- Email: [email protected]
Influence of the filament winding tension on physical and mechanical properties of reinforced composites
Composites: Part A 33 (2002) 1615–1622 www.elsevier.com/locate/compositesa
Influence of the filament winding tension on physical and mechanical properties of reinforced composites P. Mertiny, F. Ellyin* Advanced Composite Materials Engineering Group, Department of Mechanical Engineering, 4-9 Mechanical Engineering Building, University of Alberta, Edmonton, AB T6G 2G8, Canada Received 25 January 2002; revised 18 October 2002; accepted 23 October 2002
Abstract In this experimental investigation the influence of the applied tow tension during filament winding on the physical and mechanical properties of glass-fibre reinforced polymeric composite tubulars, was studied. Pressure retaining tubular products used in the transportation/storage of fluids are generally subjected to a variety of loading conditions during their service life; thus tubular specimens were tested under different biaxial loading ratios. The stress/strain response was recorded and functional and structural failure envelopes were developed. These envelopes indicate the leakage and final failure characteristics of the components, respectively. The mechanical properties were analysed in conjunction with the measured physical properties: ‘fibre volume fraction’ and ‘effective wall thickness’. Experimental findings demonstrate that the component strength depends on the degree of fibre tensioning. Under fibre-dominated loading conditions, higher winding tension leads to an improved resistance against failure of tubular components, whereas under matrix-dominated loading failure is delayed by reduced fibre tensioning. q 2002 Published by Elsevier Science Ltd. Keywords: E. Filament winding; B. Physical properties; B. Mechanical properties; Fibre-reinforced composites
1. Introduction In the composites industry, the process of filament winding has evolved to be the preferred, and most cost effective method, for producing pressure retaining structures from fibre reinforced polymeric (FRP) composites (e.g. piping and tanks for the transportation/storage of fluids). Although this method has been in use for an extended period of time, the effect of processing parameters has only been investigated to a limited extend. In Ref. [1] the influence of primary processing parameters (i.e. parameters that can be selected, monitored and controlled by the operator; e.g. the tow tension) on secondary processing parameters (i.e. parameters that are not directly controllable by the operator, e.g. the winding bandwidth) has been reported. However, it is often of greater importance to understand how primary processing parameters affect the quality of a component. Techniques for predicting physical part properties were reported in, for * Corresponding author. Tel.: þ 1-780-492-2009; fax: þ1-780-492-2200. E-mail address: [email protected] (F. Ellyin). 1359-835X/02/$ - see front matter q 2002 Published by Elsevier Science Ltd. PII: S 1 3 5 9 - 8 3 5 X ( 0 2 ) 0 0 2 0 9 - 9
example, Ref. [2 – 4]. Cohen [5] used the design of experiment method (‘DOE’) to identify the applied tow tension during winding (winding tension) to be the most significant manufacturing parameter for the resulting mechanical part properties. Increasing this parameter produced a higher fibre volume fraction and, at the same time, an increased strength of the investigated structures (i.e. filament wound pressure vessels). In a later publication Cohen et al. [6] demonstrated a relationship between the fibre volume fraction in hoop-dominated laminae and the failure strength of filament wound pressure vessels. Note that in this specific case the fibre volume fraction was varied by a resin removal technique (i.e. running the fibres through an orifice after impregnation) instead of controlling the winding tension. The current study is seen as an extension of other research works concerned with the performance of pressurised tubulars (see e.g. [8,9]). However, the preceding investigations do not include a discussion on the influence of processing parameters, or they are limited to specialised cases such as purely hoop-dominated fibre/loading configurations [5,6]. In this paper, results from an experimental
1616
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
study are presented. The focus is to investigate the influence of the primary processing parameter ‘winding tension’ on the performance of pressurised tubular components. First, the effect of tow tensioning on the physical properties ‘fibre volume fraction’ and ‘effective wall thickness’ are described. Using a unique tow tensioning system [7], specimens were produced applying different winding tensions. In this study an angle-ply lay-up of [^ 6083]T was chosen, which is typical of pressurised composite structures such as piping that usually incorporate lay-up configurations with winding angles between 55 and 708. The investigation on physical properties was then extended to the mechanical properties of the components. Structures in industrial applications are generally subjected to complex loading states; thus, monotonic tests were conducted applying several biaxial loading ratios. The experimental set-up was designed to facilitate testing for functional and structural failure, i.e. the loss of the fluid containment capability and the structural integrity. Functional and structural failure envelopes were developed and analysed qualitatively, providing an insight into the material behaviour for a wider range of structures and loading configurations.
2.2. Determination of the fibre volume fraction
2. Testing program and experimental procedures
2.3. Experimental procedures for the determination of mechanical properties
The fibre volume fraction was determined by a resin burnout test. Ring samples of 25.4 mm length were taken from several locations along the length of a tubular. It is important to point out that the outside surface of these sections was machined to three-fourth of their individual average wall thickness, exposing the fibres of the top cover of the fibre bed. This preparation step was necessary for the following reason: Excess resin accumulates on the outside surface during the winding process building a resin cover that strongly affects a weight-based determination of the fibre volume fraction. It was found that measurements without removing the resin cover do not result in an useful parameter, as only the ‘local’ fibre volume fraction (nf ) within the fibre bed significantly influences mechanical part properties. After machining, several sample sections from a single tubular were subjected to an elevated temperature of 540 8C for three hours leaving the pure glass fibre bed behind. Knowing the weight of each section before and after the burnout procedure, and the densities of the constituents, an average fibre volume fraction was determined for each manufactured tubular batch.
2.1. Specimen material system and fabrication method Tubular specimens were produced by the wet filament winding method. A four axis, computer controlled winding system was used for the fabrication. The winding machine was equipped with a unique numerically controlled tensioning system that allows for accurate tension control of each individual tow [7]. Eight tows of an Owens-Corning E-glass fibre material of 0.735 g/m weight were tensioned, and guided through a drum-type resin bath. As matrix material, Resolution Performance’s EPON826/EPICURE9551 two-component resin system was used, which was kept at a constant temperature of 30 8C in the resin tank. Wet tows were finally combined to a fibre band while passing though the pay-out eye of the winding machine and placed in an helical pattern onto chrome-plated steel mandrels with a diameter of 38.1 mm. Six layers were deposited creating a [^ 6083]T interwoven angle-ply lay-up. Two batches of specimens were produced by applying uniform tow tension of 26.7 and 44.5 N, respectively. After completion of the winding sequence, excessive built-up of liquid resin on the surface of the parts was removed using a foam brush. Parts were then cured in a twostage curing cycle at 80 and 120 8C for 1 and 2.5 h, respectively, (excluding ramping times of 15 min) and afterwards allowed to cool down to room temperature. Finally, mandrels were extracted from the tubulars, and specimens of a specified length were machined from the parts.
