International Journal of Fatigue 75 (2015) 5156

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

Fatigue in bamboo
Lauren Keogh, Patrick OHanlon, Peter OReilly, David Taylor
Trinity Centre for Bioengineering, Department of Mechanical & Manufacturing Engineering, Trinity College, University of Dublin, Dublin 2, Ireland

a r t i c l e

i n f o

Article history:
Received 9 September 2014
Received in revised form 28 January 2015
Accepted 2 February 2015
Available online 11 February 2015
Keywords:
Bamboo
Fatigue
Fracture
Anisotropy
Stress concentration

a b s t r a c t
This paper describes the results of an experimental programme to determine the fatigue behaviour of
bamboo. Bamboo is subjected to cyclic loading, both in the plant itself and subsequently when the material is used in load-bearing applications in the construction industry. However, there is currently no data
in the literature describing fatigue in this material. We found that sections of bamboo culm loaded parallel to the culm axis did not undergo fatigue failure: samples either failed on the rst loading cycle, or
not at all. By contrast, fatigue was readily apparent in samples loaded in compression across the diameter
of the culm. The number of cycles to failure increased as the cyclic load range decreased in a manner similar to that found in many engineering materials: fatigue occurred at applied loads as small as 40% of the
ultimate strength. Two different species of bamboo were tested and found to have different ultimate
strengths but similar high-cycle fatigue strengths. Finite element analysis was used to help understand
the progression of fatigue damage and the effect of stress concentration features. Some tentative design
rules are proposed to dene stress levels for the safe use of bamboo, taking fatigue into account.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
Bamboo is an important structural material, grown in many
parts of the world. Being very fast growing it provides a renewable
resource which is extensively employed in construction work and
other load-bearing applications. The literature on this material
contains a number of excellent studies reporting static mechanical
properties such as Youngs modulus, ultimate strength under various different types of loading, and fracture toughness. The bamboo
plant grows as a series of hollow, tubular stems known as culms.
Static mechanical property data have been published for tests carried out on intact lengths of culm [1,2] and also on specimens
machined from the culm wall [38].
In the growing plant, cyclic forces will be experienced, principally in the form of wind loading. When bamboo is used in construction, for example in scaffolding and space frame structures,
cyclic loading can also be expected. So it is surprising that there
is no data in the published literature recording the fatigue behaviour of bamboo. The bres of the bamboo plant, being relatively
stiff and strong, have been used to make bre composites in combination with various matrix materials, and several of these have
been characterised as regards their fatigue properties (e.g. [9]),
but these data give no insight into the fatigue behaviour of bamboo
when used in the form of culms.
Corresponding author. Tel.: +353 1 8961703; fax: +353 1 6795554.
E-mail address: dtaylor@tcd.ie (D. Taylor).
http://dx.doi.org/10.1016/j.ijfatigue.2015.02.003
0142-1123/ 2015 Elsevier Ltd. All rights reserved.

Our primary aim for this work was to determine whether fatigue failure occurs in bamboo and, if it does, to determine the number of cycles to failure for a range of applied cyclic loading
conditions. The material is highly anisotropic, being much stronger
and stiffer when loaded along the culm axis and much weaker
when loaded perpendicular to the axis, so we aimed to investigate
fatigue behaviour in different orientations. Finally, since fatigue
behaviour is strongly affected by the presence of notches and other
stress concentration features, we planned to carry out a limited
number of tests to investigate the response of bamboo to such
features.
2. Methods and materials
Most of the testing was carried out using the species of bamboo
known as Moso (Phyllostachys Pubescens) which is one of the most
commonly used species for structural purposes. The material was
obtained from a local supplier (Bamboo Suppliers of Ireland,
Dublin). A limited number of tests was carried out on a different
species, obtained from the National Botanic Gardens, Dublin, under
the name Dendrocalamus gigantea. All samples were tested in the
air-dried condition in which they would normally be used for
structural applications.
Mechanical testing was carried out using an Instron 8874 servohydraulic machine. Fig. 1 summarises the experimental programme. Samples were obtained in the form of culms several
metres in length. Bamboo culms are hollow tubes in which the

