Sie sind auf Seite 1von 19

Composites: Part A 37 (2006) 18971915 www.elsevier.

com/locate/compositesa

Simulating the deformation mechanisms of knitted fabric composites


M. Duhovic, D. Bhattacharyya
*
Centre for Advanced Composites Materials, The University of Auckland, Building 740, Tamaki Campus, Morrin Rd, Glen Innes, Private Bag 92019, Auckland, New Zealand Received 9 September 2005; received in revised form 19 December 2005; accepted 29 December 2005

Abstract For composite materials with complex reinforcing structures, such as knitted fabrics, a material model must be able to accommodate the most important parameters so that the material can be simulated accurately. Sometimes these important parameters can be found by examining the reinforcing structure alone. To investigate these parameters physically would be very dicult because of the scale and level of detail required; however, using advanced numerical micro-simulations a large quantity of information can be made available. By numerically simulating the tensile test of a knitted specimen the most important parts of the textile structures deformation behaviour can be identied. In this study a numerical micro-specimen has been developed by simulating the actual fabric manufacture, so that the force-determined geometry and residual stresses incurred during the manufacturing process are captured. The two-step modelling procedure is performed using a dynamic explicit nite element code. Each lament is represented by a series of rigidly connected beam elements, which undergo a complex set of contact interactions. To validate the model, numerical simulations of the warp direction load versus extension is compared with the experimental data. It has been proposed that eight individual deformation modes exist for textile materials in general. For dierent reinforcing fabrics the inuence of each of these mechanisms during fabric deformation is dierent; however, the denition of a generalised model allows user inputs for these parameters. These input data can be obtained by taking numerical measurements from the micro-specimen model, which is in itself, capable of producing dierent knitted structures by adjusting the boundary conditions that represent elements of the knitting machinery. 2006 Elsevier Ltd. All rights reserved.
Keywords: A. Glass bres; A. Polymer (textile) bre; C. Finite element analysis (FEA); E. Knitting

1. Introduction Over the past decade there has been a signicant amount of research interest in the eld of thermoplastic composites, especially fabric-reinforced sheet materials. An understanding of the unique behaviour of these sheet materials during forming and the practical viability of various forming processes are crucial in assessing the applicability to given products and uses. A relatively recent innovation in this eld is the use of knitted textiles as the reinforcement. Knitted fabrics oer a number of potential advantages over straight-bre fabrics: they can be stretched in both directions during forming, increasing the potential for forming
*

Corresponding author. E-mail address: d.bhattacharyya@auckland.ac.nz (D. Bhattacharyya).

complex and deeply curved components [13], have excellent impact, fracture toughness and energy absorption properties [46], and (in the case of 3D knits) exhibit improved out-of-plane mechanical properties [2,7,8]. Fig. 1 compares knitted fabrics with some of the other main existing reinforcement types. Until now the bulk of literature has concentrated on characterising mechanical properties of these materials, both thermosetting [6,9,10] and thermoplastic [11,12] based knitted composites. Modelling the mechanical properties (mainly stiness and tensile) has also been popular [13 17]. However, only limited literature exists on the forming property characteristics of knitted fabric reinforced thermoplastics [1820] and their processing properties are still poorly understood. In fact, most of the literature on forming properties deal with unidirectional [21,22], mostly

1359-835X/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.compositesa.2005.12.029

1898

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

Nomenclature Xcf m l G E W Fi Ti d#i dli Msi Mti knit structure size correlation factor bending modulus of yarn knit loop length shear modulus of yarn energy work force in ith element torsion in ith element axial rotation in ith element elongation in ith element s-axis nodal moment in ith element t-axis nodal moment in ith element Struct structural energy KnitStruct structural energy of the knitted fabric specimen Subscripts TOT total energy component INT internal energy component EXT external energy component KIN kinetic energy component AXIAL axial elongation energy component TORSION torsional energy component SBM s-axis bending moment energy component TBM t-axis bending moment energy component SSF s-axis shear force energy component TSF t-axis shear force energy component

Superscripts SIF sliding interface friction energy SIS sliding interface spring-contact energy

woven [2327], and to a lesser extent, braided reinforcements [28] rather than knitted reinforcements. Although the actual mechanism of sheet forming can be quite complex, a general understanding of the process and material characteristics can be gained through a series of simple experiments including hot-tension, vee-bending, dome-forming, cup-drawing, and picture-frame experiments. The experiments are designed to subject the material to selective modes of deformation allowing a more systematic approach to the analysis. One way of obtaining a physical measure of the forming behaviour is to develop kinematic models based purely on geometrical changes. These changes can be measured using grid strain analysis (GSA) where grid points before and after the forming process are used to calculate forming strains. The experiments not only provide general understanding, but also serve as a check for any predictive work that is done. When sucient data regarding the material and other processing parameters are available, numerical simulations

can provide timesaving predictions of the forming behaviour. For materials with complex reinforcing structures, such as knitted fabrics, the material model must be able to accommodate the most important parameters so that the material can be simulated accurately. An examination of the reinforcement, since the forming behaviour of a textile composite is governed by its reinforcing textile structure, can identify these important parameters. When forming with these materials, it is known that the best results can be achieved using a low matrix viscosity allowing lubricating eects to become signicant. Therefore, understanding the deformation behaviour of the reinforcing structure alone is of critical importance. To investigate these parameters physically would be very dicult, because of the scale and level of detail required; however, using advanced numerical simulations, a large quantity of information may be obtainable. The overall approach of this study involves both experimental investigations and advanced micro-numerical

Short Continuous fibres fibre mats


low

Woven UD Knitted &braided Non crimp laminates fabrics fabrics fabrics


high

Stiffness & strength

Interlaminar fracture toughness

high

low

high

low

Processability of complex shapes

Degree of isotropy Material + production cost

high

low

low

high

Fig. 1. Overview and comparison of some composite properties of the main existing reinforcements (modied from Ref. [40]).

