Characterizing the processing and performance of aligned reinforcements during preform manufacture

Composites: Part A 2lA (1996) 241-253 Copyright 0 1996 Elsevier Science Limited Printed in Great Britain. All rights reserved 1359-835X/96/$1 5.00

ELSEVIER

Characterizing the processing and performance of aligned reinforcements during preform manufacture

A. C. Long*,

C. D. Rudd”,

M. Blagdon

and P. Smith

Department of Mechanical Engineering, University of Nottingham, Nottingham NG7 2RD, UK (Received 27 July 1995; revised 2 November 1995)

University

Park,

Liquid moulding processes are now finding a wide range of applications for both structural and semistructural components. This has been facilitated in part by the development of computer-aided engineering tools for structural analysis and process modelling. However, the accuracy of these tools is dependent on the available material property data, which are usually determined using two-dimensional flat plaque experiments. This approach may not be satisfactory for complex component geometries, as preform manufacture often results in large variations in both fibre orientation and volume fraction. In recent years, several authors have developed reinforcement deformation or ‘drape’ models that may be used to predict the fibre architecture at the design stage. These models are usually based on the assumption that reinforcement deformation is facilitated by inter-fibre shear, whereas in reality a number of mechanisms are available. In this study, reinforcement deformation is characterized using an automatic strain analysis system. This is applied to a number of generic components with increasing depth of draw, enabling the validity of the interfibre shear model to be established. The effects of deformation on reinforcement permeability and component structural properties are then established using experiments based on sheared reinforcements. (Keywords:liquid composite moulding; reinforcement deformation; process model&g)

INTRODUCTION Liquid moulding processes are becoming increasingly popular for the manufacture of structural and semistructural components from composite materials. One of the major benefits is perceived to be the separation of reinforcement preparation from the moulding cycle by the use of reinforcement preforms. The most common method of preform manufacture involves the forming of reinforcements between matched moulds. Reinforcements may consist of either random or aligned fibres, with the latter preferred for structurally demanding components. In particular, the use of engineered fabrics, consisting of aligned fibres stitched together to provide a zero-crimp reinforcement, is becoming increasingly popular for structural applications due to improved in-plane mechanical properties over woven materials. However, there are a number of associated problems, including the limited formability of the reinforcement materials which can lead to the occurrence of wrinkles in deep-drawn components’, Preform design is usually based on the required mechanical performance, which is routinely assessed * To whom correspondence

should

be addressed

using laminate analysis techniques. Processing considerations are also important, and to this end a number of computer simulations have been developed for the impregnation phase during liquid moulding2-5. The accuracy of both structural and process models is dependent on the quality of the input data. Property measurements are usually based on flat plaque data, which ignore the effects of fibre reorientation during preform manufacture. This may result in a significant departure from anticipated processing and performance characteristics. In recent years a number of authors have presented simulations of aligned fibre reinforcement deformation, usually based on a kinematic mapping of fibres to a three-dimensional surface6-“ . However, there appears to have been relatively little effort to characterize fabric deformation experimentally. Similarly, the effects of fabric shear on processing and mechanical properties have received little attention. The purpose of this study is to characterize reinforcement deformation and to establish the effect on reinforcement permeability and laminate mechanical properties. The deformation characteristics of reinforcements are established using an automated strain analysis system originally developed to

247

Preform processing

and performance

characterization:

(as demonstrated in Figure I). Fibre buckling, leading to preform wrinkling, can also occur when the fabric is sheared beyond its physical limit. This may be anticipated by measuring a fabric locking angle, which is likely to be a reinforcement-specific property. Assuming that inter-fibre shear is the only available mechanism, the fabric can be approximated as a pinjointed net. Based on this assumption, a deformation modelling package has been developed at the University of Nottingham”‘“. The position of every fibre is determined using a kinematic draping algorithm, based on solving the equations of intersection between the fabric and the surface to be draped (i.e. the preforming tool). The surface is modelled using triangular and quadrilateral flat patches, allowing the equations to be solved explicitly. To obtain a unique draped pattern, two intersecting fibres are constrained to the surface. These are defined by geodesic paths intersecting at the point of initial contact between the reinforcement and the preforming tools. The drape simulation has been applied to a number of complex components. An example is shown in Figure 2, which illustrates the predicted fibre pattern for one quarter of an automotive wheel-hub draped with a O/90” reinforcement grid. Maximum deformation is predicted midway along the circumference of the component, with a minimum inter-fibre angle of 28”. This level of shear would be expected to result in reinforcement wrinkling during preform manufacture.