Mechanical testing was done under stress control using a facility for multiaxial loading of tubular specimens [10]. Different biaxial loading ratios (given by the ratio of Hoop to Axial stress [H:A]) were applied by internal pressurisation and axial loading of the specimens. These ratios were: [1H:0A], i.e. a pure hoop loading, [3H:1A], i.e. a loading condition in which—based on netting analysis—the resulting stress coincides with the applied fibre direction, [2H:1A], i.e. an unconstrained loading condition (‘pressure vessel loading’), and [1H:15A], i.e. an effectively pure axial loading condition (note that a low internal pressurisation was necessary to facilitate the investigation of leakage events). Tubular specimens were provided with aluminium end tabs resulting in a 102 mm gauge section. A strain gauge rosette with a perpendicular grid was aligned in axial and hoop direction of the tubular and bonded at the mid-length of the specimen. Hoop and axial stresses (sH and sA ) were computed according to Eqs. (1) and (2) from the recorded internal pressure pi and the applied axial load FA. Together with the hoop and axial strain gauge readings (1H and 1A ) the global stress –strain response was obtained
sH ¼
IDðpi 2 po Þ 2 2spo 2s
ð1Þ
sA ¼
ID2 ðpi 2 po Þ 2 4po sðID þ sÞ FA þ 4sðID þ sÞ psðID þ sÞ
ð2Þ
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
where s is the effective wall thickness (as defined in Section 3.2), ID the inside diameter and po the atmospheric pressure. The monotonic stress-strain response under different loading ratios is only one aspect of describing the mechanical properties of a pressurised tubular structure. Of greater interest are the circumstances under which a part exhibits failure. Two characteristics can be distinguished: (a) a functional failure event in which the part is no longer able to contain the fluid and (b) a structural failure event in which the part loses its load-bearing capability. Both failure events can either occur simultaneously or as two distinct events. To facilitate an experimental investigation of the two at the same time, tubular specimens were equipped with a bladder system [11]. Hydraulic oil was used to pressurise a specimen from within a rubber bladder. The fluid filled the inside of the bladder as well as an annulus between the bladder and the wall of the tubular. During testing the loss of hydraulic fluid from the pressure intensifier was measured and data was normalised according to the compressibility of the hydraulic fluid and the expansion characteristics of the specimen and the testing system (e.g. hoses). The event of a functional failure is indicated by an extra loss of fluid (defined to be a loss of 1% of the specimen inside volume, i.e. 2 ml), as leakage occurs through crack openings in the specimen wall (i.e. matrix cracks). After the fluid in the annulus has penetrated through the wall, the rubber bladder acts as a seal enabling continued testing to the point of structural failure. A sudden drop of the internal pressure and/or the applied axial load indicates a structural failure. Strain gauge readings are not suited for detecting a structural failure, as effects associated with a functional failure (i.e. matrix cracking and fluid leakage) impair the gauge bond making the readings less reliable after occurrence of a functional failure.
3. Influence on physical properties 3.1. Fibre volume fraction Fibre volume fraction values (nf ) showed a strong correlation with the applied winding tension (see Fig. 1). In the case of the 26.7 N winding tension, an average fibre volume fraction of 70.8% was determined with an average absolute deviation from the mean of 0.05%. For the 44.5 N winding tension a fibre volume fraction value of 74.0% was computed with a variability of 0.38%. This accounts for a significant increase in fibre volume fraction of more than an absolute value of 3%. Note that the fibre volume fraction for samples with an unworked resin cover was found to be inconsistent. Values of 65.4 and 67.4% were measured with a high variability of 0.90 and 2.00% for the low and high winding tensions, respectively. A strict influence of an increase in winding tension cannot be ascertained in every case. An increased
1617
Fig. 1. Dependency of fibre volume fraction on the applied winding tension.
tow tension displaces more resin from within the fibre bed, which in turn accumulates on the outside of the part; thus the overall fibre volume fraction remains theoretically unchanged. However, a high variability is caused by (a) the manual operation of removing excessive resin build-up and (b) resin drip-off. 3.2. Effective specimen wall thickness In conjunction with basic geometrical information the thickness of the load bearing fibre bed (i.e. the effective specimen wall thickness, s ), can be approximated using the volume of deposited fibre material and the fibre volume fraction of the part [1]. More complex models for this purpose were described by for e.g. Banerjee et al. [12]. In the current study, the method proposed in Ref. [1] was chosen for its simplicity. An analytical determination of the wall thickness was preferred over measurement techniques, as (a) distance measurements using calliper type devices are difficult to obtain (the resin cover falsifies any measurements) and (b) the use of optical means (e.g. micrographs) often necessitates the destruction of the specimen. Optical methods are also generally tedious, especially, if a number of samples needs to be investigated. In Eq. (4) the thickness of the fibre bed is given by the sum of single cover thicknesses (t ) (i.e. thickness of two interwoven laminae, given by Eq. (3)), which are calculated by the number of applied tows (TOW), the tow weight (TEX), the fibre density (rf ) and the fibre volume fraction (nf ). Necessary geometrical information is the inside diameter of the part (ID) and the winding angle (a). Finally, the number of winding circuits (C ) needs to be taken into account for a volume-based calculation (The width of a fibre band is generally not sufficient to complete one laminae in a single pass of the machine carriage, hence it requires several passes, i.e. circuits resulting in an interwoven two-layer structure.). Using Eqs.(3) and (4) the thickness of the fibre bed was determined to be 1.27 and 1.21 mm for the lower and higher fibre volume fraction, respectively. Note that
1618
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
the overall average specimen thickness including the resin cover was found to be 1.38 mm tðCOVER#nÞ ¼
loading condition was applied (i.e. zero axial load). Consequently, the actual loading ratios deviate from the
TEX TOW C nf rf pððID=2Þ þ tðCOVER#1Þ þ · · · þ tðCOVER#n 2 1ÞÞcosðaÞ
s ¼ tðCOVER#1Þ þ tðCOVER#2Þ þ · ·· þ tðCOVER#nÞ
ð4Þ
To verify the proposed method, the effective wall thickness was measured for a specimen at several points around the circumference (angle increments of 108) using a toolmaker microscope. Based on these measurements, it was found that the calculation produced a slightly smaller thickness value with a relative error of less than 10%. However, due to the pronounced unevenness of the outside of the fibre bed, it was difficult to assess a distinct dividing line between the fibre bed and the resin cover. Hence, it can be concluded that results are strongly dependent on the quantity and location of measurements, as well as on the applied method.
