Holz als Roh- und Werkstoff (2005) 63: 1–10

DOI 10.1007/s00107-004-0533-7

O R I G I N A L A R B E I T E N Æ O R IG I NA L S

J. W. G. van de Kuilen Æ H. J. Blass

Mechanical properties of azobé (Lophira alata)

Published online: 23 December 2004

Springer-Verlag 2004

Abstract The use of tropical hardwoods in structural dieser Arbeit wurden Biege-, Druck- und Scherversuche
applications requires knowledge of the mechanical mit Prüfkörpern in Bauholzabmessungen ausgewertet
properties. Even though long experience is available und die entsprechenden charakteristischen Steifigkeits-
with the species azobé (Lophira alata), much of the und Festigkeitskennwerte bestimmt. Die Ergebnisse ze-
current knowledge is based on tests performed on small igen, dass Bongossi bessere mechanische Eigenschaften
clear specimens for bending, compression and shear. aufweist als bisher angenommen. Die Beziehung zwis-
Since tropical hardwoods are used in structures with chen statischem Elastizitätsmodul und Biegefestigkeit
high economic value, a sound knowledge of the wurde bestimmt. Der Höheneinfluss von Bongossi zeigte
mechanical properties allows for better optimisation and sich weniger ausgeprägt als für Nadelholz.
for minimal use of raw material. Bending, compression
and shear tests have been performed on material in
structural sizes. Characteristic values show that in most 1 Introduction
loading cases the structural capacity of the material is
distinctly higher than currently assumed in practice. The Tropical hardwoods are used in structures where both
relationship between static modulus of elasticity and sound mechanical performance and good natural dura-
bending strength has been determined. The depth effect bility are required. Hardwoods are used in bridges, sheet
was shown to be less pronounced than for softwoods. pile walls, sluice doors, heavy piles and harbour works.
One of the most important hardwoods used in the
Festigkeitskennwerte von Bongossi (Lophira alata) Netherlands for these applications is the species azobé
(Lophira alata), also known as ekki. This species is
Zusammenfassung Der Einsatz von Konstruktionsholz widely used not only in the Netherlands, but also in
aus tropischen Hölzern setzt die Kenntnis der mecha- Belgium, the UK, and Germany. Nowadays, structures
nischen Eigenschaften dieser Hölzer voraus. Obwohl made with azobé are also exported from the Netherlands
Bongossi seit vielen Jahren als Bauholz vor allem in to other countries such as the USA and Japan.
bewitterten Konstruktionen verwendet wird, wurden Undoubtedly, an accurate knowledge of the mechanical
mechanische Kennwerte für Biege-, Druck- und Scher- properties of the species is of key importance for eco-
beanspruchung vor allem aus der Prüfung kleiner feh- nomic designs. The species was introduced in the
lerfreier Prüfkörper abgeleitet. Da tropische Hölzer Netherlands just after the Second World War, when the
überwiegend in ökonomisch bedeutenden Konstruktio- Dutch Ministry of Road and Waterworks needed a
nen eingesetzt werden, erlaubt eine bessere Kenntnis der durable and reliable alternative for many sluice doors,
mechanischen Eigenschaften optimierte Konstruktionen which were traditionally made from oak, treated with
sowie Einsparungen insbesondere beim Rohstoff. In creosote or a similar organic compound. The alternative
was found in azobé. Azobé has a dark red/brown colour,
interlock grain and is found in the tropics of West
J. W. G. van de Kuilen (&) Africa. The durability of azobé is well known, and the
Faculty of Civil Engineering and Geosciences, material has proven its performance in many structures
Delft University of Technology, 5048,
2628 CN Delft, The Netherlands that were built during the last decades (Wood Hand-
E-mail: [email protected] book 1999). With the introduction of new design codes,
H. J. Blass
such as the Eurocode 5 (2002), that are based on char-
Lehrstuhl für Ingenieurholzbau und Baukonstruktionen, acteristic values, the need for a thorough investigation of
Universität Karlsruhe (TH), 76128 Karlsruhe, Germany the mechanical properties of this species arose. Many of
2