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

1899

simulations to form the knowledge base which is used for predictive macro-numerical forming simulations. 2. Knitted fabrics Knitting is the process of manufacturing textile structures with a single yarn or set of yarns moving in only one direction. Unlike weaving where the yarns crossover one another, knitted fabrics are produced by looping the yarn through itself to make a chain of stitches which are then connected together, Fig. 2. There are two types of knitted fabric, those produced by weft knitting and those produced by warp knitting, where warp and weft refer to the knitting directions. In a weft knit the fabric is essentially produced with one yarn whereas in a warp knit the number of yarns used depends on the required fabric width. The knitting directions are also termed wale and course as shown in Fig. 2. 2.1. The milano and rib knit structures 3.1. Prepreg ow mechanisms In this study, two weft knitted structures, the full milano and 1 1 rib are investigated, see Fig. 3. Both fabrics are essentially 2D but unlike the plain knitted structure, knit loops occupy two planes, with a mirrored geometric conguration that ensures the fabric is balanced. One reason for these structures to be chosen is that the rib structure generally possesses an unusually high degree of elasticity. As a result, yarns made form high performance bres with very little inherent elasticity can produce a reinforcing fabric with decent stretchability. For the milano rib structure this is even more the case. 3. Textile composite deformation mechanisms The hierarchy of deformation modes for this family of composite materials can be broken down into three categories: prepreg ow mechanisms, macro-level fabric deformation modes and micro-level fabric deformation modes, each When the textile fabric reinforcement is combined with the matrix to form the composite prepregs, a set of deformation modes are introduced, which may be referred to as top-level deformation modes since they involve the movements of the reinforcement (macro- and micro-level fabric deformation mechanisms) and the matrix. They are, in fact, the conformation modes of the composite prepreg sheet or group of sheets, as is usually the case, during the forming process. The hierarchy of the top-level deformation modes consist of transverse ow, resin percolation, interply shear and intraply shear as summarised by Cogswell [29] and revisited by many others [30]. 3.2. Macro-level fabric deformation modes The four types of macro-level fabric deformation modes, as shown in Fig. 4, describe the deformations observed when looking at the fabric as a whole. However, the way in which each fabric complies to these modes is dierent and can be attributed to the deformations occurring within the textile structure itself. These sub-structure or microlevel deformation modes are the real mechanisms behind textile deformations and need to be identied in order to understand the materials behaviour. 3.3. Micro-level fabric deformation modes Micro-level fabric deformation modes exist through the interaction of structured yarns within the fabric. Fig. 5 shows what are generally believed to be the eight microlevel deformation modes for textile fabrics. Inter-yarn slip as shown in Fig. 5(a) occurs when the yarns that construct the fabric move over one another. It is one of the modes of deformation belonging almost exclusively to knitted fabrics. In this mode of deformation the

Fig. 3. Schematic of (a) 1 1 rib and (b) full milano rib structures.

of which contains a number of dierent mechanisms. This hierarchical denition is a method by which a textile composite material can be studied at dierent levels of its material structure.

Fig. 2. Schematic of knitted fabrics [9]: (a) weft knit, (b) warp knit.

1900

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

(a)

(b)

(c)

(d)

Fig. 4. Macro-level fabric deformation modes: (a) transverse compression, (b) in-plane tension, (c) in-plane shear, (d) out-of-plane bending.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

Fig. 5. Micro-level fabric deformation modes: (a) inter-yarn slip, (b) inter-yarn shear, (c) yarn bending, (d) yarn buckling, (e) intra-yarn slip (inter-bre friction), (f) yarn stretching, (g) yarn compression, (h) yarn twist.

friction between the yarns becomes important since it determines where the onset of buckling will be as well as the magnitudes of the forming forces required. Fortunately the matrix and bre chemical sizing (coatings) usually lubricate the yarn to help this mode of deformation. Inter-yarn shear is a common mode of deformation in many woven fabrics. This is where the yarns rotate about their crossover points to accommodate the required deformation, Fig. 5(b). In fact, this type of mechanism has been reported to occur also in multi-layered continuous brereinforced composites as outlined by Krebs et al. [26], Martin et al. [31] and Christie [21] and is commonly referred to by many researchers as the trellising eect [3235]. In knitted fabrics depending on the orientation of the reinforcement large in-plane tension and in-plane shear can be accommodated, but not in the same direction. Yarn bending or straightening, shown in Fig. 5(c), is the most signicant deformation mode in many textiles. It is the most inuential mode in knitted fabrics because of the knit loop geometry. Straightening also occurs to a lesser degree in woven and braided fabrics depending on the amount of crimp or yarn undulation present in the fabric structure.

Fibre buckling is an unfavourable mode since material movement through this mode creates what are considered as defects, although it is quite dicult to observe these with complex structures such as knits and braids. Out-of-plane buckling usually occurs when the in-plane modes cannot accommodate the required deformation. In-plane buckling can also occur but is less likely due to in-plane geometric constraints, Fig. 5(d). Intra-yarn slip shown in Fig. 5(e) coupled with yarn bending, Fig. 5(c), are the biggest contributors to a textile fabrics force displacement curve. Intra-yarn slip is where the continuous bres within the yarn slide past one another along the length of the bre because of changes in bre curvature during bending and unbending. Yarn stretching, Fig. 5(f), while not so prominent at early stages of fabric deformation, is certainly present and becomes a signicant contributor to the deformation at larger strains. Although the reinforcing bres used in composites are relatively brittle and exhibit very high stiness moduli, strains of up to 5% through this mode can still occur. Another fabric deformation mechanism to consider is yarn compression, Fig. 5(g), where forces at yarn crossover

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

1901

points compress the laments in the yarn and cause them to atten out and conform to the curvature of perpendicular yarns. Like bre stretching this can also be considered relatively insignicant and only really starts to contribute to the load extension curve once the aforementioned mechanisms have been exhausted. Finally, yarn twist, Fig. 5(h), which has been observed in knitted fabrics and not so much in woven fabrics contributes further resistance to fabric deformation. This is where the yarn is subjected to one full turn during the manufacture of the fabric in order to create the looping structure of the knit. The twist creates a resistance to the increase in yarn curvature during fabric deformation. 3.4. Textile fabric forcedisplacement curve The relative importance of each of these deformation mechanisms is fabric specic and in some cases certain deformation modes may not be utilised at all. During the deformation of a textile fabric, combinations of these mechanisms occur simultaneously and the inuence of each mode changes continuously throughout the entire deformation. Fig. 6 shows the force displacement curves for both woven and knitted fabrics in general, again for comparison, with regions of the curves identied to show where certain deformation modes are thought to be of greatest inuence. It can be seen that the curves for both the woven and knitted fabrics follow similar trends. Intra-yarn or interbre friction (a) is most inuential at initial stages of both curves starting out as the static friction that needs to be overcome to initiate the sliding of long bres past one another. For both fabrics the next major regions (b) and (c) are caused by bending/unbending and resistance to twist. Friction is still present from here on but is in the form of a lower dynamic friction force. Owing to its structure, a knit-

