Composites: Part B 30 (1999) 523–534

Design and manufacturing considerations for ply drops in composite

structures
D.S. Cairns*, J.F. Mandell, M.E. Scott, J.Z. Maccagnano
Montana State University, DOE EPSCoR Wind Energy Cluster, Bozeman, MT 59717, USA
Received 17 February 1998; accepted 28 May 1998

Abstract
Thickness variations are required to optimize the design of modern laminated composite structures. These thickness variations are
accomplished by dropping plies along the length to match varying in-plane and bending loads. This results in a structure which is matched
to stiffness and loading requirements. Unfortunately, these ply drops produce internal and local stress concentrations as a consequence of
geometric discontinuities and shear lag. In this study, we explore various factors for design of composite structures with ply drops. These
factors include: thicknesses, ply stacking sequences, ply drop geometries and manufacturing considerations. In addition, fatigue loading is
considered with respect to delamination initiation and growth. A strong sensitivity to the position and the manufacturing details of ply drops
is shown for fatigue damage initiation and growth. All studies were conducted on a low-cost E-glass/polyester composite system. The results
indicate that it will be difficult to completely suppress damage and delamination initiation in service. However, it was found that, in many
cases, there is a threshold loading under which there is little growth after initiation is noted. Factors affecting this threshold are analyzed via
the virtual crack closure method in Finite Element Analysis and verified experimentally. Design rules for ply dropping are presented on the
basis of these results. q 1999 Published by Elsevier Science Ltd. All rights reserved.
Keywords: Ply-drops; B. Delamination; Strain energy release rate; B. Fracture; B. Fatigue

1. Introduction In addition to this, a thicker material was constructed to

represent current industry trends for low-cost, high-volume
Ply drops in composite materials are a relevant design production of composite structures. For applications, differ-
consideration in many structures, especially those with ent methods of stopping and arresting delaminations were
varying loads throughout the structure. As an example, explored to determine if they were practical solutions to
each ply in a composite laminate without bending needs stopping or slowing delaminations.
to carry the same amount of strain when a ply is dropped.
There is a reduction in area that places the thinner section
under a higher stress, transitioning via shear lag. The ply 2. Description of specimens
drop introduces a stress concentration, causing a crack to
form and propagate along the layer that forms the ply drop. In this experiment, all the specimens were made using
There are three possibilities for crack growth: resin transfer molding (RTM) equipment. The plies used in
this experiment were made by Knytex in Texas (a wholly-
1. The driving force is high enough to propagate the crack
owned subsidiary of Owens/Corning, currently named
until the specimen fails.
Owen Corning Fabrics). These plies are made from
2. If the critical level is not reached, the delamination will
Owens/Corning E-Glass fibers in various preform config-
start but will arrest at a certain length and remain
urations. The zero degree layers were D155 material and
constant at that length (arrest).
the ^ 458 layers were DB120. The matrix was a room
3. The stress is so low that delamination never initiates
temperature curing polyester resin manufactured by Inter-
within 500 000 cycles.
plastic Corporation in Minnesota (CoRezyn 63-AX-051).
In this study, different laminate configurations were After the plates were made, 25.4 mm wide specimens
tested that included both internal and external ply drops. were machined and the edges were polished to minimize
edge effects from the machining. The coupons were then
* Corresponding author. post cured at 608C for 2 h. Post-curing assures that the
1359-8368/99/$ - see front matter q 1999 Published by Elsevier Science Ltd. All rights reserved.
PII: S1359-836 8(98)00043-2
524 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534