the design rules in the new design codes have been ver- O (old). In total, seven new and three old samples are
ified for softwoods, but hardwoods have a different analysed. All data are also analysed as being part of one
structure and may show different strength behaviour. population. Furthermore, one sample for compression
Azobé is assigned to strength class D60 of EN 338 (EN (C) and one sample for shear (S) are analysed.
1912: 1998). One of such behaviours is the volume effect
(Weibull 1939). The failure mechanism of softwoods is
governed by the amount and location of growth defects 2.1 Bending tests
such as knots. Knots are virtually non-existent in a
tropical hardwood such as azobé, and the failure mode is 2.1.1 Sample N1
generally based on grain deviation. Many studies have
been carried out regarding the volume effect in timber, The timber dimensions of the specimens were 50·150
but almost all of them focussed on softwoods (Barrett and 3,300 mm in length. The beams were tested edgewise
and Griffin 1989; Barrett and Fewell 1990). Currently, in four-point bending with a span of 3,000 mm with the
the volume effect for bending and tension is dealt with in loads at 1/3 of the span. Span to depth ratio was 20. In
European standard EN 384 and Eurocode 5 (2002) total, 46 beams were tested, all having moisture contents
according to the following equation and based on a above the fibre saturation point. The moisture content
reference depth of 150 mm: varied between 35 and 52%. The density was based on

 green volume and oven-dry mass. The density of azobé
150 0:2 determined, according to this method can be adjusted to
kh ¼ ð1Þ 12% moisture content by adding 200 kg/m3 according
h
to (Houtinstituut 1961).
The equation is mainly determined on the basis of
test results of softwoods. Bohannan (1966) found an
exponent of 0.11 for clear wood bending strength, 2.1.2 Sample N2
based on a reference depth of 200 mm, while Fewell The second sample consisted of 40 sheet pile boards with
and Curry (1983) found 0.4 on the basis of Canadian a depth of 30 mm. All boards were tested in four-point
and British structural softwoods, also for a reference bending and in wet condition. The test span was 750 mm
depth of 200 mm. The coefficient of variation in and consequently the span to depth ratio was of 25. The
hardwoods is generally between 15 and 20%, whereas width of the boards varied between 150 and 276 mm.
in softwoods a coefficient of variation above 20% is The static modulus of elasticity and the bending strength
common, even if the material is machine strength gra- were determined (Houtinstituut 1977).
ded (Frühwald and Thelandersson 2001). In addition to
the new design codes, new strength-grading methods
have become available, such as grading using stress 2.1.3 Sample N3
waves or ultrasound. However, the applicability of
these methods is limited to softwoods (mainly spruce This sample consisted of 40 sheet pile boards with a
and pine); grading of tropical hardwoods is performed depth of 60 mm and in this case, the span was
to a much lesser extent. Therefore, the analysis also 1,500 mm, giving the same span to depth ratio as that of
included the relation between modulus of elasticity sample 2. The widths of these boards varied between 179
(static or dynamic) and bending strength in order to and 315 mm and the moisture content was between 25
establish an appropriate method to grade tropical and 43%. In this case, the oven-dry density was also
hardwoods and azobé in particular. Besides bending determined, besides the static modulus of elasticity and
strength and stiffness, compression strength and shear the bending strength (Houtinstituut 1977). The oven-dry
strength are also analysed regarding their relevance in density was adjusted to 12% moisture content similarly
the design of columns (compression) and short-span to Sample N1.
bridge decks (shear).
2.1.4 Sample N4

2 Materials and method Thirty-nine sheet pile wall boards with a depth of
60 mm, a width of 300 mm and a length of 5,700 mm
Over a period of more than 40 years, tests have been were tested flatwise in four-point bending to determine
performed with azobé. Generally, the material was ob- the relationship between wet density, mass, dynamic
tained from sawmills, but in some cases material from modulus of elasticity and static modulus of elasticity
building sites was also obtained where the material had with the strength. Another 50 boards were also tested for
been in service for a number of years. In the following static and dynamic modulus of elasticity, but not for
following section, the origin of the data and the tests and strength, since these had to be used for full-scale testing
measurements performed are given. Material that was of sheet pile walls. The boards were obtained from five
directly obtained from the sawmills is indicated with N different suppliers in the Netherlands. The test span was
(New) and material from building sites is indicated with 5,400 mm with the loads at 1/3 of the span. Due to the
 3