Fibre slip

Fibre rotation or shear Figure 1 Fabric deformation mechanisms

Figure 2 Automotive wheel-hub draped with O/90” fabric

/

motor

Stepper Vertical ~;i;on;:g

Rotary transducer

screw

positioning screw

Preform

/Stepper

Turntable

PC containing:Data acquisition board CAMSYS

motor

Data acquisition & con tro/

t

software

Figure 3 ASAME automatic strain measurement system

characterize sheet metal formability. By applying this to engineered fabrics formed to various component geometries, it is possible to assess the accuracy of the kinematic model and to establish formability limits.

REINFORCEMENT

DEFORMATION

Drape modelling

Deformation modelling for aligned fibre reinforcements is usually based on the assumption that deformation is restricted to inter-fibre shear6-9. In reality, a number of mechanisms are available, the most dominant of which are shear and relative slip at the fibre crossovers

248

A. C. Long et al.

Experimental

analysis

Although several authors have developed deformation models similar to that described above, relatively little experimental evidence has been published to support this approach. To establish the validity of the model, a technique is required to measure the fibre positions and orientations within a deformed fabric. This could also be used to characterize the deformation mechanisms exhibited by a particular reinforcement. Previous studies by the authors have relied on comparison of the predicted and measured fibre volume fraction distributions”. Although this method has shown the model to be reasonably accurate for the wheel-hub described above, it can only provide average properties related to the fibre distribution and does not give information on local fibre orientations. A method has been developed based on measuring the deformation of a square grid printed onto the fabric, using a technique developed by the Formability Research Group at Ford Motor Company. This uses a system known as the CAMSYS Automated Strain Analysis and Measurement Environment (ASAME). This equipment, shown schematically in Figure 3, consists of a turntable on which the deformed specimen is placed and a digital camera positioned using steppermotor-driven lead screws. The deformation of the surface is measured by capturing two digital images of the grid from different orientations. The position of the

Preform processing

and performance

characterization:

A. C. Long et al.

(4

(4

Figure 4 Fibre distributions for 7 mm disc: (a) predicted pattern; (b) measured pattern

camera and the orientation of the turntable are monitored using a PC, allowing the three-dimensional coordinates of each grid intersection to be calculated using the CAMSYS software package. To establish the applicability of the above process to reinforcement deformation, experiments were carried out using Tech Textiles E-BXhd 936 f45” reinforcement. This material is of particular interest, as it is designed to be a ‘high drape’ reinforcement, in which inter-fibre slip is thought to be an important deformation mechanism. A 6.4mm square grid was screen printed onto one surface of the fabric. The material was deformed over a number of male forms using a vacuum bag, with the deformed specimens maintained using a polyurethane aerosol varnish. The deformed grids were then scanned using the ASAME system to provide the position of each grid intersection. These were post-processed to calculate the local inter-fibre angle and grid spacings. which were used to estimate the local degree of inter-fibre slip (calculated as a percentage engineering strain between intersections). Five 120mm diameter discs with heights of 7, 14, 19, 26 and 38 mm were used to establish the effect of depth of draw on fabric deformation. In each case, the fabric was considered in four quadrants, each of which was

Fibre distributions for 14 mm disc: (a) predicted pattern; (b) measured pattern

Figure 5

analysed using the process detailed above. Drape analyses were carried out for each disc to enable comparison of the predicted and measured fibre architectures. The results suggested that the accuracy of the pin-pointed deformation model decreased with increasing disc height. This is demonstrated by Figures 4-7, which show the predicted and measured fibre patterns for 7, 14, 19 and 26 mm discs, respectively (the deformed pattern for the 38 mm disc could not be fully scanned due to the presence of large wrinkles). In each case the surface is shaded to represent local shear, and for the measured patterns fibre segments are also shaded to indicate percentage inter-fibre slip. It is apparent that as the disc height increased inter-fibre slip became increasingly important, resulting in a reduction in fabric shear. The minimum fibre angles observed in each quadrant and at each disc height are plotted in Figure 8. This shows that although the predicted inter-fibre angle continued to reduce with increasing disc height, the measured angles levelled off at around 32” due to fibre locking. Any further deformation was accommodated by

249

Preform processing

and performance

characterization:

A. C. Long et al.

(4

(4

Figure 6 Fibre distribution for 19mm disc: (a) predicted pattern; (b) measured pattern

either inter-fibre slip or wrinkling. It should be noted that although there is clearly some discrepancy between predicted and measured minimum angles (ranging from 12” at 7 mm to 31” at 38 mm), the agreement for the majority of measurements was significantly better. The experiments described above suggest that although shear is probably the most important deformation mechanism, some engineered fabrics may exhibit a considerable degree of inter-fibre slip for deep-drawn components. More generally, the results suggest that the pin-jointed deformation model is applicable to relatively simple surface geometries where the fabric does not approach its practical locking angle. Both locking angle and the shear/slip relationship are thought to be fabricspecific phenomena, which may be characterized using material deformation data. It is hoped that the ASAME process can be used to characterize a range of materials, both to establish the validity of the kinematic deformation model and to assist in the development of a more accurate simulation.

250

Figure 7 Fibre distributions for 26 mm disc: (a) predicted pattern; (b) measured pattern

80 7

*Predicted

I

. Measured

70 -: 80 -s! P 4

50--

.

. 40--

e P L

-.-

30 --Mi;;i~&%t&&&&&

-.-

1--

20 -10 --I

OT 0

5

10

15

20

25

30

35

Depth of Draw (mm)

Figure 8

Minimum interfibre angles for discs of varying height

40

Preform processing

and performance

A. C. Long et al.

characterization:

PERMEABILITY As described in the Introduction, several authors have developed simulations of the injection phase during liquid moulding. These models rely on an accurate knowledge of the reinforcement permeability, which is usually determined using a simple flow experiment within a two-dimensional cavity. This approach neglects the effects of reinforcement deformation which may cause a significant departure from measurements based on undeformed (roll-stock) reinforcements. In particular, a local reinforcement shear causes changes in both fibre orientation and volume fraction. For a reinforcement with ply angle f0, the fibre volume fraction can be calculated from

u.

. 02.