Table 1
Ply and adhesive material properties

Material E11 E22 (GPa) E33 (GPa) v12 v13 v23 G12 (GPa) G13 (GPa) G23 (GPa) t (mm)
a
D155 08 ply 38.4 8.9 8.9 0.29 0.29 0.29 7.4 7.4 4.6 0.45
DB120 ( ^ 45)8 ply a 11.8 11.8 7.9 0.43 0.28 0.28 7.4 4.1 4.1 0.68
Adhesive b 2.6 0.31
a
Nominal Vf ˆ 45%
b
Isotropic

styrene in the polyester has cross-linked to attain maximum laminates tested and the properties of each laminate, includ-
composite strength. Nominal material properties may be ing a description. In these initial configurations there was
found in Table 1. only a single ply drop and the material was the same thick-
The ply configurations were chosen by selecting the most ness as the DD5 material. A minimum of five (5) test repli-
fatigue resistant material from previous studies [1]. This cates were used for the studies.
was the DD5 material with a 38% fiber volume in a [0/ ^ In Fig. 2, it is shown how the laminates behaved, up to
45/0]s laminate configuration construction using the fabrics delamination initiation. The strain energy release rates (Gc)
as mentioned earlier. The fatigue data for the DD5 material were determined from the critical load (Pc) using the follow-
is illustrated in Fig. 1. ing strength of materials formula from Ramkumar and
Whitcomb [2], and later by Rhee [3].
A2 E2 2 A1 E1
3. Static tests Gc ˆ P2c 1†
2wA2 E2 A1 E1
The first test conducted was a static test which was used Where: the subscript 1 is for the thin section and 2 is for the
to determine the critical load for delamination initiation and thick section of each test coupon, the A is the cross-sectional
to determine how a delamination in each configuration area and E is Young’s modulus, respectively, for each part
would form and grow. Micro-Measurements WK series of the specimen, and w is the width of the coupon.
strain gages were mounted on the thick and thin sections A diagram of the model used can be seen in Fig. 3. The
of the coupons to obtain a more accurate measurement of corresponding nominal critical strain energy release rates
strain than could be obtained by using an extensiometer. The for static delamination initiation are 0.12, 1.34 and
initial tests were the ESA, ESB and ESC laminates. The test 1.61 N mm 21 for the ESA (single interior ply drop), ESB
matrix was expanded to include such effects as thickness, (interior ply drop of first continuous interior 08 ply), and
manufacturing parameters and repair. A guide to all of the ESC (center 08 ply drop), respectively. It is interesting to
laminates is given in Table 2, which contains a list of the note that the 08 Double Cantilever Beam (DCB) test and the

Fig. 1. Fatigue data for DD5 laminate.

 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534 525

Table 2
Laminate configurations, specifications and descriptions

Laminates Lay-up Ex (GPa) Vf (%) Average Comments

thickness
(mm),
thick/thin

ESA [0 a/(0/ ^ 45/0)s] 24.0 35.2 3.81/3.10 Single exterior ply drop
ESB [0/0 a/ ^ 45/0/0/ ^ 45/0] 24.8 35.1 3.7/3.00 Single interior ply drop
ESC [0/ ^ 45/0/0 a/0 ^ 45/0] 24.3 35.3 3.61/3.02 Single center ply drop
ESE [0 a/(0/ ^ 45/0)3] 30.5 39.7 4.52/3.81 Thicker laminate, single
exterior ply drop
ESF [0/0 a/ ^ 45/0/(0/ ^ 45/0)2] 33.3 39.7 4.70/4.06 Thicker laminate, single
interior ply drop
ESG [0 a/0 a/(0/ ^ 45/0)3] 32.7 38.9 4.95/3.81 Thicker laminate, two
exterior ply drops
ESH [0/0 a/0 a/ ^ 45/0/(0/ ^ 45/0)2] 34.5 38.9 4.70/3.56 Thicker laminate, two
interior ply drops
ESJ [0 a/(0/ ^ 45/0)s] 26.7 37.5 3.56/3.05 Single exterior ply drop,
w/‘Z-Spiking’
ESK [0 a/(0/ ^ 45/0)s] 29.3 37.4 3.53/2.90 Single exterior ply drop,
w/Hysol adhes.
a
Indicates ply or plies dropped.

