Composites: Part A 43 (2012) 748–757
Contents lists available at SciVerse ScienceDirect
Composites: Part A journal homepage: www.elsevier.com/locate/compositesa
Influence of process induced defects on the failure of composite T-joint specimens R.S. Trask ⇑, S.R. Hallett, F.M.M. Helenon, M.R. Wisnom Advanced Composites Centre for Innovation & Science (ACCIS), University of Bristol, Queen’s Buildings, University Walk, Bristol BS8 1TR, UK
a r t i c l e
i n f o
Article history: Received 16 July 2011 Received in revised form 21 November 2011 Accepted 23 December 2011 Available online 16 January 2012 Keywords: A. Polymer–matrix composites (PMCs) B. Defects D. Mechanical testing E. Joints/joining
a b s t r a c t An experimental investigation of the failure load of T-joint structures has been undertaken, with a focus on the influence of process induced defects within the deltoid area. Static pull-off tension tests have been conducted for a comparative assessment. An extensive literature review on the subject to place is also presented. The role of the deltoid area in stabilizing the primary load bearing plies and minimising the volume fraction reduction in the off-axis plies is critical to maximise the failure load and minimise the performance variability of the T-joint structures. A deltoid area reduction of 25%, with no change in the external geometry, yielded a similar mechanical performance to the nominal baseline specimen but with increased variability. However, a reduction of 50% in the deltoid area yielded a strength reduction of 33%. The findings of this study suggest that the reduction in the deltoid area can be tolerated within certain limits. Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction Laminated composites are used in a wide range of structural applications. One of the difficulties in structural design is that of joints, both composite-to-composite and composite to other materials. A particular case is that of a 90° joint, where a laminate or sandwich construction is required to terminate and transfer load to an adjacent laminate or sandwich panel. This is a particularly challenging design case since in-plane loads, typically carried by the fibres in a composite laminate, are required to be transferred to the characteristically much weaker through-thickness direction of the adjoining structure. Additionally bending loads can be imparted at the joint. A variety of solutions have been proposed for this problem which in general can be described as T-joint structures. To date, a wealth of research has been undertaken on: (1) the optimisation of the T-joint geometry; (2) detection and influence of internal defects; and, (3) the role of hydrothermal and processing temperatures on structural performance. These three themes, and a brief introduction of the T-joint structure and terminology, form the composition for the remaining part of this introductory section.
1.1. T-joint definition Composite T-joint structures have found broad application in aeronautics, astronautics and ship construction for a long time. In recent years, researchers have experimentally and/or numerically investigated the failure mechanism and strength behaviour of both ⇑ Corresponding author. E-mail address:
[email protected] (R.S. Trask). 1359-835X/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.compositesa.2011.12.021
glass fibre reinforced plastic (GFRP) and carbon fibre reinforced polymers (CFRPs) laminates. A general representation of the different T-joint structures, and examples of loading conditions, considered within the open literature is given in Fig. 1. This also attempts to capture the nomenclature used to describe the T-joint structure. In general the horizontal component is representative of the primary structure and often described as the hull, flange or skin. The vertical component of the T-joint is referred to as the web or bulkhead and is representative of the internal structure or stiffening arrangement. The web is supported by a fillet, noodle or deltoid section (clearly evident in Fig. 1a) primarily to assist in the formation of the external overlaminate, thus ensuring smooth load transfer between the vertical and horizontal sections. The curved laminate section is sometimes referred to as the boundary angle.
1.2. Optimisation of T-joint geometry and load carrying ability Shenoi and co-workers have extensively investigated T-joint structures typically used in naval applications (Fig. 1a), particularly for the UK mine-countermeasure-vessels (MCMVs), where the most common hull structure configuration is a stiffened single GFRP skin construction with multiple watertight bulkheads. In the early work by Shenoi and Violette [21] their numerical finite element analysis of T-joint structures identified the role of the fillet radius as the primary influence on T-joint structures. Shenoi and Hawkins [20] extended this investigation to consider T-joints under 45° pull-off loading. This work confirmed that the structural behaviour and failure modes of the joints are very dependent upon geometry and material selection as observed in the previous study. The influence of disbonding within the T-joint structure has been
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757
749
Fig. 1. T-joint geometries considered by different authors, (a) GFRP construction considered by Shenoi and co-workers, (b) Whittingham et al. [24] and Kesavan et al. [12] investigated the potential of structural health monitoring, (c) in an effort to suppress damage formation in the deltoid, Qin et al. [18] investigated the influence of an end cap over the vertical web, (d) the influence of different deltoid geometries was investigated by Panigrahi and Pradhan [15], and (e) Davies and Ankersen [7], investigated stiffener/ skin joint strength subjected to pull-off loads.