\

00

~ 25

~~~~-

30

35

The permeability of a deformed zero-crimp reinforcement can be estimated by assuming the fabric to be composed of two unidirectional (UD) layers. The axial permeability (k,) of each layer is assumed to be related to fibre volume fraction by the following:

55

65

with ply angle: E-BXhd 936

s Experimenlal-The-

.

.

.

60

(0)

Figure 9 Variation in permeability reinforcement

.

?

? -\ .

(2)

It should be possible to determine the constants A and m experimentally, although in practice it is difficult to carry out flow experiments using truly UD reinforcement. Therefore the constants were chosen to provide close agreement between theoretical and experimental permeability values. As the transverse permeability (k,) is negligible in comparison with the axial value, it is assumed to be zero to simplify the analysis. The principal permeabilities of the combined bidirectional reinforcement can then be approximated by applying a transformation to the axial values and employing a simple addition rule3, giving the following for flow parallel to the major axis of the fabric (corresponding to the bisector of the two fibre axes): k, = k, co? 0

50

Ply

07

k, = A( 1 - Vr)‘”

*i;,e

40

(3)

Note that this is based on a simplified version of the expression proposed by Advani et a1.12, which should be used if experimentally determined permeability values were used in place of equation (2). To determine the validity of this approach, permeability tests were carried out using sheared reinforcements. A constant flow-rate radial flow arrangement was used, involving the injection of an SAE 30 oil into a 2 mm cavity at approximately 7 ml s-i. The principal permeabilities of the reinforcement were calculated from the pressure distribution within the cavity using the method described in our earlier paper”. Two layers of reinforcement fabric were used in each case, each of which was pre-sheared using a four bar linkage, with the fabric clamped to the laboratory bench along one edge. Two series of experiments were carried out based on different zero-crimp reinforcements:

01

0

~

25

_~

_~_

30

35

ALle

40

Ply Figure 10 Variation reinforcement

??

??

in permeability

50

55

60

65

(l3) with ply angle: E-LT 850

Tech Textiles E-BXhd 936 gmp2 style 3074 it 45” high drape reinforcement; Tech Textiles E-LT 850 grn-’ style 1036 O/90” reinforcement.

The resulting variations in permeability with ply angle are shown in Figures 9 and 10. The values of the constants used in equation were (2) 936 A==49.0x 10p9m2 for E-BXhd and A = 17.5 x 10p9m2 for E-LT 850, with m = 7.18 used for both materials. In both cases these values generally result in an excellent agreement between predicted and measured permeabilities over the range of ply angles. It is also apparent that although the permeability of E-BXhd 936 is consistently higher than that of E-LT 850, similar trends are observed for both fabrics. The permeability decreases for ply angles above 45” due to the combined effects of fibre reorientation and increased volume fraction. For ply angles below 45”, a slight increase occurs for low levels of shear as fibres become aligned towards the direction of flow. However, as the ply angle is decreased (below 40”), the associated increase in fibre volume fraction becomes dominant and the permeability begins to fall.

251

Preform processing

MECHANICAL

and performance

characterization:

Table 1 Fibre and matrix properties (from manufacturers’ data)

PROPERTIES

The effect of reinforcement shear on the subsequent laminate mechanical properties can be estimated in a number of ways. At the simplest level, the analysis suggested by Krenchel can be used to calculate a reinforcement efficiency factor which can be substituted into a modified form of the rule of mixtures to give an estimate of laminate stiffness”. The problem is complicated slightly for a sheared fabric, as there is an associated increase in fibre volume fraction [given by equation (l)]. An alternative and more robust approach is to use classical laminate theory14 to predict the properties of the laminate which effectively consists of unidirectional (UD) plies. The properties of each UD ply may be estimated using the Halpin-Tsai equations. Longitudinal modulus and Poisson’s ratios are estimated from the fibre and matrix properties using the rule of mixtures: El = EfV,+&(I

- Vf)

(4)

v12 = z+V,+z&(l

- V,)

(5)

al (1 -

Glass fibre

73.0

29.9

Wf

- MIn) + J-&I)

0.22

2605

4.47

1.62

0.38

1120

Dow Derekane 8084 vinyl ester resin

3.30

1.20

0.38

1130

25 7

t

Experimental

-

Krenchel ..-- ..Lamlnate

Theory

30

35

/&e(e)

40

50

55

60

65

Figure 11

Variation in tensile modulus with ply angle: E-BXhd 936 reinforcement with Cray Valley 6345.001 polyester resin

(7)

25 -

??

Experimental

-

Krenchel

...

..Laminate

55

60

Theory1

I

in which M is the appropriate modulus and c is an empirically determined constant. By applying a stress transformation based on the required ply angle, and then applying a unit strain to the off-axis laminate stiffness matrix, it is possible to predict the corresponding laminate tensile moduli. To assess the validity of each of the approaches described above, flat plaque mouldings were manufactured by resin transfer moulding using aluminium tooling with a cavity thickness of 3.5mm. Preforms were produced using three layers of reinforcement which were pre-sheared using the method described above. Two material combinations were used: Tech Textiles E-BXhd 936 reinforcement with Valley Totale 6345.001 unsaturated polyester (with 2% by mass Akzo Perkadox 16 catalyst); Tech Textiles E-LT 850 reinforcement with Derekane 8084 vinyl ester resin (with 1% by Interox TBPEH catalyst).

Density (kgm-l)

Ply

rlVf)

q = Wf

Poisson’s ratio