high span to depth ratio of 90, a special test set-up was 2.1.10 Sample P1
built to allow for very large deflections and rotations at
the supports and the loading points. All the boards had The population sample P1 comprises the data available
a trapezoidal tongue and groove profile. There were no from all the bending samples. The mechanical properties,
indications that this influenced the strength determined mainly bending strength and modulus of elasticity, are
in any way (Van der Linden et al. 1996). The moisture analysed statistically. The total number of specimens for
content of the boards varied between 17 and 50%. bending strength and modulus of elasticity is 258, while
for the density and the dynamic modulus of elasticity only
the data from samples N4, N5 and N6 are used.
2.1.5 Sample N5

Sixteen beams with a depth of 150 mm and a width of

75 mm were tested in four-point bending. The beams 2.2 Compression tests
were delivered directly from two sawmills and all had
moisture contents above the fibre saturation point. Dy- Compression tests (Sample C1) were performed using
namic modulus of elasticity was determined based on specimens taken from sample N1. Specimens were
determination of eigenfrequency using a Grindo-sonic manufactured with a cross section of 50·150 mm and a
machine and with TNO-Wave equipment. length of 250 mm. In total, 43 specimens were tested in
compression. The moisture content was determined and
was above the fibre saturation point in all cases.
2.1.6 Sample N6

This sample contained 25 beams with a depth of 200 mm 2.3 Shear tests
and a width of 75 mm. This sample is comparable to
sample N5 and was meant to study a possible depth Shear tests (Sample S1) are particularly difficult to per-
effect. Three sawmills delivered the material, all of which form, since whatever type of specimen is chosen, the
was tested in wet condition. Of these 25 beams five introduction of secondary stresses is inevitable. Standard
beams, five beams were rejected due to too large grain test specimens as specified for instance in ASTM D143
deviations. Beams with a grain deviation of more than (1995) determine the strength of small clear specimens
1:10 are not allowed according to the visual grading (Rammer and McLean 1996). The results are difficult to
rules for azobé in the Netherlands (NEN 5480). interpret for structural applications. European standard
EN 1193 specifies a method with bonded steel plates on
2.1.7 Sample O1 the surface. This method is applicable to species easily
adhering to steel, but for a species like azobé it is not
From the harbour of Rotterdam, old mooring posts possible. Therefore, shear tests have been performed on I-
were obtained. From the original sizes of 120·120 mm, shaped beams in a five-point bending test. Although in
20 boards were manufactured with a depth of 55 mm this case bending stresses parallel to the grain develop as
and a width of 118 mm. Five of these boards showed well as compression stresses perpendicular to the grain,
large defects and were excluded. The boards were tested this method does give reliable results that can be used in
in wet condition in four-point bending with 1,500 mm structural applications (Leicester and Breitinger 1992),
span, resulting in a span to depth ratio of 27. Schickhofer (2001). Twenty-five beams were shaped into
an I beam as shown in Fig. 1.
2.1.8 Sample O2 The test set-up is shown in Fig. 2. The span/depth
ratio is 948/110=8.6 compared to 20 or more for all the
This sample consisted of 20 boards from a sheet pile wall bending samples. The sizes of the supports and the
located in Alphen a/d Rijn. The depth of the boards loading points were 120 mm in length. However, the
varied between 50 and 60 mm, while the width varied data analysis has been performed on the assumptions of
between 135 and 260 mm. All boards were tested in point loads and point supports, in order to be able to use
saturated condition. The boards were sawn from sheet the values for practical design situations.
piles that had been in service for 15 years.

2.1.9 Sample O3 2.4 Statistical data analysis

This sample consisted of 22 boards from a sheet pile wall 2.4.1 Normal, lognormal and Weibull distributions
located in Koudekerk a/d Rijn. The depth of the boards
was 50 mm in all cases, while the width varied between The test data have been analysed statistically. Average
153 and 283 mm. The span in the test set-up was again values, standard deviations and minimum and maxi-
1,500 mm and consequently the span to depth ratio was mum values were determined for each sample and
30 in this case. The boards were sawn from fully satu- for the population sample P1, respectively. Sample
rated sheet piles that had been in service for 18 years. P1 was analysed to verify whether the properties
4