considered by Dharmawan et al. [9]. The authors undertook a finite element investigation of different disbond locations and their influence on the strain distribution using a crack tip element (CTE) methodology. The authors reported that the critical strains to failure were significantly altered by joint geometry. Phillips and Shenoi [16] considered the role of different fibre architecture (woven roving (WR) cloth and plies of chopped strand mat (CSM) with a polyester Crystic 489 resin) on the failure response of Tjoint structure. The authors noted failure was initiated at the interface of chopped strand mat layers and the woven roving layers within the fillet radius, when loaded under 45° pull-off loading as well as three point-bending. In an effort to enhance the baseline performance of the T-joint specimens, numerous authors have considered the inclusion of dissimilar materials, novel fibre architecture including z-direction reinforcement, and optimisation of the fillet geometry. Blake et al. [4] investigated the static structural response of a GFRP composite T-joint (configuration as detailed in Fig. 1a) containing a viscoelastic fillet using a progressive damage model under 3-point bending. The study highlighted the importance of the overlaminate region for both load transfer and the point of damage initiation with failure governed by interfacial disbonding between the overlaminate and the flange. The experimental results showed good qualitative and quantitative agreement with the predictive damage model. Cope and Pipes [6] investigated the application of an epoxy fillet material in the deltoid region whilst varying the local joint geometry. The application of a smooth or radius deltoid geometry,
compared to a triangular deltoid insert, in joining the vertical and horizontal CFRP laminates maximised joint strength. Typically, a joint employing a 12.7 mm radius at the intersection was found to provide the greatest strength with reasonable stiffness. Panigrahi and Pradhan [15] numerically simulated the failure mechanism of modified elliptical adhesive deltoid GFRP T-joints by means of a 3D finite element analysis (FEA) for location of damage onset. Through the variation of the width/height (b/h) ratio of the elliptical profile the authors observed a reduction in the peak normal stresses by increasing the base width, although this approach tended to promote mode I failure at the end of the overlaminate. The authors recommended the application z-reinforcement to avoid this induced failure mode. Kumari and Sinha [13] applied linear elastic finite element analysis using a thick shell element specially developed to compute all the stress components in each ply of the T-joint specimen and predict ply failure. The authors noted that an increase in radius between two fixed end points of arc length at the web/skin interface increases the pull off strength of the T-joint since the increasing radius reduces the stress concentration at the web/skin interface. Although the failure load can be controlled by the radius arc length and the end points of the overlaminate, the overall stress patterns and location of failure within the joint was not observed to alter. This confirmed that the web/skin interface zone is the critical failure zone and that the radius arc length and its end points have to be chosen efficiently to derive the maximum strength. Davies et al. [8] developed a new interface element with a monotonic
750
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757
(exponential) force/displacement to model the onset of delamination/debonding in composite T-joint structures and subsequent propagation. The authors showed that the interface element initiates and propagates correctly, without needing an initial flaw. Davies and Ankersen [7] extended their earlier research with the inclusion of experimental validation of different fillet radii and different web thicknesses. As previous authors identified, the T-joint’s strength depended crucially on the stress concentrations in the triangular region bounded by the two fillet radii and the skin: the fillet or deltoid region (see Fig. 1f). In an effort to improve the baseline performance of T-joint structures Xu et al. [25] investigated the tensile and shear strength performance of stitched, non-stitched and co-bonded T-joints manufactured by resin transfer moulding, RTM. In terms of mechanical properties, the authors reported the RTM-made T-joint holds superiority over both the stitch-RTM and co-bonding methods. Stickler and Ramulu [23] analyzed the tensile strength of transverse stitched T-joints both experimentally and through numerical simulation. Through the variation of the fibre insertion tow modulus, the filament count and insertion depth, the authors were able to increase the damage initiation threshold under flexural loading, whilst the T-joint deflection could be controlled by varying the fibre insertion depth. Allegri and Zhang [1] investigated the benefits of Z-fibre pinning to improve the bonding strength of composite T-joints. The authors reported that the insertion of Z-fibres in the flange-to-skin bond improved the damage tolerance capabilities by increasing both ultimate strength and the failure pull-off displacement. In conclusion it was noted that Z-fibres are particularly effective in retarding the delamination/debond growth in the flange-to-skin bond in pure mode I loading; nevertheless, the onset of debonding cracks was not influenced by the Z-fibre’s presence. The design of T-joint specimens with complex 3-dimensional woven fibre architecture was considered by McIlhagger et al. [14]. In this study, the authors successfully manufactured 3-D woven reinforced T-joint specimens with high fibre volume fractions and low void content using autoclave manufacturing; although a number of manufacturing problems had to be resolved to achieve this level of quality, i.e. the caul plate radius must be kept to a minimum and high processing pressures must be applied uniformly to eliminate any resin rich areas at the web/ flange intersection. Soden et al. [22] considered the effect of through-the-thickness reinforcement on the in-plane strength of composite T-shaped specimens, woven on conventional weaving apparatus. The authors found the in-plane strength of 3D woven T-sections to be superior to laminated samples. The authors also noted increasing the levels of through-thickness interlinking