End Notched Flexure Test (ENF) tests [4,5] for these same machine. In the ESA laminate, a single exterior ply was
materials are 0.49 and 2.27 N M 21, respectively. All of the dropped and the initial maximum cyclic stress of 207 MPa
laminates are mixed-mode, with the external ply drop more in the thin section was chosen. The delamination initiated
Mode I dominant, and the interior ply drops more Mode II easily and grew rapidly as can be seen in Fig. 4. It was not
dominant, to be discussed later. until the maximum stress was reduced to 120 MPa (corre-
sponding to approximately 0.5% engineering strain) that the
delamination finally arrested after some growth, compared
4. Fatigue testing to the static strength of ESA of 452 MPa. The next laminate
tested was the ESB laminate which consisted of a single
From the critical loads determined from the static tests, internal ply drop. This laminate schematic can be seen in
the fatigue load of each laminate could be approximated. All Fig. 5 later.
of the tests in this study were run at R ˆ 0.1 (minimum/ From the static testing this laminate had approximately a
maximum stress) on an Instron 8501 servo-hydraulic testing 68.9 MPa higher stress level before delamination initiation

Fig. 2. Stress vs % strain for ESA, ESB, ESC laminates.

526 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534

However, it is noted that the analysis for Gc is based on a

strength of materials approach from Eq. (1). This may not be
adequate for the complicated mixed mode problem. A
nonlinear finite element analysis, as shown in Fig. 9, is
Fig. 3. Model used for Gc calculations. currently being used to investigate this problem. It was
determined that a nonlinear analysis was necessary because
than the ESA laminate. The internal ply has two shear of the membrane/bending coupling in the surrounding plies.
surfaces to resist delamination and the geometry causes a This is illustrated by the deformed mesh of Fig. 9. The
less severe discontinuity. This was accounted for when bridging plies react out the asymmetric loading via bending
calculating the strain energy release rate (Gc) values. Eq. and membrane stresses and the combination is dependent on
(1) was divided by two to account for the two individual the loading. Errors as high as a factor of 2 in Gc calculations
crack fronts, consistent with the strength of materials can result if the coupling is ignored. Fig. 10 is a detailed
approach used in its development. The results for the ESB closeup of the ESH interior ply drop. From Fig. 10, it is
laminate show that the threshold stress for forming delami- expected that, as a consequence of the geometry, mixed
nation (276 MPa) was more than twice as high as for the mode fracture occurs in the resin rich region and the subse-
ESA laminate. Results for the ESC laminate were similar to quent fractured ‘window.’ To account for different mode
the ESB laminate, with only a slight gain in the maximum types and mixes, the virtual crack closure method described
fatigue stress level. The ESC laminate differs from the ESB later was used for analysis [6].
laminate in that the ply drop is in the center of the laminate
rather than immediately under the surface, which reduced
the additional bending stress that results as a consequence of 5. Virtual crack closure method
asymmetry. From the graphs of the ESB and ESC laminates,
delamination resistance in fatigue can be seen in Figs. 6–7, A finite element analysis (FEA) model was constructed to
respectively. examine the initialization, propagation and arrest character-
To compare the different laminates on the same scale a istics of delaminations resulting from plydrops. Three
graph of fatigue delamination growth rate vs normalized models were developed, each corresponding to ESA, ESB
energy release rate is given in Fig. 8. This shows that and ESC samples with various delamination lengths. These
although the strain energy release rate is higher for the models used the material and geometric characteristics of
ESA than for the ESB laminates, when normalized with the glass/polyester coupons listed in Tables 1 and 2. Both
their respective Gc, the crack growth rate for both laminates types of plies in the coupons (D155 and DB120) are
can be correlated with a single relationship. These results modeled as 0.508 mm thick. Each plydrop is dropped across
are extremely encouraging since they imply that the damage a 2.54 mm length. The total model length for all samples is
growth rate for different ply drop configurations may be 91.4 mm long with the plydrop starting at 38.1 mm down
predicted from limited testing on a given material. the length. A two-dimensional plane stress model reduced

Fig. 4. ESA laminate, delamination rate vs cycles.

D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534 527

Fig. 5. Sketch of ESB laminate.