ð3Þ
theoretical one. However, differences are small, i.e. the actual loading ratios for specimens with a high and low fibre volume fraction are 2.064 and 2.067, respectively, (based on the effective wall thickness). From Eqs. (1) and (2) one gets
› FA › pi ¼ p ›t ›t
A H
IDðID þ sÞ ID2 2 2 4
! ð5Þ
4.2. Biaxial monotonic stress – strain characteristics
Internal pressurisation was applied using a loading rate (›pi/›t ) of 0.462 kPa/s in the case of pure axial loading, and 4.626 kPa/s for the other three loading conditions. These rates were chosen in order to provide sufficient time for matrix cracks to develop/connect, and fluid to penetrate along cracks through the pipe wall. The selected loading ratios were attained by adjusting the following axial loading rates (›FA/›t ): 2 5.28 N/s for [1H:0A], 2 1.63 N/s for [3H:1A], 0.0 N/s for [2H:1A], and 15.84 N/s for [1H:15A]. These values are based on the overall average specimen wall thickness and were obtained from Eq. (5). The exception is the [2H:1A] loading case, where a true unconstrained
(a) [1H:0A]: Fig. 2 shows the measured global stress – strain response along with the computed strains transverse (1T) and parallel (1P) to the fibres. In this test the functional failure of the tubular occurred almost simultaneous with the structural failure, i.e. gradual leakage was not observed. The presence of a negative transverse strain component is deemed to be the reason for this type of behaviour (see Section 5.1). (b) [3H:1A]: Results from this test reveal a stress range with pronounced leakage between functional failure and the structural failure event (see Fig. 3). A positive transverse strain greater than 0.25% was measured within this range, where also a strong non-linearity in the otherwise almost linear stress –strain response was recorded for the axial and transverse direction (i.e. the matrix-dominated directions). (c) [2H:1A]: Similar to the preceding test, the hoop and parallel to the fibres stress –strain response of this loading (i.e. fibre-dominated components) is nearly linear before
Fig. 2. Biaxial stress–strain response for a pure hoop [1H:0A] loading ratio, 1A, 1H: axial and hoop strain (measured); 1T, 1P (calculated) transverse and parallel to fibre strain.
Fig. 3. Biaxial stress–strain response for a [3H:1A] loading ratio, 1A, 1H: axial and hoop strain (measured); 1T, 1P (calculated) transverse and parallel to fibre strain.
4. Influence on mechanical properties 4.1. On the effect of the axial loading rate
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
1619
axial stress value. At this point the specimen also lost its load-bearing capability, consequently, the functional and structural failure coincide in this loading case. 4.3. Structural failure characteristics
Fig. 4. Biaxial stress–strain response for a [2H:1A] loading ratio, 1A, 1H: axial and hoop strain (measured); 1T, 1P (calculated) transverse and parallel to fibre strain.
functional failure (see Fig. 4). However, a more pronounced non-linear behaviour is noticeable in the matrix-dominated directions. After functional failure at a transverse strain of 0.35% an extremely non-linear behaviour was recorded. The matrix-dominated strain components trend towards high positive values, whereas the hoop strain shifts towards high negative values. This behaviour indicates a strong change in specimen geometry. After functional failure the matrix increasingly breaks down allowing the fibres to realign in the resultant loading direction. This behaviour was subsequently accompanied by a significant change in length and diameter of the specimen. However, these changes were not measurable with the existing test set-up, hence, stresses were calculated assuming unchanged specimen dimensions. (d) [1H:15A]: Fig. 5 shows the stress – strain response of a typical specimen in the almost pure axial loading case. Strains in fibre direction were negligibly small; thus the load was carried to a great extend by the matrix material. At 0.32% transverse strain functional failure occurred at a low
Fig. 5. Biaxial stress– strain response for an effectively pure axial [1H:15A] loading ratio, 1A, 1H: axial and hoop strain (measured); 1T, 1P (calculated) transverse and parallel to fibre strain.
A visual inspection of the fractured specimens provides further insights into the mechanisms accompanying structural failure. In Fig. 6 the damaged specimens are presented for the [3H:1A], [2H:1A] and [1H:15A] loading cases. Note that a specimen tested under a pure hoop loading is not shown, since these specimens failed by bursting leaving only fragments behind. And thus, no localised damage could be assessed. This type of behaviour indicates that within the structure a high amount of energy was stored in an evenly distributed manner. Under a [3H:1A] loading, burst failures with fractured surfaces parallel to the fibre direction could be observed. Failure generally occurred within the gauge section of the specimen. This type of damage is characteristic of a fibre dominated failure mode where fibres are loaded parallel to the resulting loading direction. The strength of the fibre structure is well utilised in this loading/ lay-up combination. In the [2H:1A] loading case, localised failure occurred near the end tabs, indicating a stress concentration in this zone. As the fibres were fixed in the annulus of the end-tabs, the aforementioned fibre realignment after matrix breakdown led to an additional localised fibre bending; thus, the potential of the fibre structure could not be fully utilised. In the pure axial loading case the fibre bed immediately pulled apart after the breakdown of the matrix, indicating a poor utilisation of the potential strength of the structure. 4.4. Biaxial monotonic failure envelopes Test results for specimens with an average fibre volume fraction of 70.8 and 74.0% were compiled into four biaxial
Fig. 6. Photograph of specimens failed under (a) [3H:1A], (b) [2H:1A] and (c) [1H:15A] loading conditions, and (d) undamaged specimen (from left to right).
1620
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
Fig. 7. Biaxial functional and structural stress failure envelopes.
failure stress envelopes (shown in Fig. 7). As expected the structural failure envelopes encompass the functional type. In the pure hoop loading case, structural and functional failure points practically coincide at the highest recorded hoop stresses. Generally, failure stresses show a good consistency for the different loading ratios. The exception is the [3H:1A] loading case, where values exhibit considerable scatter. This is attributed to variations in the strength of the fibre structure, which is predominantly load bearing in this case. However, mean values still follow a general trend that is similar for the structural and functional failure type. This trend can be described as follows: Under [3H:1A] loading a higher winding tension (i.e. higher fibre volume fraction) produced components with greater strength (strength meaning resistance against failure). However, in the case of the other loading ratios, a higher winding tension produced parts with equal or even less strength. In particular under increasingly matrix dominated loading (i.e. [1H:15A], [2H:1A]) test results indicate that component strength was decreased by an increased tow tension. Note that this study does not discuss the presence of residual stresses that may arise from the manufacturing procedures, e.g. curing specimens at elevated temperatures. Process models for the determination of fibre pretension are available (see for example [2 – 4]). However, a timedependent relaxation of residual stresses also needs to be considered, as polymer matrix materials are generally of a visco-elastic nature (see e.g. [13,14]). Accounting for these effects, i.e. relaxation of initial residual stresses resulting from the applied winding tension due to curing at elevated temperatures and visco-elastic effects, are beyond the scope this experimental study.
could be observed, without any other visible damage to the pipe. This observation suggests that a network of cracks develops, starting from inter-laminar matrix damage, i.e. long cracks forming parallel to the fibres. These cracks bridge intra-laminar zones and coalesce allowing fluid to penetrate through the wall of the tubular. It is reasonable to assume that inter-laminar matrix cracking is caused by strains acting transverse to the fibre direction. Note that this leakage mechanism is only operative, if an opening mode, i.e. a positive displacement, is present. Hence, the typical leakage event was absent under [1H:0A] loading where transverse strains were found to be compressive. From the recorded test data, strains parallel and transverse to the fibre direction at the leakage load were computed for the remaining three loading ratios. As shown in Fig. 8, functional failure occurs approximately at an average value of 0.3% transverse strain, independent of the applied loading ratio. The exhibited scatter of data points is probably due to: (a) the strain gauge readings being affected by the developed damage and (b) the strain gauges being positioned close to undulations of the interwoven fibre structure, where the local stress state varies from those locations where a uniform fibre structure exists. However, the observation of a common failure indicator (i.e. transverse strain) confirms that the functional failure envelopes in fact properly represent the material behaviour. 5.2. Qualitative analysis of structural failure events A uniaxial fibre/matrix structure is preferably loaded in fibre direction. Failure occurs when the tensile load exceeds a certain critical value (the maximum tensile strength, sPmax). In the following, the observed structural failure behaviour will be qualitatively analysed applying the same principle. In the [3H:1A] loading case the resulting stress caused by the biaxial loading coincides with the applied winding angle of 608. The stress/strain response and the investigation of fractured specimens indicated that for this case fibres were actually experiencing a predominantly tensile loading (see Sections 4.2 and 4.3). Hence, the stress
5. Discussion 5.1. Qualitative analysis of functional failure events Matrix damage is regarded to be the primary cause for functional failure, as fluid weeping through the pipe wall
Fig. 8. Strains parallel and transverse to fibres at functional failure loads.