ted fabric has more curved yarn to extend whereas a woven fabric, no matter what its degree of crimp, has far fewer. For knitted fabrics, inter-yarn slip (d) also contributes to the shape of the curve in this central region starting at around 10% extension and ending once the forces at the yarn crossover points become too large. Finally yarn compression and extension (e) and (f), no doubt present throughout the entire extension, become most dominant at the latter parts of the curves and together can contribute up to 5% of the total extension of a knitted fabric. It is important to note, that while the dierent regions of the forcedisplacement curves shown in Fig. 6 have been drawn with similar orders of magnitude for explanation purposes, this is usually not the case in practice. For example, actual forcedisplacement curves of woven fabric can look quite steep and linear indicating that regions (e) and (f) are the most dominant. Furthermore, for most fabrics, region (a) will be signicantly smaller than the rest of the curve. 3.5. The role of the matrix With the previous three sections in mind it is interesting to consider what role the thermoplastic matrix plays. Besides allowing for the top-level or prepreg deformation modes to occur it will, depending on its physical state, aect the behaviour of the lower level deformation modes in different ways. As a result, the forming characteristics of the material may change to become more favourable or less. This has been the basis for a large body of work done by Duhovic [36] which investigates the forming characteristics of knitted fabric composites by subjecting them to a number of in-plane, single curvature and double curvature forming experiments. From these experiments it has been concluded that the elastic behaviour of the molten composite is very similar to the dry fabric at low forming rates (<100 mm/ min) and temperatures above 180 C. It was also shown in the double curvature forming experiments that choosing a higher forming temperature resulted in a more uniform strain and thickness distribution. In fact, in all the cases the matrix could be considered to somewhat restrict the overall bre movement. It is also interesting to note that in some cases the molten matrix might not inltrate between bres and the bre sizing provides lubrication, although this would not be the case for commingled yarns. However, in both cases the frictional energy component can be expected to be inuenced most by these phenomena. 4. Modelling the manufacture of the reinforcement architecture

Force

Woven Fabric * f

Knitted Fabric *

e b d

a c

Displacement (in warp & weft directions)


Fig. 6. Textile fabric force displacement curve. *Note that the orders of magnitude for some regions and the scale of the two curves have been modied for explanation purposes.

In Section 3 it has been established that the most dominant factors inuencing the sheet forming behaviour of knitted fabric thermoplastics are associated with the micro-mechanics of their reinforcing structures. Therefore, understanding the deformation behaviour of the reinforcing structure on its own becomes extremely important. In

1902

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

this section, a model is developed to quantitatively analyse the contributions of the deformation mechanisms described in Section 3. By setting up a model exible enough to evaluate these mechanisms for some specic knit structures, more accurate simplied models can be developed for describing the forming behaviour of textile composite materials in a nite element software such as PAMFORMTM. Furthermore, the model could be developed to analyse the behaviour of the knitted reinforcement in dierent forms of the matrix (i.e. molten and solid), along with their failure modes, which could also be analysed in detail. There may also be application in the textile industry for the design of atbed knitting machinery. Although the knit manufacturing model is capable of producing models of many dierent weft-knit structures; the number is only limited by what could be produced on actual atbed knitting machinery; in this study, the 1 1 rib structure has been chosen to develop the model. 4.1. Model set-up The manufacture of knitted fabric is a high-speed dynamic contact problem, where knitting needles move back and forth at speeds of up to 1.5 m/s. This makes knit manufacture particularly suitable for modelling using the explicit dynamics code commonly utilised for crash and forming simulations. PAMCRASHTM has been utilised because of its extensive range of material models and contact algorithms most suited to the knitting process. Most of the software functionality required in a crash simulation is also required for knitting (e.g. beam self contact, function on/o sensors, simulation restart, etc.). Only a small quantity of the fabric is required for the analysis, therefore the number of needles used in the simulation has been limited to ve. On real weft knitting machinery there are hundreds of needles, the actual number that are engaged is dependent on the desired fabric width and structure. For the 1 1 rib however, it is possible to produce a coherent narrow strip of fabric, which captures the repeating unit and can be used for experimental

comparisons, using only ve knitting needles, as shown in Fig. 7. 4.2. Model input: knitting machine parameters Simulating the operation of a atbed-knitting machine requires in-depth knowledge of all the physical parameters that make the machine work. Some of these important knitting machine parameters include: the relationship between the knitting bed movement and displacement proles of the front and back knitting needles or the cam proles, the needle spacing or gauge, needle and bed geometries, yarn feed friction and take-up spring stiness, fabric take-up velocity and take-up spring stiness, knitting needle latch kinematics and friction as summarised in Table 1. While many of the knitting machine parameters can be modelled directly, the yarn feed mechanism, including feed friction and the take-up spring, is simulated using non-linear bar elements to simplify the model. Each lament in the yarn uses its own non-linear bar element describing the yarn feed properties, this way the denition is not dependent on the number of laments and more or less number of laments can be added depending on the computer resources that are available. Parameters such as the fabric take-up velocity and the yarn feed friction, which are dicult to measure physically, are estimated and adjusted according to the visual quality of the resulting knit, just as is done by technicians with real knitting machinery (an insuciently large numerical value for the fabric take-up velocity causes poor loop formation, tangling and needle jamming). The fabric take-up mechanism, a spring-loaded rotating roller, is simulated by moving springs attached to the take-up bar, applying a near constant tension. All ve knitting needles include kinematic pin joints that simulate the needle mechanism required to produce the fabric structure. The needle latches open and close according to the movements made by the needles and contacts encountered by the needles against the yarn and needle latches against the main body of the needle itself. A needle latch rotational friction resistance is also prescribed to restrict latch movement under its own inertial forces. 4.3. Model input: material property parameters Material property parameters form another important part of the model. Fortunately, continuous lament glass bre yarn can be accurately represented as a purely linear elastic material. Each lament in the yarn is represented by a series of interconnected linear elastic circular beam elements whose bending, tension and torsion forces are transmitted between one another. The element size, 0.2 mm, has been carefully chosen to allow accurate representation of the fabric geometry but also to keep the solving time reasonable. Other elements of the machinery such as knitting needles and the machine bed are treated as rigid bodies and the information on stresses and strains in these

Fig. 7. Real ve needle knitting of 1 1 rib weft knitted fabric.

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915 Table 1 Summary of important knitting machine parameters

1903

1904 Table 2 Yarn material properties Material property Filament diameter Density, E-glass Poissons ratio, m Tensile modulus, E Second moment of area, Ix Second moment of area, Iy Polar second moment of area, J

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

Value 17 lm 2.54e06 kg/mm3 0.2 73 GPa 4.1e09 mm4 4.1e09 mm4 8.2e09 mm4

elements is not required. A summary of the yarns material properties is given in Table 2. To simulate contact between the glass laments and other elements of the knitting machinery, the code uses contact algorithms that check for penetrating nodes within a space around each element. In this simulation the space is dened using the average diameter of the laments, 17 lm. Any penetration is then resisted by a contact stiness (lamentlament, lamentneedle compression stiness in the presented case) calculated by averaging the tensile moduli of the two materials in contact [37]. For additional control, a scaling variable called the penalty scale factor (PSF) is also introduced. Using the self-impacting contact type, contacts between individual laments, laments and the knitting machinery as well as laments contacting themselves at dierent points can all be accounted for. 4.4. Model input: non-physical parameters Apart from the physical parameters many non-physical parameters, such as the time step scale factor (TSSF), contact search accelerator (CSA), penalty scale factor (PSF) and material damping factors, play important roles in the simulation ensuring solution stability and results within a sensible timeframe. A list of the most important non-physical parameters and their descriptions is presented in Table 3. In explicit dynamic nite element procedures, the stable time step is calculated by using information about the element size, material density and stiness. To ensure the stability of the solution this value is multiplied by the TSSF. For large strain simulations using shell elements, such as sheet metal forming, a suitable TSSF usually lies between 0.7 and 0.9. For crash simulations where elements undergo more radical movements during a single time step, the factor may need to be set as low as 0.3 [37]. In the knitting
Table 3 Important non-physical parameters Important non-physical parameter Time step scale factor Contact search accelerator Penalty scale factor Material damping ratios Abbreviation TSSF CSA PSF Description Time step control multiplier Controls search frequency per n cycles Contact stiness multiplier based on E values Controls material oscillations