promoted a decrease in the load required for damage initiation but with an increase in the peak failure load. The application of braided fibre architecture within a T-joint specimen was assessed by Chen et al. [5] using a finite element cohesive interface model to predict delamination failure. The results indicate the application of a cohesive model is feasible although careful consideration is required with the mixed-mode failure criterion. Zimmermann et al. [26] have investigated ultra-thick CFRP noncrimped fabric (NCF) composite T-joints (where material thickness exceeds 60 mm) tested in three load cases: tension, compression and axial bending. Due to the presence of transverse shear and normal stresses a 3D modelling approach using a commercial FE code (MSC Marc) was employed. The experimental results indicated specimen failure for the tension case was initiated at the deltoid tip and propagated through the web; in compression, delamination damage was contained in both fillet radii; in the asymmetric bending case, the authors noted a more complex stress combination exists which was observed to initiate delamination damage within the fillet radius on the tension side of the specimen (at 1/3rd of the laminate thickness). In the 3D modelling the transverse shear
and normal stresses are identified as the main failure cause and the study employed solid composite brick elements as an efficient way to model thick structures. The authors acknowledged that this methodology was incapable of calculating accurate shear stresses on a ply level; they maintain that usable results were achieved by discretisation of the component with multiple elements through the thickness. 1.3. Influence of hydrothermal and processing temperatures on performance A further consideration on the evolution of the damage within a T-joint geometry concerns the influence of environmental effects. Hygrothermal environments have been considered by Rao et al. [19] on the tensile pull-off force required for T-joint failure. In this study the specimens were exposed to a controlled environment until they attained a saturated moisture concentration of 0.7% (typically 60 days). The authors reported a 20% reduction in the joint strength which they attributed to the presence of residual hygrothermal stresses, although the failure location remained the same. The influence of thermal variation during the manufacturing cycle has been considered by Apalak et al. [2]. The authors numerically analyzed the nonlinear thermal stresses of T-joints in a nonhomogeneous temperature field. The mechanical and thermal mismatch of the adhesive and adherend materials gave rise to incompatible thermal strains along the adhesive–adherend interfaces. High thermal stress distributions through the adhesive and adherend regions in the neighbourhood of the adhesive–adherend interfaces were observed by Apalak et al. [2]. This study was extended further [3] to consider the thermal-nonlinear elastic stress analysis of specimens loaded through different support conditions. The increased support length did not have the effect of reducing the normal and shear stresses in the joint, and in fact the authors identified the occurrence of the stress concentrations inside the adhesive fillets and in the regions of the adhesive layer-composite plate interface as the critical regions for design consideration. The authors proposed modification to the edge geometry of the joint would be more effective in reducing the peak stresses for the considered thermal and structural boundary conditions. 1.4. Concluding observations on current scientific understanding The proceeding discussion has endeavoured to capture some of the numerous investigations (whether experimental, analytical or numerical simulation) that have been undertaken on laminate Tjoint structures. In general, these research programmes have tried to understand the use of different designs and geometries, and the influence of different material types on the structural performance of the T-joint specimen. This review has identified an absence of research into the influence of manufacturing defects on the mechanical performance of the T-joint specimen. The control of manufacturing variability, whether autoclave processing or via the resin infusion processes, is critical to the widespread acceptance of polymer based composite materials as a material of choice for advanced structural applications. Potter et al. [17] generated a taxonomy of defect states in composite parts, recommending the need to re-categorise the current definition of defects into design induced defects, which might more properly be regarded as features, and process induced defects which are truly defect states that can be influence by correct choice of manufacturing conditions. In the context of T-joint structures, the nature of the design promotes both design induced defects, namely geometrical errors (fibre misalignment; both wrinkled fibres and fibre direction errors around the fillet) and spring-in distortion and residual stresses, and process induced defects, such as voidage, delamination and consolidation pressure variation around the fillet [17].
751
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757 Table 1 Prepreg ply orientations and thicknesses for T-joint specimens (where 0° in the web section is aligned to the primary loading direction as illustrated in Fig. 2). Generic designation
Position designation
Prepreg lay-up
Thickness (mm)
Vertical section (web)
Outer wrap Outer 0° layer 90° Layer Inner 0° layer 90° Deltoid Platform laminate
Two layers of ±45 Seven layers of 0 Four layers of 90 Three layers of 0 90° Deltoid (60/0/ 60/0)3s
0.500 0.875 0.500 0.375 – 3.000
Deltoid Base panel (flange)
The work presented here investigated the influence of process induced defects (located within the deltoid region) on failure mechanisms occurring in a laminated composite T-joint structure, when loaded in tension pull-off. This was achieved by varying the amount of material that was placed in the deltoid region of a Tpiece specimen for a fixed external geometry. In addition to extending the technical understanding of process-induced defects and their interaction on (1) design features and (2) damage modes, the experimental data has been used for input to and validation of finite element models [10]. The reader is referred to this publication for further analysis of this complex problem.