12 intervals of 10 N/mm2 was determined starting at

40 N/mm2. The properties (average l and standard
deviation r) for the lognormal distribution are deter-
mined as
P sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P P ffi
lnðxi Þ n lnðxi Þ2
ð lnðxi ÞÞ2
l¼ and r ¼
n nðn
1Þ
ð2Þ
The three-parameter Weibull distribution is written
as
k
P ðf Þ ¼ 1
eð
ðxi
x0 Þ=mÞ ð3Þ
where P(f) is the probability of failure, m the scale
Fig. 1 Cross section of shear specimens (in mm) parameter, k the shape parameter and x0 the location
Abb. 1 Querschnitt der Scherversuchskörper (in mm)
parameter.
The location parameter x0 may be chosen to be zero.
In that case, the three-parameter Weibull distribution
reduces to a two-parameter distribution. Both two- and
three-parameter Weibull distributions have been applied
to describe the test results. The values of the three-
parameter Weibull distribution, have been determined
using the freely available software Dataplot (NIST
2001). For the three-parameter Weibull distribution
Dataplot takes the lowest value of the test data as the
Fig. 2 Five-point bending test with two spans of 948 mm location parameter.
Abb. 2 Fünf-Punkt-Biegeprüfung mit zwei Spannweiten von je 948
mm

2.4.2 Least-square method for three-parameter

can best be described by a normal, lognormal or a Weibull distribution
Weibull distribution. Benjamin and Cornell (1971) sug-
gest using nine intervals for the sample size concerned, A simple least-square method (lsm) was developed to
but this would lead to more complex boundaries for determine the influence of the location parameter x0 on
each interval. In this case, a histogram of the data with the value of the scale and shape parameters m and k,

Table 1 Overview of test samples and test results

Tabelle 1 Übersicht über die Stichproben und Versuchsergebnisse

Sample No. Width Depth Span Ratio Moisture Density (kg/m3) Static MoE
(mm) (mm) (mm) S/D (%)
Avg. Remarks SD Cov. Avg. SD

N1 46 50 150 3,000 20 35–52 830 Oven-dry 21 2.5 16,700 2,190

N2 40 150–276 30 750 25 >30 – – – – 16,450 2,330
N3 40 179–315 60 1,500 25 25–43 880 Oven-dry 23 2.6 18,490 2,470
N4 39 300 60 5,400 90 17–50 1,120 12% 12 11 15,660 1,720
N5 16 75 150 3,000 20 >30 1,110 12% 25 2.5 17,620 2,330
N6 20 75 200 4,000 20 >30 1,160 12% 23 2.0 16,290 1,670
O1 15 118 55 1,500 27 >30 – – – – 19,030 3,060
O2 20 135–260 50–60 1,500 30–25 >30 – – – – 18,490 2,420
O3 22 153–283 50 1,500 30 >30 – – – – 17,380 2,370
P1 258 – – – – – 1,120 12% 59.5 5.3 17,200 2,510
(based on N4, N5, N6)
Thickness Width Length
(mm) (mm) (mm)
C1 43 50 150 250 35–51 – – – – – –
Sizes (mm)
S1 25 See Fig. 1 8.6 >30 12% – – 13,670 1,330
a
Corrected values for mean and standard deviation: Eqs. 6 and 7
 5

respectively. The procedure to determine x0, k and m is

as follows. First, the cumulative probability function
was determined based on the histogram of Fig. 3. Sec-
ondly, the value of x0 was varied from 0 to 50 N/mm2,
the latter being close to the minimum strength value
found in the tests. For each value of x0, the least sum of
squares between the test data and the Weibull distribu-
tion parameters k and m was determined where k and m
were optimised using the Solver routine in Excel. Then,
the minimum value of the least sum of squares was taken
as the optimal value for x0, k and m. In this way, it is
ascertained that data in the lower and upper tail of the
distribution are weighted equally as data in the centre,
even though in the centre much more test data are
available. This is an alternative for tail fitting, which Fig. 3 Distribution of bending test results (Sample P1)
Abb. 3 Verteilung der Biegefestigkeit von Bongossi
requires large datasets (Ranta-Maunus 2001).

