistic curves in Fig. V4. which show a marginal improvement in the ballistic performance for anisotropic fabrics, which is of the same order as the scatter in the simulations due to the variability in the pull-out strength of each set of fibers, ‘On the contrary, changes in affinity led to large differences in the ballistic response and nonwoven fabrics which present non- affine deformation posses better ballistic properties both in terms of the ballistic limit and of the residual velocity after penetration (Fig. 14), The nonwoven fabrics with non-affine deformation (third and fourth rows in Fig. 15) showed much higher strains and simi lar stresses than those that underwent affine deformation (first and second rows in Fig. 15) and were able to absorb much more energy daring the impact, These results are in perfect agreement with the effect of non-affine deformation in the mechanical response under -Hergueta etal, 20166 in-plane deformation (Martine 6. Conclusions The ballistic perform *dnonwo- ven fabric has been analyzed by means of a physically-based constitutive model for this type of materials, which was re- cently developed and validated for quasi-static in-plane deforma- tion (Martinez-Hergueta et al, 2015; 20160). The model, which was mostly calibrated by means of independent micromechanical and quasi-static tests, was able to capture the ballistic limit of the ‘as-received material as Well as of samples which have been pre deformed along different orientations (TD and MD). Moreover, the anisotropy of the nonwoven as well as the localization of damage around the impact point were adequately predicted, Nevertheless, the model tended to underestimate the residual velocity of the projectile above the ballistic limit in the as-received specimens as well as in those pre-deformed along TD because final penetration took place in these cases by fiber fracture due to thesmal softening (instead of fiber disentanglement and pull-out from the fabric) and this mechanism was not explicitly introduced in the model os 06 04s 03 Fig. 12, Contour plot of he san and sess forte predefonmed fis impacted at 300 as 125 ys after he impact (2) Maximum principal garni stl, ‘Simple pre-detomed by 40% along MD (8) Von Mises Stes, oy, he sample pe-detoraned by AO along MD. The bok lines inthe gues tad or te Boundary of Mons erg enema Jura of Sos ana Sica 106-107 (2017) 56-67 6 (a)T040x Ss tomosox gz — Before impact (a)Moaoy — Aer impact ig. 13, Polar representation ofthe Aber ODF W before apd 125 ps afer impact at 300. mis in the specimens pre-formed by 40% along TD and ND (a Point X 2 () Point Vin 7 tn te specimen preteormed along TO. c) Pit X ad (Pont ¥ in 7 the specimen predeormed long, MD. 400 = Aine & FBorple Connectivity © Ate & ‘rote Comectvity ' = NonAine 300 Isotropic Conmectity & 5 Nonlin ritee Camectvty 200 100 4 Residual velocity (m/s) ° eee io seemnae o 4 250 278° 300° 328" 380 375400 Initial velocity (m/s) ig. 1. Parametcl study ofthe inlunce ofthe network afi and connect ip on the ballsie response of the nonwoven. Aine ad on-afe deformation se characterized by y= and =O inf. (o.respetvy lope and topic fiber connectivity ae ven by fy = and jm Ea especies, ‘The physically-based model predicted the fiber orientation and the localization of damage. Furthermore the model also provided numerical predictions of the strain and stress fields which were analyzed for each configuration to give insight into the deforma- tion micromechanics of the nonwoven fabric during impact, En- ergy was dissipated by the tensile deformation of the fabric around the impact point in an elliptical region whose boundaries were controlled by the wave propagation with different velocities along, ‘TD and MD. Final penetration of the fabric around the ballistic limit occurred by the disentanglement of the fiber from the fab- ric network around the impact point. Pre-deformation of the fabric around TD or MD led to a reduction in the ballistic performance, ‘which was much more noticeable in the former case, The numer- ical simulations showed that pre-deformation along TD consumed 2 large fraction of the energy dissipation capacity ofthe fabric and reduced the efficiency of the fabric against ballistic impact. Hov- ‘ever, these effects Were ameliorated in the case of pre-deformation along MD because deformation was non-affine ang the active fiber fraction was smaller in this orientation. Thus, much larger pre- ‘deformation along MD was necessary to exhaust the energy ab- sorption capability of the nonwoven fabric. ‘The constitutive model was finally used to explore the influence ‘of non-affine deformation and of the anisotropy in the fiber con- nectivity on the impact response. It was found that the presence ‘of non-alfine deformation enhanced the energy absorption capa- bility of the nonwoven because the deformability increased while the load bearing capability did not change. The anisotropy in the fiber connectivity did not significantly modify, however, the