2. Materials and experimental details 2.1. Materials and manufacturing processes
20mm diameter loading roller
40 mm
The material selected for the present study was Hexcel’s IM78552 carbon epoxy unidirectional pre-preg tape with a nominal ply thickness of 0.125 mm (typical of an aerospace grade
composite material). The material was cured in an autoclave according to the manufacturer’s specifications, with temperature stages of 120 °C for 1 h, followed by 180 °C for 2 h. A temperature ramp of 2 °C per minute was used for all stages. A pressure of 7 bar was applied for the duration of the cure cycle. The exact ply orientations, thickness and stacking sequence used within this study are given in Table 1 with Fig. 2 illustrating the manufactured T-joint specimen. The T-joint specimen was manufactured in three key stages. This process is shown schematically in Fig. 3. The first stage involved the manufacture of the platform, namely (60/0/ 60/0)3s. Once fully cured, the platform was cut to size and prepared for the assembly of the perpendicular sections to form the T-joint profile. One surface of the platform was prepared for bonding by shot blasting with grit followed by degreasing with acetone. Once completed the platform was ready for the final assemble stage of manufacturing. The second stage of manufacturing involved the manufacture of the deltoid insert. A length of prepreg was cut to match the area of the deltoid as calculated from the nominal geometry. This was then heated and rolled to form a rod. The prepreg
40 mm
(a)
(b)
±452 07 904 03
(60/0/-60/0)3s
(c)
(d)
Fig. 2. (a) T-joint experimental testing configuration, (b) schematic illustration of the test set-up. (c) Microscopy section detailing fibre stacking sequence in platform (60/0/ 60/0)3s and overlaminate in the deltoid lower corner of nominal T-joint specimen, and (d) schematic representation of lower corner of nominal T-joint specimen. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
752
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757
Stage 1: Assembly and autoclave manufacture of platform
Stage 2: Step 1: manufacture of deltoid
Stage 4: Removal of T-joint specimens from autoclave and prepared for testing
Stage 2: Step 2: manufacture of vertical section
Stage 3: Assembly ready for autoclave curing of T-joint specimens
Fig. 3. Schematic representation of the T-joint manufacturing process, (Stage 1) assembly and autoclave manufacture of 16-ply composite platform using flat tooling, (Stage 2, Step 1) RT manufacture of UD composite deltoid, consolidated against steel roller under 1Bar vacuum pressure, (Stage 2, Step 2) lay-up of composite plies against prepared tooling forming the overlaminate and half of the vertical section, (Stage 3) assembly of left and right vertical sections, with deltoid and composite platform (grit blasted on lower surface) located into position for autoclave processing, (Stage 4) removed T-joint specimen ready for trimming operations and machining into individual samples. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
rod was then placed in the cavity formed by two touching 10 mm diameter steel bars. Finally, the prepreg was heated and vacuum consolidated against the tooling to form the deltoid shape. The third stage involved the lay-up of the two halves of the vertical section and upper part of the platform over two (degreased) aluminium tooling blocks with a 5 mm radius. To minimise any manufacturing variation within the vertical section the individual prepreg layers were grouped together (typically in threes or fours depending upon their position within the stacking sequence) and vacuum consolidated (away from the formers) prior to their layup over the aluminium tooling blocks. Once laid-up, the two halves of the vertical section were pushed together, the deltoid material was inserted into position (from stage 2), and the grit blasted and degreased side of the platform (from stage 1) was placed directly on top. The assembly was then prepared and bagged-up for the autoclave curing cycle. Once removed from the autoclave and de-bagged, the edges of the assembly were cleaned to remove any composite flash using a diamond saw. At this stage it was noted that the platform section had a slight degree of ‘spring-in’, i.e. the platform had a slight curvature upwards towards the aerofoil section. A typical T-joint specimen by the described method is shown in Fig. 2. After manufacturing, the T-joints were stored in a temperature but not humidity controlled environment prior to machining (typically this ranged from 3 days to 10 days). The T-joint assembly (250 mm wide) was cut into individual specimens – overall height 80 mm, length 130 mm, with a nominal width of 20 mm, using a diamond coated slitting saw and water soluble oil coolant. The specimen width was selected to follow similar experimental and numerical investigations undertaken within the composite group