2.4.3 Determination method of characteristic the Student distribution (Confidence level=0.75) and n
values of samples is the number of tests.
The value of tvd depends on the degrees of freedom v
European standard EN 384 specifies a method for the (=n
1). This means that the fewer test results available,
derivation of characteristic values for samples with at the larger the value of tvd results. For materials where the
least 40 test results (Fewell and Glos 1988) . Because not standard deviation of the population is known a priori,
all samples fulfil this requirement, characteristic values the degree of freedom may be based on n=¥. In this
of the individual samples as well as for the population case, there are nine samples available from many dif-
sample have been determined using the Bayesian method ferent sources, and it may be assumed that the standard
given in ISO 2394 (1998). According to this method, the deviation of the whole population is known from Sam-
fifth-percentile characteristic value of a sample Rk can be ple P1. As a result, tvd is taken as 1.64 to determine a
determined as: samples’ fifth-percentile strength value. The character-
s
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 istic value of Sample P1 has also been determined as the
1 fifth-percentile value of the fitted distributions.
Rk ¼ mR
tvd sR 1þ ð4Þ For the shear tests, the procedure for the determi-
n
nation of the characteristic value is more complicated. In
where mR is the mean value of the sample, sR the sample each five-point bending test, two shear tests are per-
standard deviation of the sample, tvd the coefficient of formed, one in each span. Consequently, every time a

(N/mm2) Dynamic MoE (N/mm2) Bending strength (N/mm2)

Cov. Min. Max. Avg. SD Cov. Min. Max. Avg. SD Cov. Min. Max. Rk

13.1 12,300 21,800 – – – – – 93.8 14.8 15.8 66.2 131.9 69.3

14.2 12,570 23,730 – – – – – 104.8 13.3 12.7 78.8 147.9 82.7
13.4 11,350 22,140 – – – – – 106.1 18.2 17.2 55.5 126.9 75.9
11.0 12,160 21,600 17,250 1,640 9.5 13,870 22,460 78.5 12.2 15.6 49.5 105.4 58.2
13 12,760 21,020 18,800 2,250 12.0 15,160 21,910 79.1 14.4 18.1 49.6 105.3 54.8
10 11,950 20,270 18,020 1,490 8.3 14,800 20,460 78.6 7.5 9.6 58.8 92.5 66.0
16.1 14,770 24,140 – – – – – 109.6 11.9 10.9 91.6 129.2 89.4
13.1 15,420 21,660 – – – – – 109.4 10.3 9.4 91.5 127.2 92.1
13.6 13,190 22,910 – – – – – 106.0 16.5 15.6 82.1 140.0 78.3
15 11,350 24140 17,730 1,850 10 13,870 22,460 96.1 18.9 19.7 49.5 147.1 65.0
(based on N4, N5, N6)
Compression strength (N/mm2)

– – – – – – – – 69.7 6.5 9.3 56.4 84.3 58.9

Shear strength (N/mm2)
9.8 10,860 15,630 – – – – – 17.7a 2.1a 12.0 13.4 20.2 14.2
6

beam fails, the lowest value of the ‘two’ tests is found

and the result has to be adjusted statistically. Van
Douwen et al. (1958) have shown that for this case the
following relationships apply:
x1 ¼ x2 þ c1 r 2 ð6Þ

r1 ¼ c 2 r2 ð7Þ
where x1 and x2 are the average value of the whole and
the tested population, c1 and c2 correction factors for the
mean and standard deviation r1 and r2 are the standard
deviations of the whole and the tested population.
For symmetrically loaded specimens as in the five-
point bending tests, and assuming that the results can be
described by a normal distribution, the correction fac-
tors are c1=0.68 and c2=1.21. With these results, the
characteristic shear strength has been determined using
Eq. 4. The modulus of elasticity of each beam has been
determined in four-point bending by applying a load
while the middle support was removed. The density of
each beam was also determined.

3 Results and discussion

Table 1 gives the average values, standard deviation and

coefficient of variation and minimum and maximum
value of all samples tested including the population P1
sample. The latter is done to obtain information on the
whole population and to be better able to derive (char-
acteristic) values for the species and relationships
between properties, such as the modulus of elasticity
and strength. In addition, it becomes possible to deter-
mine, which statistical distribution function fits the data
best.