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
1621
Fig. 10. Mohr’s circle illustration for comparison of the [3H:1A] and [2H:1A] structural failure behaviour. Fig. 9. Mohr’s circle illustration for comparison of the [3H:1A] and [1H:0A] structural failure behaviour.
parallel to the fibres at structural failure under [3H:1A] loading is assumed to be the maximum tensile strength of the fibre/matrix structure, sPmax. Mohr’s circles were chosen to illustrate the conditions under pure hoop loading, see Fig. 9. The [3H:1A] loading condition is represented by the top circle acting as a scale for comparison with the circle for the [1H:0A] loading case (shown below). The second circle was constructed using the maximum tensile strength of the fibre/matrix structure, sPmax, and the given loading ratio. It can be observed that the hoop stress, sH, is necessarily greater than that in the first circle. This corresponds with the behaviour shown in the structural failure envelope (Fig. 7). Moreover, the transverse stress, sT, for the pure hoop loading case is shown to be considerably less than that under [3H:1A] loading. This then supports the assumption that significant matrix cracking did not occur under the pure hoop loading. Comparing the failure behaviour under [3H:1A] and [2H:1A] loading using a similar approach as presented above supports the assumption of a localised load concentration in the fibre structure (see Section 4.2). In the bottom part of Fig. 10, a Mohr’s circle for the [2H:1A] loading was constructed using a fibre angle of 558, which is the limit case for fibre realignment after matrix breakdown (found by conducting netting analysis for the particular loading ratio). Matching the stresses parallel to the fibre direction (i.e. sP ¼ sPmax ) for the two considered loading ratios, resulted in nearly identical hoop stress values sH. This does not correspond with the experimental results, where the hoop stress is found to be less in the [2H:1A] loading case. However, incorporating a decreased maximum strength of the fibre structure (i.e. sP , sPmax) gave a representation that quantitatively matched the experimental findings. Such a decrease was already suggested in Section 4.3 based on the presence of additional bending loads near
the specimen end tabs. Depending on the amount of strength reduction, axial stresses at failure, sA, can be higher or lower than the corresponding axial stresses in the [3H:1A] loading case; thus explaining the intersection of the structural failure envelopes for the high and low fibre volume fraction.
6. Conclusions In this investigation the effect of the primary manufacturing parameter ‘winding tension’ on physical and mechanical properties of filament wound [^ 6083]T composite tubulars was studied. Specimens were manufactured using different winding tensions, and mechanical testing was conducted under different loading ratios. Results from the mechanical testing were compiled into functional and structural failure stress envelopes, and the observed mechanical behaviour was analysed qualitatively to show the plausibility of the experimental findings. Linking the applied winding tension to the physical properties of a part, as well as to the mechanical response at failure, led to the following conclusions. † Increasing the filament winding tow tension significantly increased fibre compaction; thus the primary processing parameter ‘winding tension’ is suited for controlling the fibre volume fraction of the manufactured components. † Mechanical testing revealed that stresses at functional and structural failure depend on the applied winding tension. † Comparison of test results from specimens wound with a low and high tow tension indicated that the mechanical properties do not follow a simple unidirectional trend. Generally, an increase in winding tension has a beneficial effect on the component strength in the case of a fibre dominated loading, whereas under a matrix dominated
1622
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
loading condition a slight decrease in the failure strength was obtained by increasing fibre compaction.
Acknowledgements This research was made possible by the contributions from TransCanada Pipelines Limited, the Alberta Intellectual Infrastructure Partnership Program (IIPP), the Canadian Foundation for Innovation (CFI), the Natural Sciences and Engineering Research Council (NSERC) and the University of Alberta. In addition, the authors would also like to acknowledge the contributions made by the members of the ACME group and the technical staff at the Department of Mechanical Engineering (J. Wolodko, G. Meijer and B. Faulkner).
References [1] Mertiny P, Ellyin F. Selection of optimal processing parameters in filament winding. In: Proceedings of 33rd International SAMPE Technical Conference. Seattle; November 2001. p. 1084–95. [2] Lee SY, Springer GS. Filament winding cylinders: I Process model. Compos Mater 1990;24:1270–98. [3] Calius EP, Lee SY, Springer GS. Filament winding cylinders: II Validation of process model. Compos Mater 1990;24:1299 –343. [4] Lee SY, Springer GS. Filament winding cylinders: I Selection of the process variables. Compos Mater 1990;24:1344–66.
[5] Cohen D. Influence of filament winding parameters on composite vessel quality and strength. Composites Part A 1997;28A: 1035–47. [6] Cohen D, Mantell SC, Zhao L. The effect of fibre volume fraction on filament wound composite pressure vessel strength. Composites Part B 2001;32B:413–29. [7] Wolodko J, Mertiny P, Meijer G, Martens M, Ellyin F. Development of a facility for filament winding GFRP tubulars. In: Proceedings of 46th International SAMPE Symposium. Long Beach; May 2001. p.1271–82. [8] Soden PD, Leadbetter D, Griggs PR, Eckold GC. The strength of a filament wound composite under biaxial loading. Composites 1978; 9(4):247–50. [9] Soden PD, Kitching R, Tse PC. Experimental failure stresses for ^558 filament wound glass fibre reinforced plastic tubes under biaxial loading. Composites 1989;20(2):125–35. [10] Ellyin F, Wolodko J. Testing facilities for multiaxial loading of tubular specimens. In: Kalluri S, Bonacuse PJ, editors. Multiaxial fatigue and deformation testing techniques. ; 1997. p. 7–24. ASTM STP 1280. [11] Wolodko J. Biaxial fatigue and leakage characteristics of fiber reinforced composite tubes. Ph.D. Dissertation. University of Alberta; 1999. [12] Banerjee A, Sun L, Mantell SC, Cohen D. Model and experimental study of fibre motion in wet filament winding. Composites Part A 1998;29A:251 –63. [13] Hu Y, Xia Z, Ellyin F. Mechanical behaviour of an epoxy resin under multiaxial loadings. Part I: experimental study. Polym Polym Compos 2000;8:11–8. [14] Chen Y, Xia Z, Ellyin F. Evolution of residual stresses induced during curing processing using a viscoelastic model. Compos Mater 2001;35: 522 –42.