at Bristol. These results have been reported in Hill et al. [11] and Chen et al. [5]. It should be noted that any defect generated as a result of varying the deltoid geometry will run the entire length of the moulding and are therefore prismatic in nature, thus independent of specimen thickness. The void type defects in contrast are significantly smaller (typical ranging from 10 lm to 50 lm, with some isolated voids as large as 0.25 mm in diameter – (this is discussed in Section 4) and their size relative to the specimen width means that the 20 mm specimen width is sufficiently wide to not be affected by the location of individual defects in the width direction. Furthermore, in practice much of composites failure is dominated by free edge effects which are independent of specimen width, in which case it would not improve the configuration to use a wider specimen. The magnitude and influence of these free edge stresses are discussed in Hélénon et al. [10] where finite element models are able to describe the relevant stress fields. When machining, the inverted assembly was firmly clamped against the flat base-plate of the milling machine, applying a slight force to the curved platform section. It was observed that the platform curvature was identical in all specimens regardless of deltoid area and typically exhibited a spring-in of 0.5°. Over the testing gauge length the curvature was minimal and consistent. 2.2. Tension pull-off tests The individual T-joint specimens were tested by applying a tension (pull-off) load to the vertical (web) section while the base panel (flange) was simply supported by rollers at either end. The load was applied in displacement control at an extension rate of 1 mm per minute until the initiation of the first failure mechanism was
753
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757
recorded. The exact loading configuration is shown in Fig. 2. In this configuration the perpendicular section is clamped (distance 40 mm from the underside of the platform) and the two loading rollers are positioned 80 mm apart. To position the specimen within this loading arrangement, the right-hand loading roller is released and moved out to allow the specimen to fit within the hydraulic grips. The right-hand loading roller is then repositioned 80 mm away from the other loading roller using a set of digital callipers. Throughout the tests both the left hand roller and the upper hydraulic grips are fixed as datum points. In these tests only the position of the right hand roller could introduce an experimental error (typically ±0.5 mm). 3. Experimental results 3.1. Deltoid area definition The amount of material inserted into the deltoid area was altered to ascertain its influence on the failure load and failure mode on the T-joint specimens. Quite extreme variations in the deltoid area were used to gauge the influence of severe manufacturing anomalies, namely a 25% and 50% reduction in the deltoid area. In reality, both test geometries were slightly different from their targeted areas, namely 26% (range 24–29% below nominal), and 46% (range 44–49% below nominal). To appreciate the deltoid variation Fig. 4 has been included for comparative purposes. In this investigation, the nominal deltoid area has been calculated to be 22.9 mm2 from the known fillet radius and number of plies used in the construction of the vertical section, assuming the individual plies consolidate to the manufacturer’s recommended thickness of 0.125 mm. The ‘nominal’ specimens were manufactured with the aim of obtaining the optimised deltoid area of 22.9 mm2. The results shown in Table 2 indicate the deltoid area was within 5% of this target indicating that the consolidation process during assembly and autoclave processing is within manufacturing tolerances to obtain the correct packing density and hence volume fraction required. 3.2. Batch quality The reduction in material in the deltoid area had a pronounced effect on the consolidation and the overall quality of the plies within the deltoid and the surrounding ‘fillet’ region of the specimens. Fig. 4 indicates that the greatest influence was observed on the 0° and 90° fibre stack, both in terms of thickness profile and variation around the deltoid. To determine the exact magnitude, a series of measurements were taken to capture the ply variations within this region. Firstly, in an effort to understand the positional control of the different ply groupings around the fillet, a series of measurements were made from a datum point, exactly at the mid-point on the lower edge of the deltoid, up through the fillet profile at an angle of 45°. This is illustrated in Fig. 5, positions 1–4, and
7.25mm 5mm radius radius
7.25mm 5mm radius radius
2mm
Table 2 Summary of deltoid quality for the three different specimen configurations. Number of specimens given in brackets. Specimen ID
Theoretical Nominal 26% Below nominal 46% Below nominal
Deltoid area (mm2) Mean
CV
% Difference from theoretical
22.9 24.0 (5) 16.8 (7) 12.3 (8)
– 1.0 3.5 4.4
– 4–6% 24–29% 44–49%
reported in Table 2. Secondly, the change in thickness profile of the 90° fibre stack, on both the left hand side and the right hand side of the deltoid was considered. The fibre stack thickness has been measured in four positions around the fillet radius, indicated as positions (a–d) in Fig. 5, and reported in Table 3, to capture the influence on the neighbouring plies. The results in Tables 2 and 3 are reported the average measured values and coefficient of variation (CV) of at least five specimens, and the percentage difference from the theoretical nominal. Tables 3 and 4 clearly show the influence of the under-sized deltoid on the neighbouring plies. Interestingly, the absence of the full ‘quota’ of reinforcement in the deltoid has not resulted in a reduction in the packing density of this region but rather a general shift of the ply interfaces towards the datum point (as indicated in Table 3 and shown more clearly in Fig. 4) and the collapse of the packing density in the 90° ply, almost double the nominal value (as indicated in Table 4). The three photographs in Fig. 4 are all sized to the same scale, with a radius profile included to indicate the correct location for the inside edge of the 0° ply. In comparison to the nominal deltoid (Fig. 4 right) the