simulation where only bar and beam elements are used, the factor lies between 0.1 and 0.2. The reason for such a low value may arise because of the ambiguity of onedimensional element nodal rotations. For example, if a beam element undergoes rigid body rotation larger than a certain value during one time step then it may be unclear in which direction that element has rotated in order to get to that position. It is this shortcoming of the Lagrange (or even Updated Lagrange) formulation when using bar or beam elements that is the main factor inuencing the solution stability in a knitting simulation. This is discussed in more detail in Section 4.7.3. With the large number of contacts involved, the contact search accelerator (CSA) is another important non-physical parameter determining the frequency of the contact search per n number of time steps. To ensure contacts are detected and maintained throughout the simulation, the search is performed at every time step. 4.5. Simulating the mechanics of the knitting process Given the linear density of the yarn, the density of E-glass and lament diameter, the number of laments in the real yarn can be calculated as approximately 120. While the current simulation uses only 20 laments to help reduce the solving time, a forcedisplacement curve of the resulting simulated fabric should exhibit a curve similar to that of the actual fabric (which contains 6 times higher the number of laments) only lower in magnitude. The yarn itself is modelled as a hexagonal close packed arrangement of all the laments, a simplication over real yarn, which is usually spun, or air textured to provide lateral cohesion and resistance to damage during knitting [38]. However, it is even possible to simulate these characteristics. Extra boundary conditions can be applied to the ends of the yarn or the laments can be assembled together with centre

Fig. 8. Initial state of knitting simulation.

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

1905

distances smaller than their diameters, causing the laments to y apart initially, as if under the external force of an air jet. A simplied version of the initial state of the simulation is shown in Fig. 8. The duration of the entire knitting simulation is 50 ms, somewhat faster, around 5 times, than practically achievable knitting speeds on conventional machines due to the loss in yarn feed control and dangerously high needle bed movements. However, simulating at a faster than normal knitting speed is necessary in order to keep the solving time reasonable. Fig. 9 shows the production stages of the simulation (see Electronic Annex 1 in the online version of this article for the dynamic simulation), which performs two and a half full cycles, or 5 passes, producing enough fabric

for the second stage of the analysis where all the knitting machine elements are stationary and a numerical tensile test of the specimen is performed. 4.6. Model verication 4.6.1. Geometrical comparisons The rst and most obvious way of checking the simulation is by geometrical comparisons. Visual checks of the geometry during the knitting process are of particular importance and are analogous to the visual checks made by knitting machine operators. Any problems are almost certainly due to incorrect knitting machine parameters or in the simulation case, incorrect boundary conditions.

Fig. 9. The six stages of the knitting simulation.

1906

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

Fig. 10. Geometrical comparisons of complex 1 1 rib formation.

Fig. 10 shows plan and isometric views of the real and simulated ve-needle knitting process showing very close structural similarities. Further comparisons can be made after the knitting procedure is completed on a relaxed portion of the fabric specimen, where a relaxed specimen is one that is not acted upon by any forces other than its own internal reaction forces. For the numerical case, the tensioning bar used dur-

ing the knitting phase was relieved until it gave a at-lined force reading of 0.06 N. The geometry data was then exported back into a CAD package for accurate measurement. The physical specimen was analysed and measured using microscopic digital images taken at 10 magnication. Fig. 11 shows the two specimens in their relaxed state along with their average loop height, width and length, where the averages reect measurements taken from nine fully formed

Fig. 11. Comparison of loop geometry for: (a) numerical and (b) physical specimens.

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

1907

loops across three rows of both specimens. While the two specimens were produced using exactly the same knitting parameters, their resulting 1 1 rib geometries look quite dierent with the average loop height, width and length all measuring larger in the physical specimen. One obvious reason for the dierence is that the physical yarn contains six times the number of laments used in the numerical specimen. Therefore, its signicantly higher yarn bending rigidity results in a dierent relaxed state geometry and force displacement behaviour of the specimen. Another reason why the structures look dierent can be attributed to friction and shear resistance between individual bres (inter-bre shear). The fact that the numerical specimen contains no inter-bre frictional restraint, which is also a function of the number of bres in the yarn, tends to suggest that the forces here are large enough to have an inuence at least on the relaxed geometry of the fabric. That is, the order of magnitude of the forces generated by interbre friction, is comparable to the forces generated by bending for the specimen in its relaxed state and is therefore a signicant factor in determining what the geometry of the relaxed structure should look like. A diculty arises from the fact that the specimens, in spite of being knitted using the same knitting machine parameters, exhibit signicant relaxed state geometric discrepancies and a method to compare knitted structures made from yarn of dierent bending rigidity and dierent loop sizes needs to be identied. A summary of the generalised behaviour and the mechanics of knitted fabrics by Grosberg in Hearle [38] highlights an important observation made by several researchers relevant to this course of study. It has been found empirically that the load to give a xed extension is proportional to either m/l3 or G/l3, where m and G are the bending and shear moduli, respectively, and l is the knit loop length (or average knit loop length, l = Lav) [38]. Depending on the type of bres that are involved (i.e. continuous or discontinuous) and their frictional properties, the elongation resistance is largely due to bending energy changes or shear energy changes [38]. In the case of high modulus continuous lament yarn, the relationship between the forcedisplacement curves generated from 1 1 rib fabric yarn containing dierent number of laments and dierent loop lengths (which occurs as a consequence of having dierent numbers of laments since the loop length or loop size depends on the yarn bending stiness) seems more likely to be correlated using m/l3. However, from the diculties experienced during knitting it is known that the friction coecient between E-glass bre laments along with their sizing is very high. The loop length L or average loop length Lav, two measures commonly used by textile researchers and shown in Fig. 11 become particularly important now, since they can be used to calculate the correlation factor Xcf, dened in Eq. (1). X cf m l3 or G l3 1

Using an Xcf governed by m/l3 and the geometrical measurements made on both the experimental and numerical specimens in Fig. 11, it is shown in the next section how well the simulation represents the real life system without incorporating the eects of inter-bre or intra-yarn friction. 4.6.2. Experimental versus numerical forcedisplacement curves To provide a form of experimental verication which would prove the validity of force and energy readings taken from the simulation, warp direction tensile tests were carried out on 68 tex E-glass bre yarn 1 1 rib fabric specimens, 5 needles wide, with a specimen length dened by ve full knitting cycles. The set-up is shown in Fig. 12. To prevent any distortion or damage to the knitting structure, the hat-shaped specimen holders were placed into position during the knitting procedure. The experiment used a 20 N load cell and the average curve for ve tests is shown in Fig. 13. Note that the specimens show a small amount of mechanical property variation as no two loops had exactly the same geometry and might have incurred a varying degree of yarn damage during knitting. The curves also show that the initial inter-bre friction, or region (a) in Fig. 6, is a relatively small contributor to the total energy. However, the overall behaviour is similar to that shown in Fig. 6. Fig. 14 compares the numerical and experimental force displacement curves. Both curves follow a similar trend and only appear to be dierent in size. Using an Xcf value of 1.64, calculated from the known and measured ratio values of m(6) and l(1.54), respectively, the experimental curve is redrawn and compared with the numerical results to show a very good agreement. The fact that the curves

Fig. 12. Test set-up for 1 1 rib strip tensile test.