52

L. Keogh et al. / International Journal of Fatigue 75 (2015) 5156

diameter and wall thickness stay fairly constant along the length:
we used culms from plants which were 23 years old, with diameters in the range 3247 mm and wall thicknesses of 2.9
4.4 mm. Culms also have periodic nodes at branch points: we
avoided taking samples from close to the nodes, using only the
internodal material.
Our choice of test specimen was made based on the experience
of ourselves and others in conducting monotonic tests on this
material, and also knowledge of how the material is loaded in
use. Because bamboo is extremely anisotropic (having much
greater strength and stiffness in the longitudinal direction) it
invariably fails during monotonic tests by the initiation and
propagation of longitudinal cracks, irrespective of the specimen
shape and applied load [1,4]. Machining of standard test samples
such as waisted tensile specimens or compact tension specimens,
which has been done previously [6,10] is complicated by the need
to obtain these from the tubular culms. As a result these samples
must either be very small, or they must have a curved shape.
Furthermore, the composition and properties of the material vary
through thickness, the outer layers having more bres and more
hardening by lignication, making them stiffer and stronger [10]
but probably less tough [6] than material closer to the centre. For
these reasons we decided to test samples in the form of simple

tubes obtained by cutting lengths from the culm. These tubes were
loaded in two different ways, as follows.
Axial compression tests were carried out on samples cut to a
length equal to twice their diameter; diametral compression tests
were carried out on samples of length 50 mm. In both cases compressive load was applied through parallel platens. Because fatigue
did not occur in axial compression (see below) most of the tests
were carried out in diametral compression. To reduce the effects
of variability in specimen shape and size we used the following
approach: one sample was loaded monotonically until failure
occurred at a certain force, Fu. Fatigue tests were then carried out
on adjacent samples from the same culm, which had very similar
dimensions. Each of these samples was loaded with a maximum
force Fmax which was a xed percentage of Fu. Cyclic loading was
carried out at a frequency of 1 Hz and load ratio R (equal to Fmin/
Fmax) of 0.1.
Testing was continued until complete failure of the sample:
damage developed in a three-stage process described below. To
investigate the effect of stress concentrations, some samples of
Moso bamboo had 3 mm diameter holes drilled mid-way along
their lengths at one of the locations where initial failure occurred
in the plain samples. Some samples had longitudinal grooves
machined into them at these same locations (see Fig. 1). Monotonic

Fig. 1. (a) Samples were cut from internodal regions of the bamboo culm and tested in the form of cylinders. Compression was applied in the axial and diametral directions:
some diametral specimens contained stress concentration features (holes or grooves). (b) Diametral compression gives rise to high tensile stresses at four locations around
the circumference: the stress on the inside surface at points N and S is approximately twice as high as the stress on the outside at points W and E.

L. Keogh et al. / International Journal of Fatigue 75 (2015) 5156

loading tests were carried out on these samples, as well as fatigue

tests with cyclic loads of 50%, 60% and 70% of the static failure
loads previously established for plain, unnotched samples.
A stress analysis for the diametral loading test was carried out
using ANSYS nite element software. The sample was modelled
as a tube of diameter 40 mm, wall thickness 4 mm and length
50 mm. Steel platens were modelled, through which a xed displacement was applied. Frictional contact was modelled at the platen/sample interfaces. The mesh element size and the number of
iterations were varied until the solution converged. As regards
the mechanical properties of the bamboo material, linear elastic
behaviour was assumed; two different models were created. In
the rst model the material was assumed to have a constant
Youngs modulus and Poissons ratio. Values of E = 1.85 GPa and
m = 0.3 were found to accurately predict the initial part of the
experimental force/deection curve: these results are in agreement
previous ndings [2]. In our second model we attempted to simulate the variation of properties which occurs from outside to inside.
Nogata and Takahashi modelled bamboo as a uniaxial bre composite in order to predict Youngs modulus in the longitudinal
direction [10]. We used their approach, adapting it to study the
transverse direction. We created a model having ve separate layers through-thickness and varied the Youngs modulus of each layer according to the percentage of bres as measured by Nogata and
Takahashi. The Youngs modulus values of the individual layers (in
GPa, from outside to inside) were: 4.01; 2.54; 1.86; 1.61 and 1.27.
We found only small differences between these two FE models: the
multi-layer model gave a slightly higher stress (by 14%) at points N
and S (as dened in Fig. 1b) and a slightly lower stress (by 3%) at
points W and E. The stress results from these models were also
in agreement with analytical solutions [1113] to within 10%.
We also modied our model to investigate the effect of introducing
a stress concentration feature and the changes in stress that occur
during the three stages of damage progression (described below).