Influence of the filament winding tension on physical and mechanical properties of reinforced composites P. Mertiny, F. Ellyin* Advanced Composite Materials Engineering Group, Department of Mechanical Engineering, 4-9 Mechanical Engineering Building, University of Alberta, Edmonton, AB T6G 2G8, Canada Received 25 January 2002; revised 18 October 2002; accepted 23 October 2002
Abstract In this experimental investigation the influence of the applied tow tension during filament winding on the physical and mechanical properties of glass-fibre reinforced polymeric composite tubulars, was studied. Pressure retaining tubular products used in the transportation/storage of fluids are generally subjected to a variety of loading conditions during their service life; thus tubular specimens were tested under different biaxial loading ratios. The stress/strain response was recorded and functional and structural failure envelopes were developed. These envelopes indicate the leakage and final failure characteristics of the components, respectively. The mechanical properties were analysed in conjunction with the measured physical properties: ‘fibre volume fraction’ and ‘effective wall thickness’. Experimental findings demonstrate that the component strength depends on the degree of fibre tensioning. Under fibre-dominated loading conditions, higher winding tension leads to an improved resistance against failure of tubular components, whereas under matrix-dominated loading failure is delayed by reduced fibre tensioning. q 2002 Published by Elsevier Science Ltd. Keywords: E. Filament winding; B. Physical properties; B. Mechanical properties; Fibre-reinforced composites
1. Introduction In the composites industry, the process of filament winding has evolved to be the preferred, and most cost effective method, for producing pressure retaining structures from fibre reinforced polymeric (FRP) composites (e.g. piping and tanks for the transportation/storage of fluids). Although this method has been in use for an extended period of time, the effect of processing parameters has only been investigated to a limited extend. In Ref. [1] the influence of primary processing parameters (i.e. parameters that can be selected, monitored and controlled by the operator; e.g. the tow tension) on secondary processing parameters (i.e. parameters that are not directly controllable by the operator, e.g. the winding bandwidth) has been reported. However, it is often of greater importance to understand how primary processing parameters affect the quality of a component. Techniques for predicting physical part properties were reported in, for * Corresponding author. Tel.: þ 1-780-492-2009; fax: þ1-780-492-2200. E-mail address: [email protected] (F. Ellyin). 1359-835X/02/$ - see front matter q 2002 Published by Elsevier Science Ltd. PII: S 1 3 5 9 - 8 3 5 X ( 0 2 ) 0 0 2 0 9 - 9
example, Ref. [2 – 4]. Cohen [5] used the design of experiment method (‘DOE’) to identify the applied tow tension during winding (winding tension) to be the most significant manufacturing parameter for the resulting mechanical part properties. Increasing this parameter produced a higher fibre volume fraction and, at the same time, an increased strength of the investigated structures (i.e. filament wound pressure vessels). In a later publication Cohen et al. [6] demonstrated a relationship between the fibre volume fraction in hoop-dominated laminae and the failure strength of filament wound pressure vessels. Note that in this specific case the fibre volume fraction was varied by a resin removal technique (i.e. running the fibres through an orifice after impregnation) instead of controlling the winding tension. The current study is seen as an extension of other research works concerned with the performance of pressurised tubulars (see e.g. [8,9]). However, the preceding investigations do not include a discussion on the influence of processing parameters, or they are limited to specialised cases such as purely hoop-dominated fibre/loading configurations [5,6]. In this paper, results from an experimental
1616
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
study are presented. The focus is to investigate the influence of the primary processing parameter ‘winding tension’ on the performance of pressurised tubular components. First, the effect of tow tensioning on the physical properties ‘fibre volume fraction’ and ‘effective wall thickness’ are described. Using a unique tow tensioning system [7], specimens were produced applying different winding tensions. In this study an angle-ply lay-up of [^ 6083]T was chosen, which is typical of pressurised composite structures such as piping that usually incorporate lay-up configurations with winding angles between 55 and 708. The investigation on physical properties was then extended to the mechanical properties of the components. Structures in industrial applications are generally subjected to complex loading states; thus, monotonic tests were conducted applying several biaxial loading ratios. The experimental set-up was designed to facilitate testing for functional and structural failure, i.e. the loss of the fluid containment capability and the structural integrity. Functional and structural failure envelopes were developed and analysed qualitatively, providing an insight into the material behaviour for a wider range of structures and loading configurations.
2.2. Determination of the fibre volume fraction
2. Testing program and experimental procedures
2.3. Experimental procedures for the determination of mechanical properties
The fibre volume fraction was determined by a resin burnout test. Ring samples of 25.4 mm length were taken from several locations along the length of a tubular. It is important to point out that the outside surface of these sections was machined to three-fourth of their individual average wall thickness, exposing the fibres of the top cover of the fibre bed. This preparation step was necessary for the following reason: Excess resin accumulates on the outside surface during the winding process building a resin cover that strongly affects a weight-based determination of the fibre volume fraction. It was found that measurements without removing the resin cover do not result in an useful parameter, as only the ‘local’ fibre volume fraction (nf ) within the fibre bed significantly influences mechanical part properties. After machining, several sample sections from a single tubular were subjected to an elevated temperature of 540 8C for three hours leaving the pure glass fibre bed behind. Knowing the weight of each section before and after the burnout procedure, and the densities of the constituents, an average fibre volume fraction was determined for each manufactured tubular batch.
2.1. Specimen material system and fabrication method Tubular specimens were produced by the wet filament winding method. A four axis, computer controlled winding system was used for the fabrication. The winding machine was equipped with a unique numerically controlled tensioning system that allows for accurate tension control of each individual tow [7]. Eight tows of an Owens-Corning E-glass fibre material of 0.735 g/m weight were tensioned, and guided through a drum-type resin bath. As matrix material, Resolution Performance’s EPON826/EPICURE9551 two-component resin system was used, which was kept at a constant temperature of 30 8C in the resin tank. Wet tows were finally combined to a fibre band while passing though the pay-out eye of the winding machine and placed in an helical pattern onto chrome-plated steel mandrels with a diameter of 38.1 mm. Six layers were deposited creating a [^ 6083]T interwoven angle-ply lay-up. Two batches of specimens were produced by applying uniform tow tension of 26.7 and 44.5 N, respectively. After completion of the winding sequence, excessive built-up of liquid resin on the surface of the parts was removed using a foam brush. Parts were then cured in a twostage curing cycle at 80 and 120 8C for 1 and 2.5 h, respectively, (excluding ramping times of 15 min) and afterwards allowed to cool down to room temperature. Finally, mandrels were extracted from the tubulars, and specimens of a specified length were machined from the parts.