Fig. 6. ESB laminate, delamination rate vs cycles.

Fig. 7. ESC laminate, delamination rate vs cycles.

528 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534

Fig. 8. Fatigue delamination growth rate vs normalized energy release rate.

Fig. 9. FEA model of ESB laminate in tension.

the expense involved with a full three-dimensional version. membrane stresses contribute to the strain energy distribu-
Energy release rates were calculated using the virtual crack tion. As extension of the laminate occurs, the bridging ply
closure method, as shown in Eq. (2). reacts to the asymmetric loading through membrane and
bending stresses. At higher strains, the membrane stresses
1F Du
Gˆ 2† become more important. This is a nonlinear effect.
2Ac
However, the virtual crack closure was found to be linear
Where: G is the strain energy release rate for a given mode, for the cases considered here. In other words, if there are
F is the force required to close the crack, Du is the nodal significant geometric nonlinearities up to the loads of inter-
displacement required to close the crack for the given mode, est, these must be accounted for, but the virtual crack
and Ac is the area of the crack closure. closure can be accomplished by scaling unit loads. By
This method allows modal distribution to be analyzed. solving for Mode I (opening mode) and Mode II (shear or
The method requires the application of virtual loads on sliding mode), the total energy release rate can be found.
nodes closest to the crack front and re-evaluation of the There are currently two schools of thought on how the
model with these loads. As discussed earlier, it was found modal release rates should be added together. One method
that it was necessary to perform a nonlinear analysis for the merely adds them, the second method is the square root of
deformed geometry as shown in Fig. 9 because the the sum of the squares [7,8].
 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534 529

Table 3
Delamination growth rates for laminates studied (mm/cycle) a

Laminate 34.4/344 275/27.5 20.6/206 13.7/137 12.1/121

21 23
ESA 1.3 × 10 3.0 × 10 3.8 × 10 24
22 24
ESB 1.4 × 10 9.1 × 10
ESC 1.1 × 10 23 2.9 × 10 24
ESE 3.0 × 10 22 1.1 × 10 24
ESF 2.1 × 10 23 4.8 × 10 24
ESG 8.6 × 10 24 2.0 × 10 24
ESH 1.4 × 10 22 7.1 × 10 25
ESJ 4.4 × 10 23 4.8 × 10 24
ESK 2.8 × 10 23 1.5 × 10 25
a
Delamination rate (mm/cycle) for Stresses (minimum stress/maximum stress in MPa), R ˆ 0.1.

Using the virtual crack closure method, ESA, ESB, and ply drop. The ESE laminate had an exterior ply drop while
ESC samples were examined for energy release rates from the ESF laminate had an interior ply drop.
crack initiation and crack arrest using experimental data and The ESE laminate had a lower delamination growth rate
loading. Using the accumulated data, it may be possible to than the ESA laminate, which had a smaller overall cross-
predict crack arrest in complex composite structures under sectional area. The Gc for both the ESA and ESE laminates,
fatigue loading. However, it is necessary to first establish a along with the ESF laminate compared with the ESB lami-
relationship between the energy release rates and the load- nate were similar.
ing before reasonable predictions can be made. Using the The ESG and ESH laminate were also thicker in cross-
virtual crack closure method, in Table 3, the percent of section, but, in an attempt to simulate a potential manufac-
Mode I and II responsible for extending the crack is turing situation, two plies were dropped at a single location
shown. In this table, the percentage of each mode for the instead of one ply. The ESG laminate contained two ply
delamination extending into the thicker section (before the drops on the outside while the ESH laminate contained
ply drop) is given, with the percent strain provided for refer- two ply drops on the interior. In both cases, these config-
ence. The analyses were conduced at the experimentally- urations behaved poorly compared with their single ply drop
measured arrest damage sizes. Most of the initiation (small counterparts. During testing of the ESH laminate, the resin
crack lengths), was dominated by Mode I crack growth. The rich area ahead of the ply drop was observed to fragment
crack also extended into the thinner section (after the ply during the fatigue test. When the coupon was in the maxi-
drop), the corresponding percentages are given in Table 4. mum stress part of the fatigue cycle a fracture in the resin
For this purpose, it is expedient to establish a larger data- pocket through the laminates width was seen. No material
base of energy release rates from the virtual crack closure was left ahead of the ply drop after the first initial cycles.
models to compare with experimental data. Of particular This is illustrated in Fig. 10 and described as the ‘window’
note, as the crack grows, the fracture is dominated by in this region.
Mode II strain energy release rates. Hence, initiation The ESG laminate with a double external ply drop dela-
appears to occur in the detailed, resin-rich area in Mode I, minated more easily than a single ply drop on the outside.
and progresses to Mode II domination. One apparent reason for this is the higher critical strain
energy release rate of the ESA laminate, the Gc of the
ESG laminate was only 1.26 N mm 21, while the Gc for the
6. Laminates ESE, ESF, ESG and ESH ESA laminate was 1.89 N mm 21. Another reason for this
was the increased bending moment created by the eccentri-
After the first three laminates were studied, it was decided city of having two layers creating the ply drop, as well as the
that other laminate configurations should be considered. The larger dropped area. A comparison of laminates which
ESE and ESF laminates were thicker laminates with a single