53

Fig. 2. Examples of cracks forming at Stage A (above) and Stage B (below), for
diametral compression tests.

3. Results
Axial compression specimens of Moso bamboo failed under
monotonic loading by longitudinal splitting. The average stress to
failure (from 11 samples tested) was 65.0 MPa, with a standard
deviation of 5.8 MPa. It was not possible to generate fatigue failures in these specimens when loaded cyclically, even at a maximum stress as high as 90% of the static strength. Samples that
did not fail monotonically could be cycled for over 100,000 cycles
to the same Fmax without failing.
The diametral compression samples of Moso bamboo, by contrast, displayed fatigue failure over a wide range of loading. Failure
occurred in three separate stages:
Stage A: longitudinal cracking at location W or E (as dened in
Fig. 1b). Cracks initiated on the outside at these points and grew
very rapidly along the entire length of the sample. These cracks
did not penetrate to the inner surface but arrested when they
had passed through approximately three quarters of the thickness
(see Fig. 2a).
Stage B: because the crack in Stage A did not cause complete
failure, the test was continued. At a later stage a crack would form
at the opposite side (point E or W respectively), developing in a
similar manner (Fig. 2b). This also did not cause complete failure.
Stage C: Final failure occurred when a crack formed on the inner
surface of the sample close to one of the loading points (point N or
S), causing complete collapse of the sample.
These three stages also occurred during monotonic loading,
where they could be distinguished by load drops giving rise to
three peaks on the force/displacement curves (Fig. 3). There was
no consistent order to the heights of these peaks, but typically they

Fig. 3. A typical force/displacement curve obtained from monotonic loading, with

maxima occurring at Stages A, B and C.

did not vary one from another by more than 20%. The average value
of the tensile stress at points W and E when monotonic failure
occurred (dened as the highest of these three peaks) was
10.4 MPa.
Fig. 4 shows fatigue data for Moso bamboo, plotting the number
of cycles to failure as a function of the maximum cyclic force, normalised by the static failure force. Failure here is dened as complete collapse of the sample (i.e. Stage C). There is a signicant
amount of scatter but it is clear that fatigue occurs at applied forces
which are considerably less than the static strength. Fig. 5 shows
results for diametral compression testing of the other bamboo species, which also showed clear evidence of fatigue failure, though
here there is less data available: further testing would be required
to completely describe the fatigue characteristics.
In Fig. 6 the plain-specimen data for Moso bamboo (from Fig. 4)
is reproduced along with data from the tests in which stress concentrations were introduced into the samples. The introduction
of holes and grooves tended to reduce both the static strength
and the fatigue strength, but there was considerable overlap with
the plain-specimen data.

54

L. Keogh et al. / International Journal of Fatigue 75 (2015) 5156

4. Discussion

Fig. 4. Fatigue data for Moso bamboo tested in diametral loading. The number of
cycles to failure is plotted against the maximum load in the cycle, normalised by the
load to cause static failure in a sample taken from the same region of the culm.

Fig. 5. Fatigue data for a bamboo species obtained from the National Botanic
Gardens, Dublin, under the name Dendrocalamus gigantea.

Fig. 6. Fatigue data for Moso bamboo samples containing stress concentration
features. The plain-specimen data (from Fig. 4) is reproduced for comparison.