Cray Valley 6345.001 polyester resin

25

where

51

0'

25

30

35

40

50

65

Ply A&s (0) Figure 12 Variation in tensile modulus with ply angle: E-LT 850 reinforcement with Dow Derekane 8084 vinyl ester resin

Cray resin Dow mass

After post-cure at 130°C for 24 h, the elastic properties were measured using an Instron 1195 universal testing machine to a method encompassing BS 2782. In the following analysis, the constant c in the Halpin-Tsai relationships was given the value 2.0 for transverse modulus (E2) and 1.0 for the shear modulus (Gr2) as suggested by Halpin”. The fibre and matrix properties used are given in Table 1. Figures 11 and 12 compare the experimental moduli at ply angles between f25” and *65” with those predicted

252

Shear modulus (GPa)

(1+e?vf)

M=M

??

Material

Tensile modulus (GPa)

07

Transverse and shear moduli are calculated using

??

A. C. Long et al.

using both the Krenchel/rule of mixtures and classical laminate models. In both cases the modulus values predicted using laminate theory are reasonably accurate across the range of ply angles, whereas those predicted using Krenchel’s model are only accurate for ply angles below f45”. This discrepancy is due to the Poisson effect, which is ignored in Krenchel’s analysis but which becomes increasingly important when the fibres are transverse to the loading axis.

SUMMARY Reinforcement deformation has been characterized using an automatic strain analysis technique. The results

Preform processing

suggest that although reinforcement shear may be dominant for many practical forming operations, interfibre slip becomes increasingly important as depth of draw is increased. This is particularly true as the fabric approaches its so-called ‘locking angle’, a fabric-specific phenomenon which may be established using the strain analysis system. This approach may be used to characterize a range of reinforcements, facilitating improvement of the kinematic deformation model developed by the authors. The effects of reinforcement deformation on subsequent processing and performance characteristics have been established by carrying out measurements using sheared reinforcements. This has allowed semi-empirical relationships to be established relating permeability and laminate modulus to the local ply angle. By integrating these methods with an accurate reinforcement deformation model, it should be possible to develop a computeraided engineering system which enables formability, flow properties and structural requirements to be considered at the preform design stage.

and performance

6

7

8

9 IO

11

12

13

14

ACKNOWLEDGEMENTS This work was funded by Ford Motor Company (USA). The authors would particularly like to thank Ken Kendall, Carl Johnson and Mahmoud Demeri for their support and assistance. Dan Morris, Roger Smith and Andrew Kingham are also thanked for their assistance with experimental work.

15

NOMENCLATURE Empirical permeability constants Tensile modulus Permeability Reinforcement superficial density Cavity thickness Fibre volume fraction Empirical Halpin-Tsai constant Poisson’s ratio Fibre density Ply angle

A, m E SO

t Owen, M.J., Middleton, V. and Rudd, C.D., Fibre reinforcement for high volume resin transfer moulding (RTM). Composites Manufacturing 1990, 1. 14 Advani, S.G. and Bruschke, M.V., A finite element/control volume approach to mold filling in anisotropic porous media. P<+n. Conzpos. 1990. 11, 398 Rudd, C.D., Rice, E.V., Bulmer, L.J. and Long, A.C., Process modelling and design for resin transfer moulding. Plastics, Ruhher Compos. Process. Applic. 1993, 20, 67 Chan, A.W. and Morgan, R.J., Computer modeling of liquid composite molding for 3-dimensional complex shaped structures. In ‘Proc. 10th Annual ASM/ESD Advanced Composites Conference’, Dearborn, MI, 1994, pp. 341-345 Let, L.J., Young, W.B. and Lin, R.J., Mold filling and cure modelling of RTM and SRIM processes. Compos. Strucr. 1994. 27, 109

A. C. Long et al.

Robertson, R.E., Hsiue, E.S., Sickafus, E.N. and Yeh, G.S.Y., Fiber rearrangements during the moulding of continuous fibre composites. 1. Flat cloth to a hemisphere. Polyn~. Compos. 1981,2, 126 Van West, B.P., Pipes, R.B. and Keefe. M., A simulation of the draping of bidirectional fabrics over arbitrary surfaces. J. Test. Inst. 1990, 81, 448 Bergsma, O.K., Computer simulation of 3D forming processes of fabric reinforced plastics. In ‘Proc. 9th Int. Conf. on Comnosite Materials (I&M-9)‘, Woodhead Publishing Co.. Cambridge and University of Zaragoza, Spain, 1993, pp. 560-567 Laroche, D. and Vu-Khanh, T., Forming of woven fabric composites. J. Compos. Mater. 1994, 28, 1825 Long. A.C. and Rudd, C.D.. A simulation of reinforcement deformation during the production of preforms for liquid moulding processes. IMechE J. Eng. Mamtf. 1994. 208, 269 Long. A.C. and Rudd, CD.. Computer Integrated design of structural preforms for liquid moulding processes. In ‘Proc. 27th ISATA--New and Alternative Materials for the Transportation Industries’, Aachen, 1994. pp. 38 I 390 Advani, S.G., Bruschke, M. V. and Parnas. R.S., Resin transfer moulding flow phenomena in polymeric composites. In ‘Flow and Rheology in Polymer Composites Manufacturing’ (ed. S.G. Advani). Elsevier Science BV. Amsterdam. 1994. pp. 465 515 Rudd, C.D., Morris. D.J., Chick, J.P. and Warrior, N.A.. Materials characterization for SRIM. In ‘Proc. 4th Int. Conf. on Automated Composites (ICAC-95)‘. Nottingham, 1995 Tsai. SW. and Hahn, H.T., ‘Introduction to Composite Materials’, Technomic Publishing Co. Inc., Lancaster, PA. 1980 Halpin, J.C., ‘Primer on Composite Materials: ,4nalysis’. Technomic Publishing Co., Inc., Lancaster, PA, 19X4

k

REFERENCES

characterization:

Vf

E 1/ P

I9 Subscriprs 1, 2

f m .Y

Principal axes of UD plies Property of fibre Property of matrix Major principal axis of laminate/reinforcement

253