Mechanical testing was done under stress control using a facility for multiaxial loading of tubular specimens [10]. Different biaxial loading ratios (given by the ratio of Hoop to Axial stress [H:A]) were applied by internal pressurisation and axial loading of the specimens. These ratios were: [1H:0A], i.e. a pure hoop loading, [3H:1A], i.e. a loading condition in which—based on netting analysis—the resulting stress coincides with the applied fibre direction, [2H:1A], i.e. an unconstrained loading condition (‘pressure vessel loading’), and [1H:15A], i.e. an effectively pure axial loading condition (note that a low internal pressurisation was necessary to facilitate the investigation of leakage events). Tubular specimens were provided with aluminium end tabs resulting in a 102 mm gauge section. A strain gauge rosette with a perpendicular grid was aligned in axial and hoop direction of the tubular and bonded at the mid-length of the specimen. Hoop and axial stresses (sH and sA ) were computed according to Eqs. (1) and (2) from the recorded internal pressure pi and the applied axial load FA. Together with the hoop and axial strain gauge readings (1H and 1A ) the global stress –strain response was obtained
sH ¼
IDðpi 2 po Þ 2 2spo 2s
ð1Þ
sA ¼
ID2 ðpi 2 po Þ 2 4po sðID þ sÞ FA þ 4sðID þ sÞ psðID þ sÞ
ð2Þ
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
where s is the effective wall thickness (as defined in Section 3.2), ID the inside diameter and po the atmospheric pressure. The monotonic stress-strain response under different loading ratios is only one aspect of describing the mechanical properties of a pressurised tubular structure. Of greater interest are the circumstances under which a part exhibits failure. Two characteristics can be distinguished: (a) a functional failure event in which the part is no longer able to contain the fluid and (b) a structural failure event in which the part loses its load-bearing capability. Both failure events can either occur simultaneously or as two distinct events. To facilitate an experimental investigation of the two at the same time, tubular specimens were equipped with a bladder system [11]. Hydraulic oil was used to pressurise a specimen from within a rubber bladder. The fluid filled the inside of the bladder as well as an annulus between the bladder and the wall of the tubular. During testing the loss of hydraulic fluid from the pressure intensifier was measured and data was normalised according to the compressibility of the hydraulic fluid and the expansion characteristics of the specimen and the testing system (e.g. hoses). The event of a functional failure is indicated by an extra loss of fluid (defined to be a loss of 1% of the specimen inside volume, i.e. 2 ml), as leakage occurs through crack openings in the specimen wall (i.e. matrix cracks). After the fluid in the annulus has penetrated through the wall, the rubber bladder acts as a seal enabling continued testing to the point of structural failure. A sudden drop of the internal pressure and/or the applied axial load indicates a structural failure. Strain gauge readings are not suited for detecting a structural failure, as effects associated with a functional failure (i.e. matrix cracking and fluid leakage) impair the gauge bond making the readings less reliable after occurrence of a functional failure.
3. Influence on physical properties 3.1. Fibre volume fraction Fibre volume fraction values (nf ) showed a strong correlation with the applied winding tension (see Fig. 1). In the case of the 26.7 N winding tension, an average fibre volume fraction of 70.8% was determined with an average absolute deviation from the mean of 0.05%. For the 44.5 N winding tension a fibre volume fraction value of 74.0% was computed with a variability of 0.38%. This accounts for a significant increase in fibre volume fraction of more than an absolute value of 3%. Note that the fibre volume fraction for samples with an unworked resin cover was found to be inconsistent. Values of 65.4 and 67.4% were measured with a high variability of 0.90 and 2.00% for the low and high winding tensions, respectively. A strict influence of an increase in winding tension cannot be ascertained in every case. An increased
1617
Fig. 1. Dependency of fibre volume fraction on the applied winding tension.
tow tension displaces more resin from within the fibre bed, which in turn accumulates on the outside of the part; thus the overall fibre volume fraction remains theoretically unchanged. However, a high variability is caused by (a) the manual operation of removing excessive resin build-up and (b) resin drip-off. 3.2. Effective specimen wall thickness In conjunction with basic geometrical information the thickness of the load bearing fibre bed (i.e. the effective specimen wall thickness, s ), can be approximated using the volume of deposited fibre material and the fibre volume fraction of the part [1]. More complex models for this purpose were described by for e.g. Banerjee et al. [12]. In the current study, the method proposed in Ref. [1] was chosen for its simplicity. An analytical determination of the wall thickness was preferred over measurement techniques, as (a) distance measurements using calliper type devices are difficult to obtain (the resin cover falsifies any measurements) and (b) the use of optical means (e.g. micrographs) often necessitates the destruction of the specimen. Optical methods are also generally tedious, especially, if a number of samples needs to be investigated. In Eq. (4) the thickness of the fibre bed is given by the sum of single cover thicknesses (t ) (i.e. thickness of two interwoven laminae, given by Eq. (3)), which are calculated by the number of applied tows (TOW), the tow weight (TEX), the fibre density (rf ) and the fibre volume fraction (nf ). Necessary geometrical information is the inside diameter of the part (ID) and the winding angle (a). Finally, the number of winding circuits (C ) needs to be taken into account for a volume-based calculation (The width of a fibre band is generally not sufficient to complete one laminae in a single pass of the machine carriage, hence it requires several passes, i.e. circuits resulting in an interwoven two-layer structure.). Using Eqs.(3) and (4) the thickness of the fibre bed was determined to be 1.27 and 1.21 mm for the lower and higher fibre volume fraction, respectively. Note that
1618
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
the overall average specimen thickness including the resin cover was found to be 1.38 mm tðCOVER#nÞ ¼
loading condition was applied (i.e. zero axial load). Consequently, the actual loading ratios deviate from the
TEX TOW C nf rf pððID=2Þ þ tðCOVER#1Þ þ · · · þ tðCOVER#n 2 1ÞÞcosðaÞ
s ¼ tðCOVER#1Þ þ tðCOVER#2Þ þ · ·· þ tðCOVER#nÞ
ð4Þ
To verify the proposed method, the effective wall thickness was measured for a specimen at several points around the circumference (angle increments of 108) using a toolmaker microscope. Based on these measurements, it was found that the calculation produced a slightly smaller thickness value with a relative error of less than 10%. However, due to the pronounced unevenness of the outside of the fibre bed, it was difficult to assess a distinct dividing line between the fibre bed and the resin cover. Hence, it can be concluded that results are strongly dependent on the quantity and location of measurements, as well as on the applied method.