Table 4
Percentage of strain energy release rates for crack growth into thicker
section a

Laminate Percent Mode I Percent Mode II

(opening) (shearing)

ESA(0.46% strain) 40.7 59.3

ESB(1.06% strain) — 100
ESC(1.07% strain) — 100
Fig. 10. Sketch of resin rich region in ESH laminate.
530 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534

Fig. 11. Comparison of internal ply drops at 275 MPa.

contained internal ply drops is presented in Fig. 11. As noted

earlier, the ESH laminate with two interior ply drops had the
highest delamination growth rate. The best configuration
was the ESF laminate, which was a thicker laminate with
a single internal ply drop. This laminate has less bending
stresses and better surrounding material load carrying
capability than the comparable ESB and ESC laminates.

7. Attempt at improving delamination resistance

After the previous configurations were tested, methods of

preventing delamination were studied. The first configura-
tion used the ESA laminate as a basis for comparison and a
second configuration used the ESB laminate as a basis. First,
random mat fabric was included between the ply drop and
the first continuous zero layer. A second modification is
called ‘feathering’. In this case, alternating tows were pulled
approximately 13 mm past the adjacent tows to provide a
less defined ply drop front and delamination site (Fig. 12).
In two other cases, the ESJ and ESK laminates, innova-
tive methods were used to prevent delamination (Fig. 12). In
the ESJ laminate, ‘Z-Spiking’ was used, which consisted of
removing the binder from the fabric and driving the fiber
tows into the lower layers. The transverse strength proper-
ties are much lower than the longitudinal properties, provid-
ing fiber reinforcement in the thickness direction results in
additional resistance to delamination. The ESK laminate
used a 25 mm wide layer of Hysol EA 9309.2NA adhesive
approximately 0.3 mm thick, between the outer ply drop and
the first continuous zero layer. This adhesive was applied
Fig. 12. Illustrations of feathering, ‘Z-Spiking’, an using Hysol adhesive to and was allowed to cure before the resin was injected into
prevent delamination. the mold. The theory behind this attempt was that by
 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534 531

Fig. 13. Comparison of ESA, JKA ‘feathered’, JKA random at 207 MPa.

applying a thin, tough adhesive, delamination initiation laminates were judged. As can be seen from Fig. 13, the
would be resisted. Also the adhesive layer enhances the feathered laminate provided a substantial decrease in the
process zone for fracture [9,10]. It is noted that ‘Z-Spiking’ delamination growth rate. However, applying random mat
and adding local adhesive can also easily be applied in a to the lay-up had little effect and possibly increased the
hand lay-up manufacturing process as well. delamination growth rate. An extensive study was not
performed here, but this result is consistent with other
studies, wherein if the process zone in the interply region
8. Results from delamination prevention techniques is suppressed with reinforcement, the critical strain energy
release rate decreases [9].
Since all of the delamination prevention techniques were In Fig. 14, the ESK laminate with the Hysol adhesive is
applied to the same layer configuration as the ESA laminate, compared with three other laminates. The ESK laminate
the ESA laminate was used as a standard by which the other demonstrated a significant improvement over the other

Fig. 14. Comparison of exterior ply drop configurations at 138 MPa.