1908

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

load of the numerical curve. The drop o in load exhibited in the numerical curve is not due to failure in the laments themselves but is a result of a boundary condition failure at the yarn feed hole. At a certain level of tensile loading, more yarn begins to pull through the yarn feed hole. It was dicult to fully constrain the specimen as the numerical procedure only allowed reallocation of the boundary conditions rather than a complete redenition. The actual readings from the simulation are presented in Section 4.7. 4.7. Detailed investigation of the mechanisms In Section 3.3 eight micro-level fabric deformation mechanisms were identied and discussed in detail. These mechanisms can be separated into two further categories, those inuenced by friction and those inuenced by the material properties and geometry of the bres. Inter-yarn slip, inter-yarn shear and intra-yarn slip are all forms of interbre friction, the dierence being that inter-yarn slip and inter-yarn shear involve the relative translation or rotation of groups of bres whereas intra-yarn slip considers the axial sliding movements within these group reference frames. Yarn bending, compression, stretching, twist and yarn buckling, which can be viewed as the reverse of yarn stretching or the result of axial compression upon a lament, are all directly related to the material properties and geometry of the bres and their energy contributions can be obtained directly from the simulation. 4.7.1. Energy derivation From the curves presented in Figs. 13 and 14, it becomes clear that friction forces play a small role in the deformation energy contribution of high modulus continuous bre 1 1 rib knitted fabric. Even though inter-bre friction eects are not active in the simulation, its signicance can still be assessed by comparing the dierence in the magnitude of the contact energy compared to the other energy components. To clarify which components of energy can actually be extracted from the simulation, a summary of the energy balance taking place is presented before proceeding any further. For the knitting, or specimen generation, Phase 1 of the simulation, the energy balance follows the expression shown in Eq. (2). ETOT EStruct ESIS ESIF EStruct W EXT INT INT INT KIN where ETOT EStruct INT ESIS INT ESIF INT EStruct KIN is the total energy present at any time in the system is the internal energy stored and absorbed by the material of the structure is the elastic energy stored by the sliding interface contact springs is the energy dissipated by the sliding interface contact friction (currently not available for beam contact in PAMCRASHTM) is the kinetic energy of the structure 2

Fig. 13. Average experimental FD curve for 1 1 rib specimen.

Fig. 14. Comparison of experimental and numerical FD curves.

match so closely in the absence of any provision in the simulation for inter-bre friction tends to suggest that for this type of yarn, contact and friction are very small contributors to the total deformation energy while bending plays an overwhelmingly dominant role. This is proved in the subsequent analysis performed in Section 4.7. Another interesting characteristic of the numerical curve and further proof of the validity of the curve, is that its gradient up until 3.5 mm extension falls below the experimental curve, again indicating the absence of a viscous friction eect between bres during yarn bending, which indeed has been left out of the simulation. The fact that no lament failure criterion has been set-up might also explain the steeper gradient near the peak

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

1909

WEXT

is the work done by externally applied forces (which only includes the velocity boundary conditions applied to the knitting needles)

The frictionless interaction between the laments means that Eq. (2) can be simplied to Eq. (3). 3 In the tensile testing phase of the simulation, Phase 2, the kinetic energy of the structure, EStruct becomes small and KIN Eq. (3) reduces to Eq. (4). ETOT EStruct ESIS W EXT INT INT 4

To examine the deformation mechanisms of the structure in detail, the internal energy stored and absorbed by the structure, EStruct , needs to be decomposed into its compoINT nents as shown in Eq. (5).
Struct Struct Struct Struct EStruct EStruct EStruct INT AXIAL TORSION E SBM ETBM E SSF ETSF

about both nodal points of each element and are dierent for the common nodal points of dierent elements. Section 4.7.3 gives a detailed explanation on the type of beam elements used. Finally, to dene a complete formula for the total deformation energy of the knit structure, EKnitStruct , the ideal case TOTAL would be to include ESIF and ESIS . ESIS is already obtainINT INT INT able, but ESIF is currently not available since friction has INT not been activated for contact type considered in this study. Nevertheless it is included for completeness and the total deformation energy of a knit structure can be dened as presented in Eq. (8). Note that ESIS and ESIF have been INT INT renamed to EKnitStruct and EKnitStruct , respectively, with all SIS SIF other quantities previously dened in Eq. (7). EKnitStruct EKnitStruct EKnitStruct EKnitStruct TOTAL AXIAL TORSION SBM EKnitStruct EKnitStruct EKnitStruct TBM SIS SIF 8

5 where EStruct is the total axial energy present at any time in the AXIAL system EStruct TORSION is the total torsional energy present at any time in the system EStruct is the total s-axis bending moment energy present SBM at any time in the system EStruct is the total t-axis bending moment energy present TBM at any time in the system EStruct is the total s-axis shear force energy present at any SSF time in the system EStruct is the total t-axis shear force energy present at any TSF time in the system For the beam theory used in PAMCRASHTM and by most FEA software packages, EStruct and EStruct are compleSSF TSF mentary quantities and should therefore not be added as part of the energy balance. As a result the internal energy of the structure reduces to Eq. (6).
Struct Struct EStruct EStruct EStruct INT AXIAL TORSION E SBM ETBM

4.7.2. Simulation results: energy contributions In Fig. 14 the force displacement curve for the numerical simulation has been shown to be in very good agreement with the experimental results with even the minor discrepancies readily explainable (i.e. the eects on the curve due to the absence of inter-bre friction). Therefore, the energy readings for the quantities, dened in Eq. (8), may be presented with a good level of condence. Fig. 15 shows the energy versus specimen extension for all twenty laments in the yarn. The energy curves show a small amount of variation, which is expected, but the trend for each is consistent. All of the curves exhibit a very low energy contribution at early stages of the tensile test before rising steeply towards tensile failure.

where EStruct AXIAL


n X F i dli ; for i 1; n 2 i n X T i d#i ; for i 1; n EStruct TORSION 2 i n X M si1 d#si1 M si2 d#si2 ; for i 1; n EStruct SBM 2 2 i n X M ti1 d#ti1 M ti2 d#ti2 ; for i 1; n EStruct TBM 2 2 i

Note that because of the way in which beam elements are formulated in PAMCRASHTM, moments are calculated

Fig. 15. Individual lament readings for axial elongation energy. Note: tension in a lament depends on its position within the yarn and the degree of contact with the needle.