specimens with the deltoid area at 26% below nominal (Fig. 4 centre), indicates a shift of a single ply thickness (0.125 mm) inside the original starting position for the first 0° ply stack. In the case of the ‘46% below nominal’ specimens, these equally exhibited a shift of the innermost 0° plies by as much as 0.3 mm towards the centreline resulting in an increased thickness of the neighbouring 90° fibre stack. The reduction in the packing density has introduced varying degrees of voidage and porosity in both the deltoid, the deltoid/0° ply interface, as well as within the neighbouring 90° plies. Critically, it is the volume fraction of the 90° ply stack and the position of the 0° load bearing plies which are outside the manufacturing tolerance, the remainder of the plies appear to have maintained their required positioning tolerances within the T-joint geometry. 3.3. Mechanical test results The results for the pull-off T-joint tests, for the three different deltoid areas, are reported in Fig. 6 with the failure loads plotted against deltoid area variation (five nominal specimens, seven specimens 26% below nominal, and eight specimens 46% below
2mm
2mm
5mm radius 7.25mm radius
Fig. 4. Variation in deltoid area in T-joint specimens (photographs taken at same scale/magnification). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
754
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757
(a)
(b) (1) (2) (3)
(c) (d)
(4) Datum point directly below deltoid tip
Fig. 5. Cross-section of ‘nominal’ T-joint specimen illustrating key measurement positions: RHS labelled (1–4) the location of the individual measurement points used in Table 3 from the central datum point (note the positioning of the arrows has been off-set to aid visualisation). LHS labelled (a–d) the four measurement positions used in Table 4 for assessing the thickness of the 90° ply stack. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Table 3 Positional locations of the deltoid edge, the inside and outside edge of the 90° ply stack and the outer edge of the fillet as indicated in Fig. 4 for all three deltoid configurations. Specimen ID
Theoretical Nominal
26% Below nominal
46% Below nominal
Position 1 Distance to sample edge from datum
Position 2 Distance from datum to outside edge of 90° ply stack
Position 3 Distance from datum to inside edge of 90° ply stack
Position 4 Distance from datum to deltoid edge
5.63 5.24 2 7
3.88 3.96 2 +2
3.38 3.53 3 +4
3.00 3.14 3 +5
Average
5.14
3.73
3.04
2.67
CV % Difference from theoretical
2
3
3
3
9
4
Average
5.14
CV % Difference from theoretical
2
Average CV % Difference from theoretical
10
11
12
3.65
2.74
2.34
1
4
5
6
23
28
Table 4 Summary of the variation in the 90° fibre stack thickness profile measured at the four positional locations (a, b, c and d) as indicated in Fig. 4b for all three deltoid configurations. Specimen ID
Theoretical Nominal 26% Below nominal 46% Below nominal
Thickness measurement of the 90° ply stack at key locations (mm)
Average % Difference from theoretical Average % Difference from theoretical Average % Difference from theoretical
Position (a)
Position (b)
Position (c)
Position (d)
0.508 0.47 7 0.48 4 0.54 +8
0.508 0.46 9 0.60 +19 0.75 +49
0.508 0.46 9 0.65 +30 0.98 +97
0.508 0.56 +10 0.84 +69 0.87 +75
nominal were tested). The T-joints specimens all behaved in a linear manner up to the point of maximum load when failure occurred instantaneously at which point the test machine was stopped and the specimen unloaded. The load–displacement response of all specimens was linear to failure with all specimens exhibiting the same bending stiffness, i.e. there was no significant difference in the platform compliance. Typically failure occurred by deltoid cracking and cracking through the 90° ply block. As detailed in Fig. 6, the nominal deltoid T-joints failed at 1330 N (CV = 9%), those with the deltoid at 26% below nominal
failed at 1296 N (CV = 18%) whilst those with the deltoid at 46% below nominal failed at 891 N (CV = 11%), a reduction of 33% in the load carrying capability of the nominal specimens. The trend observed in Fig. 6 would suggest that the reduction in the deltoid area can be tolerated within certain limits, with the acceptance of increased variation (scatter) in the mechanical performance of the T-joint. It was noted that reduction of deltoid area also caused the introduction of competing failure locations, i.e. vertical cracking in the deltoid or cracking in the 90° fibre stack with deltoid corner cracking, in part responsible for the increased scatter. However,
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757
755
Fig. 6. Pull-off load as a function of variation of deltoid area away from nominal. The insert photographs are indicative of the failure modes for each specimen configuration. All photos are fixed at the same scale. T-joint specimen with 26% below nominal deltoid area exhibited two distinctive failure modes (enlarged in Fig. 7) spanning the full range of experimental results for this specimen configuration. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