3.1 Bending strength

The distribution of the bending strength (sample P1) is

shown in Fig. 3. The average strength of the sample
of 258 specimens is determined as 96.1 N/mm2 and
the standard deviation as 18.9 N/mm2 (cov.=0.20). Fig. 4 Probability plots for the bending strength; a Normal
Figure 4a–c give probability plots of the test results, probability (N/mm2); b Lognormal probability [log(N/mm2)]; c
Weibull probability [log(N/mm2)]. Thin lines represent the distri-
based on the assumption of a normal, lognormal or butions for which the parameters are given in Table 2
Weibull distribution, respectively. From the three dis- Abb. 4 Summenhäufigkeit der Biegefestigkeit; a Normalverteilung
tributions, the normal distribution seems to best describe (log(N/mm2)); b Lognomalverteilung (log(N/mm2)); c Weibullver-
the test data over the full range, while the lognormal teilung (log(N/mm2)). Die Verteilungsparameter sind in Tabelle 2
deviates on both lower and upper tail and the Weibull angegeben
deviates around the lower tail only. The cumulative
density functions of the normal and the three-parameter distribution determined with Dataplot takes the lowest
Weibull distributions are shown in Fig. 5. The parame- value of the test data (49.5 N/mm2) as the value for the
ters of all the distributions are summarized in Table 2. location parameter. However, the three-parameter
The non-parametric value of the characteristic Weibull distribution determined on the basis of the
strength is 66.0 N/mm2. The lower fifth-percentile value histogram gives a location parameter that is far lower,
is 65 N/mm2 according to the method given in ISO 2394. namely 18.33 N/mm2. The result is that in the first case a
Therefore, the differences between the distributions are slight overestimation of the fifth-percentile value is ob-
relatively small and no preference for either one distri- tained, while in the second case it is the opposite, but the
bution can be given. The three-parameter Weibull differences remain small.
 7


150 0:103
kh ¼ ð8Þ
h

For practical applications a value of 0.1 can be used.

The characteristic values of the individual samples are
also shown in Fig. 6 and although a slight increase in
characteristic values for smaller depths can be observed,
it is small and could be neglected. The depth effect differs
from softwoods due to failure by grain deviation and the
absence of knots. However, the procedure for the
assignment of species and grades to a strength class re-
quires the application of the depth effect according to
Eq. 1, and this has a negative influence on the design
strength values for ultimate limit states of beams with a
depth of 150 mm and more.

3.3 Modulus of elasticity

For the static modulus of elasticity, the average value

and standard deviation are 17,200 and 2,500 N/mm2,
Fig. 5 Cdf for the bending strength of azobé and three statistical
respectively (cov.=0.15). In Fig. 7, the probability plots
distributions. Data of Lognormal and two-parameter Weibull are of the static modulus of elasticity are shown. Clearly, the
given in Table 2 modulus of elasticity is best described using a lognormal
Abb. 5 Summenhäufigkeitsverteilung von Bongossi und drei distribution. The parameters of the three distributions
Häufigkeitsverteilungen. Ergebnisse für Lognormal und 2-param- are given in Table 3. The coefficient of determination
eter Weibull Verteilung sind in Tabelle 2 angegeben
(R2) between the static and dynamic modulus of elas-
ticity is 0.8 and the ratio between the average value of
the dynamic MoE and the static MoE was 1.09. The
3.2 Depth effect prediction of the bending strength using the modulus of
elasticity is not very precise. The coefficient of determi-
The depth effect has been analysed by dividing all nation between static MoE and MoR (n=254) is 0.41
strength data by the average value of the bending and between dynamic MoE and MoR (n=74) it is 0.05.
strength with a depth of 150 mm. The average value of Figure 8 gives the relationship between static MoE and
the strength was 90.0 N/mm2. Bending strength values MoR. Four test results showed a modulus of elasticity of
as a function of the depth, the depth effect equation of more than 20,000 N/mm2, combined with a relatively
Eurocode 5 (2002) and the regression equation for azobé
are shown in Fig. 6. The exponent of the depth effect is
determined to be 0.103 and Eq. 1 adapted to azobé reads
accordingly

Table 2 Distribution parameters and characteristic values for the

bending strength
Tabelle 2 Verteilungsparameter und charakteristische Werte der
Biegefestigkeit