1910

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

The next set of graphs shown in Figs. 17 and 18 show a very dierent trend. For the s-axis bending moment energy the curves show a more gradual increase in energy as well as an initial internal bending energy, created in the structure during the knitting process. At the peak energy level, the bending energy has increased by a factor of ve compared to its initial internal energy value and is four times larger than the maximum energy level of axial elongation. The t-axis (perpendicular to the s-axis) bending moment energy curves shown in Figs. 19 and 20 show a signicantly lower level of energy than the s-axis energy curves. Its

Fig. 16. Total yarn axial elongation energy.

In Fig. 16 the total yarn axial elongation energy curve is shown along with six sample data points identifying the model specimens state in the mesh plot key. Only after the fourth data point does the tensile energy begin to increase, once bres in the specimen begin to straighten. At the peak axial elongation energy a maximum stress of 2058 MPa and maximum axial strain of 2.8% are achieved in lament 687. The maximum axial stress and axial strain at any time during the tensile test, is 2198 MPa and 3.0% in lament 672. Both maxima occur at the edges of the specimen in regions of high curvature and contact. A comparison of these to the documented values of E-glass bre mechanical properties [39], shows that the values here are close to the failure range (24003450 MPa). Although the graphs show the characteristics of a material property failure, it must be noted that no failure criterion has been assigned to the model. It is possible to do so by using a dierent material model and the element elimination function in PAMCRASHTM but this further complicates the model. Failure is detected by identifying element stress levels exceeding the ultimate tensile stress. As mentioned already, this may explain why the numerical FD curve shown in Fig. 14 becomes steeper than its experimental counterpart at the latter stages of the tensile test. Fortunately, failure of the specimen is quite catastrophic, as shown in Fig. 13, and the total specimen failure in the simulation can be assumed once a single element reaches the ultimate tensile stress. The peak load and drop o which are exhibited in all of the graphs are the results of leftover yarn pulling into the specimen once a certain tensile loading is reached, increasing the size of one of the loops, and relieving the load on the specimen. This is a boundary condition failure rather than an actual material property failure but because it occurs at a suciently high load, valid tensile test readings for the simulation can still be made.

Fig. 17. Individual lament readings for s-axis bending moment energy.

Fig. 18. Total yarn s-axis bending moment energy.

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

1911

tensile test with bending moment energy increasing more about one axis compared to the other. While the contribution of the bending moment energy was always expected to be high, it was dicult to even guess how much of a part the torsional energy would play. Magnied images of the milano and 1 1 rib fabric specimens always display a certain level of twist in the structure even if the yarn itself does not exhibit much initial twist, which is what was assumed for the yarn in the simulation exercise. Figs. 21 and 22 show how small the contribution of torsional energy is, with an initial internal torsional

Fig. 19. Individual lament readings for t-axis bending moment energy.

Fig. 21. Individual lament readings for torsional energy.

Fig. 20. Total yarn t-axis bending moment energy.

initial internal energy is half of that for the s-axis bending moment energy and the increase in energy during the tensile test is very low, reaching a peak value, which is four times lower than that of the s-axis bending moment energy. Since the laments have been dened with a circular symmetrical cross-section and all have their orientation nodes pointing in the same direction, this indicates that to create the 1 1 rib fabric, the yarn needs to be bent twice as much about one axis compared to the other. This phenomenon can be used to measure how three-dimensional the structure is. This characteristic continues during the

Fig. 22. Total yarn torsional energy.

1912

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

energy level of 0.01 mJ, twenty to forty times less than that of the bending moment energy. It is also interesting to note the way in which this energy component changes, with its almost linear increase up until point 3 and then an exponential increase, indicating a certain amount of dependence on the axial elongation energy curve. If a certain degree of yarn twist was incorporated into the model, it may be anticipated that this would not change the shape of the curve by much, but would only shift its position upward along the energy axis. The nal energy component curve presented here shows the total yarn contact energy, Fig. 23. Its characteristics are similar to that of the axial energy curve; however, it does not increase as much and is about four times smaller at its peak energy level. A closer inspection of this curve, compared with the axial energy curve, reveals an average energy level 4.7 times smaller after 2 mm extension (13.3 times smaller prior to 2 mm extension). The contact energy curve can also help give a clue as to how much energy is dissipated through friction eects. Using a coulomb friction law with a coecient of friction of 0.5 and assuming that the contact force is the normal force acting on an element, the only dierence between friction and contact energy is dl, the movement of the bres. For friction energy, sliding displacements are much greater than the compression displacements experienced by the contact interface springs. The magnitudes of these displacements can be sampled in the model by looking at relative movements of nodal points at the heads of the knit loops and by measuring how much loop crossover points have moved with respect to one another. Typical crossover movement is around 0.60.8 mm while inter-bre movement, depending on the location, ranges from 0.02 mm at the sides of the loops and 0.1 mm at the head of the loop. These crossover movements are 10100 times larger than the contact spring compression displacements, so the fric-

tional energy would certainly be of more signicance than the contact energy. However, the fact that the force displacement curve shown in Fig. 14 agrees so well, clearly shows that the frictional energy is not as dominant as the bending moment energy. An estimated curve would probably t somewhere between the yarn contact and yarn axial elongation energy curves with shape characteristics similar to these two curves and beginning at an initial energy level close to the torsion component. Due to the unconventional nature of some of the information required, most of the quantities were calculated element by element using the expressions for the various energies as presented in Eq. (7) and programmable macros to generate the required curves. To check the correctness of the energy calculations, all the energy components shown in Eq. (8) without the terms EKnitStruct and EKnitStruct were SIF SIS added and compared to the total internal energy output given by PAMCRASHTM minus the non-yarn energy components of the model (i.e. knitting machine elements). Fig. 24 shows the graphical form of Eq. (8) and the quantitative results of each of the components versus specimen extension. From Fig. 24, it is evident how dominant the bending moment energy is. The summation of the s- and t-axis bending moment energy curves practically t the shape of the total internal energy curve except in the region closest to tensile failure. The internal energy calculated by the software can be seen to match exactly with the sum of the calculated energy component curves, verifying that the calculations have been done correctly. An interesting question, which arises as a result of this analysis, is that with the given information, is it possible to determine which of the deformation components causes strain failure? The obvious choice is the bending moment stress, but although the energy readings for this component

Fig. 23. Total yarn contact energy.

Fig. 24. Comparison of yarn deformation energy components.