This work has succeeded in its primary aim: we have demonstrated that fatigue does occur in bamboo. Our nding that fatigue
failures happen at applied cyclic stresses as low as 40% of static
strength, at cycle numbers of the order of 105, shows that the fatigue behaviour of bamboo is similar to that of many engineering
materials. Fatigue is a phenomenon that should be taken into
account when using bamboo for any load-bearing application
where cyclic loads might arise. Our data show considerable scatter,
the number of cycles to failure varying by more than four orders of
magnitude for the same loading conditions. This is not unusual in
brittle materials, for which fatigue life tends to be controlled by
crack initiation from microscopic defects [14]. In this case additional scatter can be expected due to the natural variations in
material and geometry, though we have attempted to reduce the
effects of geometry as much as possible by testing paired samples
as described above.
Our nding that fatigue occurs readily in diametral compressive
loading but not in axial compressive loading is consistent with previous work on bre composite materials [15]. Unidirectional bre
composites loaded parallel to their bres tend not to show a large
fatigue range because the applied stress is borne largely by the
bres, which display very brittle behaviour. In contrast, loading
perpendicular to the bres leads to fatigue due to the properties
of the matrix material or the bre/matrix interface.
In use, bamboo culms performing structural functions such as
scaffolding or space-frame structures will experience complex
loading regimes. Whilst they will only rarely be loaded in exactly
the same way as our diametral specimens, they will commonly
experience bending, which causes ovalisation of the cross section
and gives rise to a stress distribution which is very similar to one
created in our tests [13]. Thus our fatigue data should be relevant
for the engineering application of bamboo.
Fig. 7 shows nite element analysis results for the diametral
compression specimens. Stresses arise mainly due to bending of
the tube walls. At points W and E there are regions of locally-high
tensile stress on the outside and compressive stress on the inside.
At points N and S the pattern is reversed, with tension occurring
on the inside. In our tests, whether monotonic or cyclic, the initial
failures always occurred at the W and E positions, with the N and
S locations failing later on. This is surprising because the tensile
stress is approximately twice as high at N and S compared to W
and E. The average tensile stress at locations W and E for monotonic
failure was 10.4 MPa: for the same applied load the stress at N and S
is 19.1 MPa.
FEA showed that cracking in one location tends to increase the
stress at other locations (see Fig. 7). In Stage A, the presence of a
crack at point E was estimated to increase the stress at point W
by 17.6% and that at points N and S by 13.7%. In Stage B, the presence of cracks at both E and W was found to increase the stress at
points N and S by 31.6% (compared to the original values for the
undamaged sample). These results help to understand how fatigue
proceeds in these tests. Initial cracking (Stage A) will occur at one
of two locations, W or E. Due to slight differences in material and
geometry, one side will be weaker than the other and so will fail
rst, causing the stress on the stronger side to increase, but only
by a relatively small amount (17.6%). This explains why in most
cases the Stage A failure did not immediately lead to total failure:
rather, a signicant number of cycles elapsed between the rst and
second failures. The onset of Stage B causes a relatively large
increase in the stresses at points N and S (31%) which were in fact
already much higher that the stresses at W and E as noted above.
Nevertheless further cycles were needed to generate failure at N
or S and thus precipitate a total collapse. A possible explanation

L. Keogh et al. / International Journal of Fatigue 75 (2015) 5156

55

Fig. 7. Results from FEA: (a) maximum principal stress contours for a plain sample; (b) likewise for a sample containing a groove; (c) stress plotted as a function of distance
around the outer circumference starting and nishing at point E, for an intact specimen and also for a groove (3 mm deep notch) at point E.

lies in the variation in material structure through-thickness, which

was mentioned above. The material on the outside of the culm is
relatively hard and strong but it appears to have a lower toughness
than material on the inside [6]. So fatigue cracks may propagate
more easily from small defects at points W and E than at points
N and S.
We conducted a relatively small number of tests to investigate
stress concentration features, so these results should be treated
with caution until further data are available. When we introduced
a longitudinal groove at the E location, this had the effect of
eliminating Stage A. As a result the force needed for a given number of cycles decreased (on average) by 21%. FEA predicts this
change quite well: the groove essentially creates Stage A at the
start of the test, and increases the stress to cause Stage B by
17.6%.
When we introduced a circular hole at E, the effect was very
variable. In some samples a crack propagated very quickly from
the hole, and on average the number of cycles needed for Stage A
did reduce (by a factor of 5.5). But on average the total number
of cycles to failure did not decrease, though this was largely due
to the fact that 3 of the 17 samples endured very large numbers
of cycles.
In a plain specimen, the initial, Stage A failure, may occur at
point E or point W and (as mentioned above) it will choose the
weaker side. The hole is introduced into one side only, which
may be the stronger side or the weaker side. If its the stronger side,
then this will have a major effect on the overall life, but if its the
weaker side it will have a negligible effect, since both sides must
eventually fail. This will tend to increase the scatter in the number
of cycles to failure. To use a simple analogy, if one throws a die
then the probably of getting any number from 1 to 6 is equal.
But if one throws two dice and chooses the higher of the two values, then the result is weighted towards the higher numbers, and
scatter is reduced. A similar argument can be made for the groove
feature: in this case our data did not show increased scatter but
this may be because the number of samples tested was small.
Tests carried out on the other bamboo species showed that fatigue failure also occurred, though the dependence of fatigue life on