532 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534

Fig. 15. ESA vs ESJ ‘Z-Spiked’ laminate.

configurations tested. The delamination in the ESK laminate free-edge effects. The results for the ESJ laminate are illu-
did not initiate until the other laminates had completely strated in Fig. 15. Note that ‘Z-spiking’ has a profound
delaminated. The ESJ laminate with ‘Z-Spiking’ is not influence on the arrest, presumably resulting from the
shown on this graph because there was no delamination extensive fiber bridging noted in these samples. The ESB
initiation in more than 200 000 cycles at the same stress laminate results, with feathered ply drops, are presented in
level. Both ‘Z-spiking’ and the Hysol adhesive delamination Fig. 16. From Fig. 16, it can be clearly seen for the ESB
did not initiate like the other laminates. In both cases, the laminate with feathered ply drops that the delamination
delamination initiated at the specimen edges, as expected started but stopped, while the ESB material with no
from previous studies [11–13]. These edge delaminations feathering continued to grow until arrest at a much larger
occurred even though the edges of the specimens were care- delamination length. In Table 2, the delamination rate for
fully polished. Polishing mitigates, but does not preclude various stress levels and laminates ESA through ESK is

Fig. 16. ESB vs JKB ‘feathered’ at 275 MPa.

 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534 533

Fig. 17. Initial ESA vs repaired ESA at 206 MPa.

presented. In this table, the R ratio (min stress/max stress) 3. Internal ply drops are more resistant to delamination than
value equals 0.1. external ply drops.
4. Repairing external plies using an adhesive is effective in
preventing delamination from starting again.
9. Repairing samples previously tested 5. The use of ‘Z-Spiking’ and adhesives in the initial manu-
facturing of parts with ply drops substantially reduces the
During testing the ESA laminate, which had an exterior initial delamination growth rate.
ply drop, the outer layer peeled away from the rest of the 6. Using feathering at the ply drops reduces the delamina-
laminate. For practical purposes, it was desired to determine tion growth rate.
whether it is possible to repair the delamination with two 7. Random mat possibly has a detrimental effect on dela-
epoxy adhesives. The first adhesive was Hysol EA mination resistance.
9309.2NA, which is blue in color, while the second one
was Hysol EA 9412, which is clear. After the initial test On the basis of these results, an optimum configuration of
run, the adhesive was injected into the delamination via dropping plies may be to have an internal ply drop, with a
syringe. A C-clamp was applied to the specimen until the combination of either ‘Z-Spiking,’ an adhesive or ‘feather-
adhesive cured. These procedures were adopted to simulate ing’ used in the construction.
practical field repairs for composite structures. The speci-
men was then retested at the same stress level and the dela-
mination growth rate was compared to the original rate, 11. Further study
which can be seen in Fig. 17. From Fig. 17, the two adhe-
sives had a similar delamination growth rate and provided a The current analysis for ply drops is based on a strength
significant reduction in delamination growth rate compared of materials approach. it is encouraging that this approach
with the original laminate. may be adequate for damage growth during fatigue as was
shown in Fig. 8. This warrants additional studies. The
strength of materials approach does not include details at
10. Conclusions and design recommendations the crack tip region. Further studies will incorporate more