ð3Þ
theoretical one. However, differences are small, i.e. the actual loading ratios for specimens with a high and low fibre volume fraction are 2.064 and 2.067, respectively, (based on the effective wall thickness). From Eqs. (1) and (2) one gets
› FA › pi ¼ p ›t ›t
A H
IDðID þ sÞ ID2 2 2 4
! ð5Þ
4.2. Biaxial monotonic stress – strain characteristics
Internal pressurisation was applied using a loading rate (›pi/›t ) of 0.462 kPa/s in the case of pure axial loading, and 4.626 kPa/s for the other three loading conditions. These rates were chosen in order to provide sufficient time for matrix cracks to develop/connect, and fluid to penetrate along cracks through the pipe wall. The selected loading ratios were attained by adjusting the following axial loading rates (›FA/›t ): 2 5.28 N/s for [1H:0A], 2 1.63 N/s for [3H:1A], 0.0 N/s for [2H:1A], and 15.84 N/s for [1H:15A]. These values are based on the overall average specimen wall thickness and were obtained from Eq. (5). The exception is the [2H:1A] loading case, where a true unconstrained
(a) [1H:0A]: Fig. 2 shows the measured global stress – strain response along with the computed strains transverse (1T) and parallel (1P) to the fibres. In this test the functional failure of the tubular occurred almost simultaneous with the structural failure, i.e. gradual leakage was not observed. The presence of a negative transverse strain component is deemed to be the reason for this type of behaviour (see Section 5.1). (b) [3H:1A]: Results from this test reveal a stress range with pronounced leakage between functional failure and the structural failure event (see Fig. 3). A positive transverse strain greater than 0.25% was measured within this range, where also a strong non-linearity in the otherwise almost linear stress –strain response was recorded for the axial and transverse direction (i.e. the matrix-dominated directions). (c) [2H:1A]: Similar to the preceding test, the hoop and parallel to the fibres stress –strain response of this loading (i.e. fibre-dominated components) is nearly linear before
Fig. 2. Biaxial stress–strain response for a pure hoop [1H:0A] loading ratio, 1A, 1H: axial and hoop strain (measured); 1T, 1P (calculated) transverse and parallel to fibre strain.
Fig. 3. Biaxial stress–strain response for a [3H:1A] loading ratio, 1A, 1H: axial and hoop strain (measured); 1T, 1P (calculated) transverse and parallel to fibre strain.
4. Influence on mechanical properties 4.1. On the effect of the axial loading rate
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
1619
axial stress value. At this point the specimen also lost its load-bearing capability, consequently, the functional and structural failure coincide in this loading case. 4.3. Structural failure characteristics
Fig. 4. Biaxial stress–strain response for a [2H:1A] loading ratio, 1A, 1H: axial and hoop strain (measured); 1T, 1P (calculated) transverse and parallel to fibre strain.
functional failure (see Fig. 4). However, a more pronounced non-linear behaviour is noticeable in the matrix-dominated directions. After functional failure at a transverse strain of 0.35% an extremely non-linear behaviour was recorded. The matrix-dominated strain components trend towards high positive values, whereas the hoop strain shifts towards high negative values. This behaviour indicates a strong change in specimen geometry. After functional failure the matrix increasingly breaks down allowing the fibres to realign in the resultant loading direction. This behaviour was subsequently accompanied by a significant change in length and diameter of the specimen. However, these changes were not measurable with the existing test set-up, hence, stresses were calculated assuming unchanged specimen dimensions. (d) [1H:15A]: Fig. 5 shows the stress – strain response of a typical specimen in the almost pure axial loading case. Strains in fibre direction were negligibly small; thus the load was carried to a great extend by the matrix material. At 0.32% transverse strain functional failure occurred at a low
Fig. 5. Biaxial stress– strain response for an effectively pure axial [1H:15A] loading ratio, 1A, 1H: axial and hoop strain (measured); 1T, 1P (calculated) transverse and parallel to fibre strain.
A visual inspection of the fractured specimens provides further insights into the mechanisms accompanying structural failure. In Fig. 6 the damaged specimens are presented for the [3H:1A], [2H:1A] and [1H:15A] loading cases. Note that a specimen tested under a pure hoop loading is not shown, since these specimens failed by bursting leaving only fragments behind. And thus, no localised damage could be assessed. This type of behaviour indicates that within the structure a high amount of energy was stored in an evenly distributed manner. Under a [3H:1A] loading, burst failures with fractured surfaces parallel to the fibre direction could be observed. Failure generally occurred within the gauge section of the specimen. This type of damage is characteristic of a fibre dominated failure mode where fibres are loaded parallel to the resulting loading direction. The strength of the fibre structure is well utilised in this loading/ lay-up combination. In the [2H:1A] loading case, localised failure occurred near the end tabs, indicating a stress concentration in this zone. As the fibres were fixed in the annulus of the end-tabs, the aforementioned fibre realignment after matrix breakdown led to an additional localised fibre bending; thus, the potential of the fibre structure could not be fully utilised. In the pure axial loading case the fibre bed immediately pulled apart after the breakdown of the matrix, indicating a poor utilisation of the potential strength of the structure. 4.4. Biaxial monotonic failure envelopes Test results for specimens with an average fibre volume fraction of 70.8 and 74.0% were compiled into four biaxial
Fig. 6. Photograph of specimens failed under (a) [3H:1A], (b) [2H:1A] and (c) [1H:15A] loading conditions, and (d) undamaged specimen (from left to right).
1620
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
Fig. 7. Biaxial functional and structural stress failure envelopes.
failure stress envelopes (shown in Fig. 7). As expected the structural failure envelopes encompass the functional type. In the pure hoop loading case, structural and functional failure points practically coincide at the highest recorded hoop stresses. Generally, failure stresses show a good consistency for the different loading ratios. The exception is the [3H:1A] loading case, where values exhibit considerable scatter. This is attributed to variations in the strength of the fibre structure, which is predominantly load bearing in this case. However, mean values still follow a general trend that is similar for the structural and functional failure type. This trend can be described as follows: Under [3H:1A] loading a higher winding tension (i.e. higher fibre volume fraction) produced components with greater strength (strength meaning resistance against failure). However, in the case of the other loading ratios, a higher winding tension produced parts with equal or even less strength. In particular under increasingly matrix dominated loading (i.e. [1H:15A], [2H:1A]) test results indicate that component strength was decreased by an increased tow tension. Note that this study does not discuss the presence of residual stresses that may arise from the manufacturing procedures, e.g. curing specimens at elevated temperatures. Process models for the determination of fibre pretension are available (see for example [2 – 4]). However, a timedependent relaxation of residual stresses also needs to be considered, as polymer matrix materials are generally of a visco-elastic nature (see e.g. [13,14]). Accounting for these effects, i.e. relaxation of initial residual stresses resulting from the applied winding tension due to curing at elevated temperatures and visco-elastic effects, are beyond the scope this experimental study.
could be observed, without any other visible damage to the pipe. This observation suggests that a network of cracks develops, starting from inter-laminar matrix damage, i.e. long cracks forming parallel to the fibres. These cracks bridge intra-laminar zones and coalesce allowing fluid to penetrate through the wall of the tubular. It is reasonable to assume that inter-laminar matrix cracking is caused by strains acting transverse to the fibre direction. Note that this leakage mechanism is only operative, if an opening mode, i.e. a positive displacement, is present. Hence, the typical leakage event was absent under [1H:0A] loading where transverse strains were found to be compressive. From the recorded test data, strains parallel and transverse to the fibre direction at the leakage load were computed for the remaining three loading ratios. As shown in Fig. 8, functional failure occurs approximately at an average value of 0.3% transverse strain, independent of the applied loading ratio. The exhibited scatter of data points is probably due to: (a) the strain gauge readings being affected by the developed damage and (b) the strain gauges being positioned close to undulations of the interwoven fibre structure, where the local stress state varies from those locations where a uniform fibre structure exists. However, the observation of a common failure indicator (i.e. transverse strain) confirms that the functional failure envelopes in fact properly represent the material behaviour. 5.2. Qualitative analysis of structural failure events A uniaxial fibre/matrix structure is preferably loaded in fibre direction. Failure occurs when the tensile load exceeds a certain critical value (the maximum tensile strength, sPmax). In the following, the observed structural failure behaviour will be qualitatively analysed applying the same principle. In the [3H:1A] loading case the resulting stress caused by the biaxial loading coincides with the applied winding angle of 608. The stress/strain response and the investigation of fractured specimens indicated that for this case fibres were actually experiencing a predominantly tensile loading (see Sections 4.2 and 4.3). Hence, the stress
5. Discussion 5.1. Qualitative analysis of functional failure events Matrix damage is regarded to be the primary cause for functional failure, as fluid weeping through the pipe wall
Fig. 8. Strains parallel and transverse to fibres at functional failure loads.