applied load was somewhat atter. The static strength was smaller
for this material, at 7.7 MPa, however the high-cycle fatigue
strengths of the two materials were similar.
The overall picture is complicated by the existence of a multistage failure process in which the culm can endure considerable
amounts of damage before nal failure occurs. High cycle fatigue
failure may be quantied as follows: total failure will occur after
100,000 cycles at R = 0.1 when the stress range (i.e. the difference
between the maximum and minimum stress in the cycle) at points
W and E is approximately 4.5 MPa. For this same loading, the stress
at points N and S rises to approximately 11 MPa in the nal stage
before complete collapse. Further work is needed to characterise
this material more comprehensively, but these initial results may
prove useful in indicating a safe working stress for bamboo culms
subjected to cyclic loading.

5. Conclusions
1. Bamboo culm samples display fatigue failure when loaded in
compression across their diameters, simulating a type of deformation which commonly occurs as a result of culm bending.
2. By contrast, no fatigue behaviour occurs when samples are
loaded in axial compression.
3. Under diametral loading, failure develops in a series of stages
during which cracks initiate and grow in different locations
around the circumference. Failure in the rst cycle (i.e. static
failure) is associated with a tensile stress of 10.4 MPa on the
outer surface and 19.1 MPa on the inner surface. High cycle failure at 100,000 cycles requires a stress range of 4.5 MPa on the
outer surface and 11 MPa on the inner surface.
4. The effect of stress concentration features such as grooves and
holes is complicated owing to the multi-stage failure process.
Initial results suggest that the main effect of such features at
points W and E is to increase scatter in the number of cycles
to failure. More work is needed to systematically study the
effect of feature size, shape and location.

56

L. Keogh et al. / International Journal of Fatigue 75 (2015) 5156

Acknowledgements
We are grateful to Martin McNulty of Bamboo Suppliers of
Ireland and to Brendan Sayers of the National Botanic Gardens,
Glasnevin, Dublin, for supplies of bamboo culms and helpful
advice.
References
[1] Chung KF, Yu WK. Mechanical properties of structural bamboo for bamboo
scaffoldings. Eng Struct 2002;24(4):42942.
[2] Torres LA, Ghavami K, Garcia JJ. A transversely isotropic law for the
determination of the circumferential Youngs modulus of bamboo with
diametric compression tests. Latin Am Appl Res 2007;37(4):25560.
[3] Amada S, Untao S. Fracture properties of bamboo. Compos B Eng
2001;32:4519.
[4] Shao ZP, Zhou L, Liu YM, Wu ZM, Arnaud C. Differences in structure and
strength between internode and node sections of Moso bamboo. J Trop Forest
Sci 2010;22(2):1338.
[5] Lo TY, Cui HZ, Leung HC. The effect of ber density on strength capacity of
bamboo. Mater Lett 2004;58(21):25958.

[6] Tan T, Rahbar N, Allameh SM, Kwoe S, Dissmore D, Ghavami K, et al.

Mechanical properties of functionally graded hierarchical bamboo structures.
Acta Biomater 2011;7(10):3796803.
[7] Lakkad SC, Patel JM. Mechanical properties of bamboo, a natural composite.
Fiber Sci Technol 1981;14(4):31922.
[8] Jain S, Kumar R, Jindal UC. Mechanical behaviour of bamboo and bamboo
composite. J Mater Sci 1992;27(17):4598604.
[9] Phong NT, Gabr MH, Okubo K, Chuong B, Fujii T. Enhancement of mechanical
properties of carbon fabric/epoxy composites using micro/nano-sized bamboo
brils. Mater Des 2013;47:62432.
[10] Nogata F, Takahashi H. Intelligent functionally graded material: bamboo.
Compos Eng 1995;5(7):74351.
[11] Young WC, Budynas RG. Roarks formulas for stressstrain. 7th ed. Boston,
USA: McGraw-Hill; 2002.
[12] Batista M, Usenik J. Technical note: stresses in a circular ring under two forces
acting along a diameter. J Strain Anal Eng Des 1996;31(1):758.
[13] Wegst UGK, Ashby MF. The structural efciency of orthotropic stalks, stems
and tubes. J Mater Sci 2007;42:900514.
[14] Taylor D. Fatigue of bone and bones: an analysis based on stressed volume. J
Orthop Res 1998;16(2):1639.
[15] Hertzberg RW, Manson JA. Fatigue of engineering plastics. London: Academic
Press; 1980.