From the foregoing results, the following general conclu- Table 5

Percentage of strain energy release rates for crack growth into thinner
sions can be drawn about delamination and delamination section a
prevention:
Laminate Percent Mode I Percent Mode II
1. For the same ply drops, thicker laminates are better in (opening) (shearing)
resisting delamination.
2. Dropping more than one ply at the same location ESB (1.06% strain) 73.8 26.2
ESC (1.07% strain) 38.6 61.4
increases the delamination growth rate.
534 D.S. Cairns et al. / Composites: Part B 30 (1999) 523–534

detailed analyses, including the fracture mode type and the graphite/PEEK laminates using cracked lap shear (CLS) specimens.
fracture mode mix to expand on some of the interesting Composite Structures 1994;29:379–382.
[4] Whitney JM, Browning CE, Hoogsteden W. A double cantilever
results of Tables 2–5. This may explain some of the differ- beam test for characterizing mode i delamination in composites. Jour-
ences in Gc for the various laminates. In particular, the nal of Reinforced Plastics and Composites 1982;1:297.
discrepancies between the DCB and ENF tests need to be [5] Carlsson LA, Gillespie Jr. JW. On the design and analysis of end
explored. notched flexure (ENF) specimen for Mode II testing. Journal of
In the future, determining the effect of spacing of ply Composite Materials 1986;20.
[6] Rybicki EF, Kanninen MF. A finite element calculation of stress-
drops should be considered. By doing this the number of intensity factors by a modified crack closure integral. Engineering
plies that can be dropped over a given length can be deter- Fracture mechanics 1977;9:931–938.
mined. If the ply drops are too close together, then a resin [7] Gong Xiao-Jing, Benzeggagh Malk. Mixed mode interlaminar frac-
rich area could possibly form, leading to failure (as in the ture toughness of unidirectional glass/epoxy composites. In: Martin
ESH laminate). However, if the ply drops are too far apart, RH, editor. Composite Materials: Fatigue and Fracture — Fifth
Volume, ASTM STP 1230. Philadelphia, PA: American Society for
then tapering the thickness may be too gradual and there- Testing and Materials, 1995:100–123.
fore, non-optimal. [8] O’Brien TK. Characterization of delamination onset and growth in a
composite laminate. In: Reifsnider KL, editor. Damage in Composite
Materials, ASTM STP 775. Philadelphia, PA: American Society for
Acknowledgements Testing and Materials, 1982.
[9] Cairns DS. Prediction of fracture toughness of multi-phase
This work was supported by the US Department of materials. In: Proceedings of the AIAA/ASME/ASCE/AHS/ASC
Energy and the State of Montana through the Montana 31st Structures, Structural Dynamics and Materials Conference,
April 1990.
DOE EPSCoR Program (contract # DE-FC02- [10] Cairns DS. Mixed-mode fracture and toughening in multi-phase mate-
91ER75681) and NREL (contract # XF-1-11009-5). rials. In: Proceedings of the ASCE Engineering Mechanics Speciality
Conference, 20–22 May 1991.
[11] Wang AS, Crossman FW. Initiation and growth of transverse cracks
References and edge delamination in composite laminates — Part 1: an energy
method. Journal of Composite Materials 1980;14(1):71–87.
[1] Samborsky DD, Mandell JF. Fatigue resistant fiberglass laminates for [12] Chan WS, Wang ASD. Free edge delamination characteristics in S2/
wind turbine blades. In: ASME Wind Symposium, Houston, TX, CE9000 glass/epoxy laminates under static and fatigue loads. In:
1996. Composite Materials Fatigue and Fracture, ASTM STP 1012. Phila-
[2] Ramkumar RL, Whitcomb JD. Characterization of Mode I and mixed- delphia, PA: American Society for Testing and Materials, 1989:270–
mode delamination growth in T300/5208 graphite/epoxy, in delami- 295.
nation and debonding of materials. In: Johnson WS, editor. ASTM [13] Allix O, Gornet L, Ladeveze P, Leveque D, Delamination prediction
STP 876. Philadelphia, PA: American Society for Testing and Mate- by continuum damage mechanics.In: Proceedings of the IUTAM
rials, ASTM, 1985:315–335. Symposium — Variations of Domains and Free-boundary problems,
[3] Rhee KY. Characterization of delamination behavior of unidirectional 22–25 April. Paris, France: Kluwer Academic Publishers, 1997.