once beyond this manufacturing limit the penalty reduction in mechanical performance is clearly severe. 4. Failure analysis investigation and discussion A microscopy failure analysis was undertaken to determine the exact failure mode, and ascertain whether the reduction in deltoid packing density promoted a shift in failure position. After mechanical testing, all specimens from the different configurations were ground by hand in steps from 600 SiC grit up to 2500 SiC grit in 2 min intervals. After grinding, the specimens were polished using 6-lm and 3-lm diamond paste each for 10 min duration and viewed using an optical microscope (Olympus SZX16 microscope with ColorView camera). Representative micrographs of the three different specimens have been included in the discussion below. Fig. 7 is an example of the failure mode exhibited by the nominal deltoid specimens, i.e. cracking located in the lower right hand corner of the deltoid, with two specimens also exhibiting some secondary cracking appearing in the 90° ply stack directly above the deltoid crack. Whilst the deltoid area has been designated ‘nominal’ in terms of area, the actual control of the geometric shape is still very dependent on the ability of the manufacturer. In Fig. 7 it is apparent that the packing density in the vicinity of the upper deltoid tip has been successfully maintained, however the control at the two lower deltoid corners is less ideal (as indicated in
Fig. 7b). The variation in ply packing density and the local variation in ply positional control are both contributory factors for damage initiation and hence the final failure load recorded for this specimen configuration. In the analysis of the failure position it was noted that whilst the position of the deltoid cracking varied specimen-to-specimen (as a result of the manufacturing process), one common feature was observed, the deltoid cracking was always in the vicinity of the thinnest section of the 90° stack (see Fig. 7b). The exact reason for this failure position requires further analysis but may be due to (1) the local increase in fibre volume fraction (Vf) in the 90° stack (arising from the poor formation of the lower deltoid tip geometry), (2) a localised Vf reduction or void in the deltoid area, or (3) the slight variation in the smooth profile of the 0° plies may have altered the load transfer path and induced a stress concentration sufficient to trigger damage initiation in the deltoid. As discussed in Section 3.2, the reduction in deltoid area (for the 26% below nominal case) resulted in a local increase in thickness of the 90° stack around the fillet. In these specimens, the occurrence of two distinctive failure modes was observed, namely at higher failure loads the damage was observed to be located through the deltoid tip (Fig. 8a), whilst towards the lower end of the failure load spectrum damage was observed to be located within the 90° stack with additional damage propagating up into the vertical section and triggered in the deltoid corner (Fig. 8b). The precise
Fig. 7. (a) Nominal deltoid geometry with representative deltoid corner cracking, (b) close-up view of corner cracking and variation in localised Vf above the crack in the 90° stack. It should also be noted that in the vicinity of the crack the 0° plies do not maintain a smooth profile around the fillet radius.
756
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757
Fig. 8. T-joint specimen with 26% below nominal deltoid area, (a) central failure crack propagating through the deltoid, high failure load, i.e. 1496 N, (b) deltoid corner failure with crack propagation extending up through the 90° fibre stack, low failure load, i.e. 1183 N.
1 mm Fig. 9. T-joint specimen with 46% below nominal deltoid area indicating the location of the failure cracks and crack–void interaction in the 90° plies. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
reasoning behind the variation between these two distinctive failure locations within the same specimen type is still uncertain. The internal positioning of the individual plies, the geometrical shape (height, width and edge profile) of the deltoid, and the extent of localised voidage (and location) will all have a direct bearing on the point of damage initiation within the global geometry, and hence the load carrying capability of the T-joint configuration. In the specimens with the deltoid 46% below nominal, the failure mode was once again located towards the lower deltoid tip and then within the 90° ply but critically the damage formation has interacted with the voids now occurring within the specimen, as illustrated in Fig. 9. In these specimens the magnitude of the voidage had a direct influence on damage initiation and hence load carrying capability of the T-joint configuration. 5. Conclusions In the course of this experimental investigation, three different batches of T-joint specimens have been manufactured and
assessed using a static tensile pull-off test to understand the role of process induced defects. In the context of this CFRP T-joint geometry with this fibre stacking sequence, the study has highlighted the role of manufacturing variability on failure. Reductions in deltoid area introduced deltoid and ply voidage through uncontrolled consolidation, allowing ply movement and fibre misalignment to develop. A deltoid area reduction of 26%, with no change in the external geometry but a shift of the innermost 0° plies of 0.3 mm, has yielded similar mechanical performance as the nominal baseline specimen but with increased statistical variation. However, a reduction in the deltoid area of 46% and a shift of the innermost 0° plies by almost 0.8 mm, again with no visible change to the external geometry, has yielded a 33% reduction in the load carrying capability of the T-joint compared to the nominal baseline specimens. The findings of this study suggest that a reduction in the deltoid area can be tolerated within certain limits, with the acceptance of increased variation in mechanical performance. However, once the process variations start to generate significant defects the reduction in mechanical performance becomes severe.