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915

1913

are the highest, this does not necessarily mean that this is the component that initiates failure. Axial elongation energy while much lower also reaches stress levels very close to the tensile failure as discussed after Fig. 16. An analysis of the bending moments shows maximum values that exceed the ultimate tensile stress of E-glass by a factor of 4.5 (maximum s-axis bending moment occurs at n1 5.28994e06 kN mm, rUTS for E-glass taken as 2400 MPa). Furthermore, the UTS value is rst reached at fairly modest levels of specimen elongation. The concern here is that the level of mesh renement is not good enough to give the correct readings for bending moments. However, it is interesting to analyse how the force displacement readings for the numerical and experimental case match so well. These issues are addressed and discussed in following section. 4.7.3. Beam elements in PAMCRASHTM and discussion of results The knitting procedure subjects a yarn to large displacements and the subsequent geometric structure has many regions of high curvature. To cope with such large displacements PAMCRASH uses the Belytschko beam element formulation, part of a family of structural elements that employ the co-rotational technique [37]. In a large displacement formulation, the idea is to separate deformation displacements from rigid body displacements since only the deformation displacements that generate strain energy. In order to perform the separation it requires a complete description of the deformed body at its current and reference (previous time step) conguration (i.e. the orientation and location of all elements and nodes at both congurations). The co-rotational technique assigns a co-ordinate system to each individual node and element. The coordinate system attached to a node is termed the body coordinate system (xb, yb, zb) and moves with the nodes while the element coordinate system (xe, ye, ze) is dened rstly by its x-axis, which originates at node n1 and then through n2 of the element. The remaining axes of the element coordinate system are dened by the elements principal inertial axes (i.e. using an orientation node n3), Fig. 25. Body and element coordinate system unit vectors are used to relate the translational and rotational transformations between the global coordinate system and both the coordinate systems. For a rigid body rotation the unit vector of the element coordinate system will be the same in the initial and rotated conguration with respect to the body
Z Y

zb yb xb
X n1

ze ye xe
n2 n3

zb

coordinate system, the same applies for the other rigid body transformations. If the unit vectors are not the same then a deformation displacement has taken place. It is in this way that all the deformation components are calculated and subsequently the forces, bending moments and torques in each of elements. These are calculated using known quantities including the Youngs and shear moduli, second moments of area, the eective cross-sectional area in shear and element lengths. The reason why the force displacement readings for the numerical and experimental case match so well, while local moment readings seem inaccurate, can be attributed to the way in which connectivity between beam elements is addressed. Beam element moments are calculated by dening their nodal connection points as a type of hinges where the stiness of a hinge is dictated by whatever stiness and second moment of area values (about the two perpendicular transverse axes) are specied by the user. Therefore depending on which material model is used (in this study only linear elastic is used since glass bres adhere to this model very well) and how much the hinge node between the two beam elements rotates, determines the moment at the beam endpoint, the element itself remains perfectly straight. With this type of formulation the overall structural response of the numerical specimen approaches the exact response of the physical specimen as the number of elements used in the simulation approaches innity. If the beam element size is small enough then correct energy readings for the entire specimen should also be achieved. However, local moment readings will not be so accurate because of the inadequate mesh resolution in localised regions of high curvature. The axial moment calculation in the beam elements is based on Timoshenkos torsion theory, which is only relevant for very small rotations and when the axial rotation of an element is larger than what can be handled by theory in one time step, the large rotation error may arise. When a very small time step scale factor is used, this problem is eliminated. The fact that the simulation does not consider initial yarn twist and friction may explain why maximum axial force readings are fairly low. More twist means a larger transverse force applied to the laments. This coupled with inter-bre friction causes a signicant amount of energy loss. Even with a signicant amount of lateral pressure, if the friction coecient is zero, as it is in the current simulation, two adjacent bres will still slide over one another very easily. If friction was incorporated then its inuence could be closely analysed, and its contribution to the deformation energy could actually be quantied for dierent coecient of friction values. 5. Conclusions This study has investigated the core elements of knitted fabric composites forming behaviour, which is dictated by eight micro-level fabric deformation mechanisms, such as

yb xb
TM

Global coordinate system

Fig. 25. The co-rotational technique used in PAMCRASH elements.

for beam

1914

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915 [2] Mouritz AP, Bannister MK, Falzon PJ, Leong KH. Review of applications for advanced three-dimensional bre textile composites. ComposPart A: Appl Sci Manuf 1999;30:144561. [3] Mayer J, Haan JD, Reber R, Wintermantel E. Knitted carbon bre reinforced thermoplastics: an overview. In: Proceedings of rst AsianAustralasian conference on composite materials (ACCM-1), Osaka, Japan, 1998. p. 401. [4] Pandita SD, Falconet D, Verpoest I. Impact properties of weft knitted fabric reinforced composites. Compos Sci Technol 2002; 62(78):111323. [5] Reber R, Haan JD, Petitmermet M, Mayer J, Wintermantel E. Failure behaviour of weft-knitted carbon bre reinforced PEEK. In: Proceedings of rst AsianAustralasian conference on composite materials (ACCM-1), Osaka, Japan, 1998. p. 407. [6] Bannister MK, Khonder OA, Leong KH, Herszberg I. The eect of architecture on the impact properties of knitted fabrics. In: Proceedings of rst AsianAustralasian conference on composite materials (ACCM-1), Osaka, Japan, 1998. p. 402. [7] Miravete A, editor. 3-D textile reinforcements in composite materials. Cambridge (England): Woodhead Publishing Limited; 1999. [8] Tong L, Mouritz AP, Bannister MK. 3D bre reinforced polymer composites. Boston: Elsevier; 2002. [9] Ramakrishna S. Characterisation and modelling of the tensile properties of plain weft-knit fabric-reinforced composites. Compos Sci Technol 1997;57:122. [10] Houtte PV, Verpoest I, Gommers B. Analysis of knitted fabric reinforced composites. Part II. Stiness and strength. ComposPart A: Appl Sci Manuf 1998;29:1589601. [11] Haan JD, Reber R, Mayer J, Wintermantel E. Tensile properties of plain weft knitted carbon bre reinforced polyamide 12 composites. In: Proceedings of rst AsianAustralasian conference on composite materials (ACCM-1), Osaka, Japan, 1998. p. 403. [12] Putnoki I, Moos E, Karger-Kocsis J. Mechanical performance of stretched knitted glass bre reinforced poly(ethylene terephthalate) composites produced from commingled yarn. Plast Rubber Compos 1999;28(1):406. [13] Tan P, Tong L, Steven GP. Modelling for predicting the mechanical properties of textile compositesa review. ComposPart A: Appl Sci Manuf 1997;28A:90322. [14] Ramakrishna S, Hamada H, Cheng KB. Analytical procedure for the prediction of elastic properties of plain knitted fabric-reinforced composites. ComposPart A: Appl Sci Manuf 1997;28:2537. [15] Huang ZM, Ramakrishna S. Micromechanical modelling approaches for the stiness and strength of knitted fabric composites: a review and comparative study. ComposPart A: Appl Sci Manuf 2000;31: 479501. [16] Houtte PV, Grommers B, Verpost I. Modelling the elastic properties of knitted-fabric-reinforced composites. Compos Sci Technol 1996; 56:68594. [17] Ramakrishna S, Huang ZM, Teoh SH, Tay AAO, Chew CL. Application of the model of Leaf and Glaskin to estimating the 3D elastic properties of knitted-fabric-reinforced composites. J Text Inst 2000;91(1):13250. [18] Takano N, Ohnishi Y, Zako M, Nishiyabu K. Microstructure-based deep-drawing simulation of knitted fabric reinforced thermoplastics by homogenization theory. Int J Solids Struct 2001;38(3637): 633356. [19] Lim TC, Ramakrishna S, Shang HM. Improvement of knitted fabric composite sheet formability by simultaneous deep drawing and stretch forming. In: Proceedings of rst AsianAustralasian conference on composite materials (ACCM-1), Osaka, Japan, 1998. p. 409. [20] Duhovic M, Bhattacharyya D. Deformation mechanisms of knitted fabric composites. In: Friedrich K, Fakirov S, Zhang Z, editors. Polymer compositesfrom nano-to macro-scale. US: Springer; 2005. p. 265. [21] Christie GR. Numerical modelling of bre-reinforced thermoplastic sheet forming. PhD thesis in Department of Mechanical Engineering, The University of Auckland, Auckland, New Zealand, 1997.