bal- lisic performance of the material. This information is very useful to gulde the microstructural design of nonwoven fabrics with im- proved properties. ‘Acknowledgement ‘This investigation was supported by the Ministerio de Economia ¥y Competitividad of Spain through the grant MAT2012-37552. In Addition, FMH acknowledges the support from the Spanish Min- istry of Education through the fellowship FPU12/02087, 0.75 0.6 0.45 0.3 0.15 Fg 18. Contour plot of he asim princi nem strain, Nonwoven fare wth aon-fine deformation sn to te fre stand for the boundary of the tansverse wave on References Bauska, Abralien, A. 2007, Computational anass of impact of 2 ballet ‘pate 9 pat 47) a poste amnates, Compr, Sct 6 (1-2), 101-173, Chowron, S. Fgura, Ey King, M. Kucocrter, Nichol, AE. Sagobie, Ws. ¢, Meas CL 2010 Mein and valiaiin of ut abi ages un er bic impact Compos Se Tena 7 (13), 2012-2022, hocon,S,Finton Ay Cendn,D. Gale, SncheeCslex V. 2002, Simao ‘ot alli impact in poyetene non-woven fk 2 interational Sy posi on Basti, pp 23-27. Onando Cunt HE 1962. An analss ofthe stem effets in wove aes wade ballistic ipa, fet, es. | 62), 495-508, Direnbergen, J. Fre, Jel By 2014, Tomands gigantic RVE sizes for 30 stots bor networks nt. [Solis Srct 5, 30-376, Faruth, Femina, ty buneopia Arar, MPourdeyhimi, Sider etm, Vile 212. Nomen modeling of samagesntsten i ken ‘chm, vn 2013. Charactetston an names modeling of complex de formaten behav in thermally bonded nonwovens. comput. Mar Se. 7 (i mnit tiabi n Mch 6@2 : lulose Abe ft ven of strctural Mesnanes (TH, nd Unwersty Inson LW, Wirrack E18, Respones of non wen syne fbr txts 0 alse impact Tech rep, No. 67-8 CM Denter Research astute tapes. ne) Sos Stat (16-1) AP SS and othe Von Mises stress. gm he viral brs mpace at 300 es 125 2) and (9) Nonuven sc with fine detrmaton snd stmpe connective) ana (e)Nonwoen brie 20 ym (MPa) afer he impact, nth sfnedeormation and ansample connect. (3 uchenk, A Uesaka, T2012 Direct simulations of fiber network deformation sd fare Mech Mater, un Cg CU a, Cn, CML Ha, CHU 200 a tin. CC Lin, CM, Hoane. CC, tou, CW Men LH. Hse, CH, Lin LH, 2008, ‘ting the design snd mapa properties of compose nonwoven abs with ‘rou lament m blero set sshion eres, Res | 79 (3) 268-74 Lin, JH Hu, CHL Meng Hi, 2005 rces of preparing 2 nonwoven | flaent Fes feat 1 (a2), 3-47, Manet, fy hej A Gide, F Gane, C Loe, Je 2016 "ence of fbr antation othe balls perfrmance of neelepunched ono veniabres Mech Mater 4, 10-116 Mating ergs, Fy Rudra A, onze, C, Mac, J, 205, Deformation and ‘energy dsption mechanisms of eed panched enone Tabs: 4 mal Seale experimental analysis Ie Sos Sct. 65, 120-1 Maines erguets Far, Ac Constr Lae, 216 "emechaical model of Rediepuncbed nenworen fees Nile Be Sankaran, Mobusher.B, Raa, S. Fer, J, 2000. Development ofr experiment ses In mpct Ene 36 (0), 1-1 a Nislantan,G, 2013, Flment- vel modeling of evar KM2 yarns for tals im Tne) ol Src nto efits. Compo, Sac 94 12, 304-3094 Nikanran, G, Keo, Mc Boget, TA. Adkinson, R.Cilespl, JW, 2010. On he ‘te cement ansis of woten frie impact wing mulscale modeling tec nies lt J Solids rut. (17, 2300-2315, Mons erg enema Jura of Sos ana Sica 106-107 (2017) 56-67 " asons EML King MJ, Soca S, 201. Modeling yam sip n woven fbi atthe ‘ining eve: simulations of bse impact F Mech. Figs. Sods 61 (1, e202, ‘Raina A Linder C, 2015. A micomectanial model wit song icons for lute in noowovens at fate deformation.) Solds Suu. 75-76, 247-259 ‘Rao. M, Duan Ket, M, Powe, B. Boge, 1, 2000, Modeling he ete of ara materi propertes apd lon onthe bllsc Impact oF 2 aimee Fabric, Compos Stet. 89), 556-506, iru A ConeiezC, Love 2010 Damage micromechanisms and noch sn Suvi of as-ber non-woven Fes an experimental and aimee Sty Mec Ps oli 8 (10, 38-1685, ‘iru A, Coils C, Lv, , 2072, A constitute model forthe in-plane me ‘hunal behavr of nonwoven abs Ink} Sols Stat 9 (17) 2215-2228 sel, Sf. 2007 Handbook of rormovens. Te Textile Insite. Woodhead Pub- sting Side [Dy Michael Perit, Git, A, Revlock, BM. Nandan, K, 203 aise yt anc Hn ti hc an pet) op Eg. Shahan, A, Mewes, L, Bowe, C, Modi, J. Gournat , 2008 Fabration and mechanical esting of ss ie entangled sandwie Deams: comparison ih honeycomb and foam sna beams Composit 9D (4) aD “ie Avan. 1, 2002 Compuainal miro thechanel model of exible woven ‘abi for ite element impact simulation. lt. J, Num, Methods Eng $3 (6), ‘asa-t278 ‘able, A, Nisiantan,G_ 2008, Basic impact of dy woven fabric composes: Fev, Appl Mech ey. GL (1 O10801~1-13 “deb HW, 5, Haro, 2009. Fite clement evaluation of projectile nose ‘iglesias perforation o high sen fabric. Comp. Struct. 87 (ysis “Teron, 2004 lps existance of woven fabric. Text, Rs 24 (8, 723-729, ‘ang Mio, ¥ Swenson, Cheeseman, Bay Yen, C- Latina, 8, 2010 Dig Hal clement proach for Simulating pact and penetation of textes Impact Ege 37 (3), 552-560 ‘Zot TL, Pomel D, 2006, Muscle construction and tage scale simulation of Stier stm Compa Mab ope

You might also like