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
1621
Fig. 10. Mohr’s circle illustration for comparison of the [3H:1A] and [2H:1A] structural failure behaviour. Fig. 9. Mohr’s circle illustration for comparison of the [3H:1A] and [1H:0A] structural failure behaviour.
parallel to the fibres at structural failure under [3H:1A] loading is assumed to be the maximum tensile strength of the fibre/matrix structure, sPmax. Mohr’s circles were chosen to illustrate the conditions under pure hoop loading, see Fig. 9. The [3H:1A] loading condition is represented by the top circle acting as a scale for comparison with the circle for the [1H:0A] loading case (shown below). The second circle was constructed using the maximum tensile strength of the fibre/matrix structure, sPmax, and the given loading ratio. It can be observed that the hoop stress, sH, is necessarily greater than that in the first circle. This corresponds with the behaviour shown in the structural failure envelope (Fig. 7). Moreover, the transverse stress, sT, for the pure hoop loading case is shown to be considerably less than that under [3H:1A] loading. This then supports the assumption that significant matrix cracking did not occur under the pure hoop loading. Comparing the failure behaviour under [3H:1A] and [2H:1A] loading using a similar approach as presented above supports the assumption of a localised load concentration in the fibre structure (see Section 4.2). In the bottom part of Fig. 10, a Mohr’s circle for the [2H:1A] loading was constructed using a fibre angle of 558, which is the limit case for fibre realignment after matrix breakdown (found by conducting netting analysis for the particular loading ratio). Matching the stresses parallel to the fibre direction (i.e. sP ¼ sPmax ) for the two considered loading ratios, resulted in nearly identical hoop stress values sH. This does not correspond with the experimental results, where the hoop stress is found to be less in the [2H:1A] loading case. However, incorporating a decreased maximum strength of the fibre structure (i.e. sP , sPmax) gave a representation that quantitatively matched the experimental findings. Such a decrease was already suggested in Section 4.3 based on the presence of additional bending loads near
the specimen end tabs. Depending on the amount of strength reduction, axial stresses at failure, sA, can be higher or lower than the corresponding axial stresses in the [3H:1A] loading case; thus explaining the intersection of the structural failure envelopes for the high and low fibre volume fraction.
6. Conclusions In this investigation the effect of the primary manufacturing parameter ‘winding tension’ on physical and mechanical properties of filament wound [^ 6083]T composite tubulars was studied. Specimens were manufactured using different winding tensions, and mechanical testing was conducted under different loading ratios. Results from the mechanical testing were compiled into functional and structural failure stress envelopes, and the observed mechanical behaviour was analysed qualitatively to show the plausibility of the experimental findings. Linking the applied winding tension to the physical properties of a part, as well as to the mechanical response at failure, led to the following conclusions. † Increasing the filament winding tow tension significantly increased fibre compaction; thus the primary processing parameter ‘winding tension’ is suited for controlling the fibre volume fraction of the manufactured components. † Mechanical testing revealed that stresses at functional and structural failure depend on the applied winding tension. † Comparison of test results from specimens wound with a low and high tow tension indicated that the mechanical properties do not follow a simple unidirectional trend. Generally, an increase in winding tension has a beneficial effect on the component strength in the case of a fibre dominated loading, whereas under a matrix dominated
1622
P. Mertiny, F. Ellyin / Composites: Part A 33 (2002) 1615–1622
loading condition a slight decrease in the failure strength was obtained by increasing fibre compaction.
Acknowledgements This research was made possible by the contributions from TransCanada Pipelines Limited, the Alberta Intellectual Infrastructure Partnership Program (IIPP), the Canadian Foundation for Innovation (CFI), the Natural Sciences and Engineering Research Council (NSERC) and the University of Alberta. In addition, the authors would also like to acknowledge the contributions made by the members of the ACME group and the technical staff at the Department of Mechanical Engineering (J. Wolodko, G. Meijer and B. Faulkner).
References [1] Mertiny P, Ellyin F. Selection of optimal processing parameters in filament winding. In: Proceedings of 33rd International SAMPE Technical Conference. Seattle; November 2001. p. 1084–95. [2] Lee SY, Springer GS. Filament winding cylinders: I Process model. Compos Mater 1990;24:1270–98. [3] Calius EP, Lee SY, Springer GS. Filament winding cylinders: II Validation of process model. Compos Mater 1990;24:1299 –343. [4] Lee SY, Springer GS. Filament winding cylinders: I Selection of the process variables. Compos Mater 1990;24:1344–66.
[5] Cohen D. Influence of filament winding parameters on composite vessel quality and strength. Composites Part A 1997;28A: 1035–47. [6] Cohen D, Mantell SC, Zhao L. The effect of fibre volume fraction on filament wound composite pressure vessel strength. Composites Part B 2001;32B:413–29. [7] Wolodko J, Mertiny P, Meijer G, Martens M, Ellyin F. Development of a facility for filament winding GFRP tubulars. In: Proceedings of 46th International SAMPE Symposium. Long Beach; May 2001. p.1271–82. [8] Soden PD, Leadbetter D, Griggs PR, Eckold GC. The strength of a filament wound composite under biaxial loading. Composites 1978; 9(4):247–50. [9] Soden PD, Kitching R, Tse PC. Experimental failure stresses for ^558 filament wound glass fibre reinforced plastic tubes under biaxial loading. Composites 1989;20(2):125–35. [10] Ellyin F, Wolodko J. Testing facilities for multiaxial loading of tubular specimens. In: Kalluri S, Bonacuse PJ, editors. Multiaxial fatigue and deformation testing techniques. ; 1997. p. 7–24. ASTM STP 1280. [11] Wolodko J. Biaxial fatigue and leakage characteristics of fiber reinforced composite tubes. Ph.D. Dissertation. University of Alberta; 1999. [12] Banerjee A, Sun L, Mantell SC, Cohen D. Model and experimental study of fibre motion in wet filament winding. Composites Part A 1998;29A:251 –63. [13] Hu Y, Xia Z, Ellyin F. Mechanical behaviour of an epoxy resin under multiaxial loadings. Part I: experimental study. Polym Polym Compos 2000;8:11–8. [14] Chen Y, Xia Z, Ellyin F. Evolution of residual stresses induced during curing processing using a viscoelastic model. Compos Mater 2001;35: 522 –42.