R.S. Trask et al. / Composites: Part A 43 (2012) 748–757
Acknowledgments The authors acknowledge the support of Rolls-Royce plc in funding this work through the Composites University Technology Centre (UTC) at the University of Bristol, UK. Also to Mike Jones at the University of Bristol for his technical understanding and assistance throughout this programme of work, Ian Chorley for composite ply cutting and Simon Chiltern for the machining of the specimens. References [1] Allegri G, Zhang X. On the delamination and debond suppression in structural joints by Z-fibre pinning. Composites Part A 2007;38(4):1107–15. [2] Apalak MK, Apalak ZG, Gunes R. Thermal and geometrically nonlinear stress analyses of an adhesively bonded composite Tee joint with double support. J Adhes Sci Technol 2003;17(7):995–1016. [3] Apalak MK, Apalak ZG, Gunes R. Effect of adhesive free-end geometry on the initiation and propagation of damaged zones in adhesively bonded lap joints. J Thermoplast Compos Mater 2004;17:103–34. [4] Blake JIR, Shenoi RA, House J, Turton T. Progressive damage analysis of teejoints with viscoelastic inserts. Composites Part A 2001;32:641–53. [5] Chen J, Ravey E, Hallett S, Wisnom M, Grassi M. Prediction of delamination in braided composite T-piece specimens. Compos Sci Technol 2009;69:2363–7. [6] Cope RD, Pipes RB. Design of the composite spar wing skin joint. Composites 1982:47–53. [7] Davies GAO, Ankersen J. Virtual testing of realistic aerospace composite structures. J Mater Sci 2008;43(20):6586–92. [8] Davies GAO, Hitchings D, Ankersen J. Predicting delamination and debonding in modern aerospace composite structures. Compos Sci Technol 2006;66:846–54. [9] Dharmawan F, Li HCH, Herszberg I, Joh S. Applicability of the crack tip element analysis for damage prediction of composite T-joints. Compos Struct 2008;86(1–3):61–8. [10] Hélénon FMM, Wisnom MR, Hallett SR, Trask RS. Numerical investigation into failure of laminated composite T-piece specimens under tensile loading. Composites Part A, submitted for publication.
757
[11] Hill GFJ, Wisnom MR, Jones MI. Failure prediction of composite T-piece specimens. In: Proceedings of the 5th international conference on deformation and fracture of composites; 1999. 10 p. [12] Kesavan A, Deivasigamani M, John S, Herszberg I. Damage detection in T-joint composite structures. Compos Struct 2006;75:313–20. [13] Kumari S, Sinha PK. Finite element analysis of composite wing T-joints. J Reinf Plast Compos 2002;21(17):1561–85. [14] McIlhagger R, Hill BJ, Brown D, Limmer L. Construction and analysis of 3dimensional woven composite-materials. Compos Eng 1995;5(9):1187–97. [15] Panigrahi SK, Pradhan B. Delamination damage analyses of FRP composite spar Wingskin joints with modified elliptical adhesive load coupler profile. Appl Compos Mater 2008;15:189–205. [16] Phillips HJ, Shenoi RA. Damage tolerance of laminated tee joints in FRP structures. Composites Part A 1998;29A:465–78. [17] Potter K, Khan B, Wisnom M, Bell T, Stevens J. Variability, fibre waviness and misalignment in the determination of the properties of composite materials and structures. Composites Part A 2008;39:1343–54. [18] Qin T, Zhao L, Huang H. Damage investigation and design of woven composite bonded joint. Key Eng Mater 2010;417–418:861–4. [19] Rao VVS, Krishna Veni K, Sinha PK. Behaviour of composite wing T-joints in hygrothermal environments. Adv Eng Aerospace Technol 2004;76(4):404–13. [20] Shenoi RA, Hawkins GL. Influence of material and geometry variations on the behavior of bonded tee connections in FRP ships. Composites 1992;23(5):335–45. [21] Shenoi RA, Violette FLM. A study of structural composite tee joints in small boats. J Compos Mater 1990;24(6):644–66. [22] Soden JA, Weissenbach G, Hill BJ. The design and fabrication of 3D multi-layer woven T-section reinforcements. Composites Part A 1999;30:213–20. [23] Stickler PB, Ramulu M. Damage progression analyses of transverse stitched Tjoints under flexure and tensile loading. Adv Compos Mater 2006;15(2):243–61. [24] Whittingham B, Li HCH, Herszberg I, Chiu WK. Disbond detection in adhesively bonded composite structures using vibration signatures. Compos Struct 2006;75:351–63. [25] Xu C, Junjiang X, Bo P, Zeling C, Hongyun L. Mechanical properties of RTMmade composite cross-joints. Chin J Aeronaut 2009;22:211–7. [26] Zimmermann K, Zenkert D, Siemetzki M. Testing and analysis of ultra thick composites. Composites Part B 2010;41:326–36.