inter-yarn slip, inter-yarn shear, yarn bending, yarn buckling, intra-yarn slip, yarn stretching, yarn compression and yarn twist. To analyse these mechanisms in terms of deformation energy they need to be grouped into ve energy components: axial energy, torsional energy, bending energy, contact energy and frictional energy. The energy contributions for a 1 1 rib knit specimen during stretching in the warp direction have been studied here by rst simulating the production of a narrow strip of 1 1 rib fabric and then verifying the model by comparing tensile tests of the physical and numerical specimens. Subsequently the individual energy contributions of the structures deformation mechanisms were evaluated. From the study it may be concluded that:  The knitting process for producing textile fabrics for manufacturing composites can be successfully computer simulated by using beam elements and the results of actual tensile tests carried out on 1 1 rib fabric agrees very well with the simulated results.  The simulation provides means of understanding the relative importance of individual micro-mechanisms that are practically impossible to isolate through laboratory experiments.  In the case of continuous high modulus bres such as Eglass, the load to give a xed extension is proportional to m/l3, where m is the bending modulus and l the average knit loop length. This allows the physical specimen, which contains six times the number of bres in the numerical specimen, to be compared with the numerical model.  Bending is by far the largest deformation mechanism contributing to the 1 1 rib structures total deformation energy and is at no point overwhelmed by any of the other deformation mechanisms. Bending is followed by torsion at low strain levels, which is surpassed by uniaxial tension at very high strain levels and nally the total yarn contact energy, which also surpasses torsion at very high strain levels.  Although the overall energy level readings for the entire specimen are correct, it has not been possible to predict which mechanism would be the cause of strain failure since localised bending stress readings are vulnerable to the mesh resolution in high curvature regions.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.compositesa. 2005.12.029. References
[1] Savci S, Curiskis JI, Pailthorpe MT. A study of the deformation of weft-knit preforms for advanced composite structures. Part 2: the resultant composite. Compos Sci Technol 2000;60(10):194351.

M. Duhovic, D. Bhattacharyya / Composites: Part A 37 (2006) 18971915 [22] Pickett AK, Cunningham JE. Numerical techniques for the preheating and forming simulation of continuous bre reinforced thermoplastics. SAMPE European conference and exhibition, Basel, 1996. [23] Cliord MJ, Long AC, Luca P. Forming of engineered prepregs and reinforced thermoplastics. The Minerals, Metals and Materials Society (TMS) 2001 annual meeting, New Orleans, 2001. [24] Boisse P, Gasser A, Hivet G. Analyses of fabric tensile behaviour: determination of the biaxial tension-strain surfaces and their use in forming simulations. ComposPart A: Appl Sci Manuf 2001;32(10): 1395414. [25] Ye L, Daghyani HR. Characteristics of woven bre fabric reinforced composites in forming process. ComposPart A: Appl Sci Manuf 1997;28A:86974. [26] Krebs J, Friedrich K, Bhattacharyya D. A direct comparison of matched-die versus diaphragm forming. ComposPart A: Appl Sci Manuf 1998;29A:1838. [27] Cherouat A, Billoet JL. Mechanical and numerical modelling of composite manufacturing processes deep-drawing and laying-up of thin pre-impregnated woven fabrics. J Mater Process Technol 2001; 118(13):46071. [28] Kuo WS, Fang J. Processing and characterization of 3D woven and braided thermoplastic composites. Compos Sci Technol 2000;60(5): 64356. [29] Cogswell FN. Thermoplastic aromatic polymer composites. Oxford (England), Boston: Butterworth-Heinemann; 1992. [30] Bhattacharyya D, editor. Composite sheet forming. Composite materials series, vol. 11. Amsterdam: Elsevier; 1997.

1915

[31] Martin TA, Christie GR, Bhattacharyya D. Grid strain analysis and its application in composite sheet forming. In: Bhattacharyya D, editor. Composite sheet forming. Amsterdam: Elsevier; 1997. p. 21745. [32] Cattanach JB, Cu G, Cogswell FN. The processing of thermoplastics containing high loadings of land and continuous reinforcing bres. J Polym Eng 1986;6:34561. [33] Rivlin RS. Networks of inextensible cords. In: Nonlinear problems of engineering. New York: Academic Press; 1964. p. 5164. [34] Green AE, Adkins JE. Large elastic deformations and non-linear continuum mechanics. Oxford (England): Clarendon Press; 1960. [35] Spencer AJM. Deformations of bre reinforced materials. Oxford (England): Clarendon Press; 1972. [36] Duhovic M. Deformation characteristics of knitted fabric composites. PhD thesis in Department of Mechanical Engineering, The University of Auckland, Auckland, New Zealand, 2004. [37] Pam-System-International, PAM-CRASHTM version 2000 level 18 notes manual, 2000. [38] Hearle JWS. Structural mechanics of bers, yarns and fabrics, vol. 1. USA: John Wiley and Son; 1969. [39] Owen MJ, Middleton V, Jones IA. Integrated design and manufacture using bre-reinforced polymeric composites. Cambridge (England): Woodhead Publishing Limited; 2000. [40] Houtte PV, Verpoest I, Gommers B. Analysis of knitted fabric reinforced composites. Part I. Fibre orientation distribution. ComposPart A: Appl Sci Manuf 1998;29:157988.

Das könnte Ihnen auch gefallen