Distribution Parameter

Mean SD f0.05

Normal (Fig. 4) 96.09 18.91 64.99

Lognormal (Fig. 4) 1.97 0.21 66.95
Distribution Parameter

Scale Shape Location f0.05

Fig. 6 Depth effect in azobé based on a reference depth of 150 mm.
Weibull 2-par. (Fig. 4) 103.78 5.77 – 62.02 The relative characteristic values of the individual samples are
Weibull 3-par. 55.38 2.73 49.52 68.18 indicated with O
Weibull lsm-cdf 2-par 103.57 5.64 – 61.17 Abb. 6 Höheneffekt für Bongossi auf der Basis einer Referenzhöhe
Weibull lsm-cdf 3-par 85.06 4.54 18.33 62.58 von 150 mm. Die charakteristischen Werte der Teilstichproben sind
mit O gekennzeichnet
8

Table 3 Distribution parameters for the static modulus of elasticity

Tabelle 3 Verteilungsparameter für den statischen E-modul

Distribution Parameter

Mean SD

Normal (Fig. 7) 17,200 2,500

Lognormal (Fig. 7) 4.23 0.063
Distribution Parameter

Scale Shape Location

Weibull 2-par. (Fig. 7) 18,290 7.21 –
Weibull 3-par. 6,600 2.48 11,350

able to determine distinctly local defects that hardly

affect the stiffness but have a large influence on the
bending strength.

3.4 Compression strength

The compression tests of sample C1 resulted in an

average compression strength of 69.7 N/mm2 and a
standard deviation of 6.5 N/mm2, all in wet condition.
The correlation between compression and bending
strength is poor (R2=0.38), even though the specimens
came from the same beam. The ratio between com-
pression and bending for the samples C1 and sample N1
is 0.74 on the average value and 0.85 on the character-
istic value, showing the influence of a lower coefficient of
variation in the case of compression stresses. The char-
acteristic value in wet condition is determined as 58.9 N/
mm2, compared to 32 N/mm2 for the value of strength
class D60 of EN 338, which is valid for dry conditions.

3.5 Shear strength

Out of 25 beams, 18 failed in a shear mode, while the

other seven failed in a bending mode. Shear failure was
either in the centre of the web or in the web near the
flange. Using Eqs. 6 and 7, the average shear strength
was determined at 17.7 N/mm2 and the standard devi-
ation was 2.1 N/mm2. Uncorrected values were 16.5 and
Fig. 7 Probability plots for the static modulus of elasticity; a 1.8 N/mm2, respectively. The coefficient of variation is
Normal probability (N/mm2); b Lognormal probability [log(N/ 11.9%, which is in the same range as for bending. No
mm2)]; c Weibull probability [log(N/mm2)]. Thin lines represent the correlation between the modulus of elasticity and the
distributions for which the parameters are given in Table 3 shear strength could be established, but since only a
Abb. 7 Summenhäufigkeit des statischen E-moduls; a Normalver-
teilung (N/mm2); b Lognormalverteilung (log(N/mm2)); c Weibull- small number of data is available, definite conclusions
verteilung (log(N/mm2)). Die Verteilungsparameter sind in Tabelle cannot be drawn. However, despite the small number of
3 angegeben data, the wet characteristic value of 13.9 N/mm2 is
considerably higher than the characteristic value of
5.3 N/mm2 as given in EN 338 for D60.
low strength. Although there is no obvious reason to
omit these data points from the analysis, R2 increases to
0.54 when this is done and the result becomes more in 3.6 Influence of moisture content on the strength
line with those obtained from softwood data (Blass and
Gard 1994). These four data points are also responsible Since, nearly all tests have been performed in wet
for the absence of a correlation between dynamic MoE condition, the characteristic values must be modified
and MoR. This indicates that stress wave analysis is not and related to dry conditions for engineering purposes.
 9

4 Conclusions

A large set of data for bending, compression and shear

strength values of azobé has been analysed. The static
bending strength can best be described using a normal
distribution, while the static modulus of elasticity best
fits a lognormal distribution. A three-parameter Weibull
distribution may also be used but the determination of
the location parameter r0 is complicated and depends on
the software and procedure used. However, the influence
Fig. 8 Relationship between static modulus of elasticity and of the choice of distribution on the lower fifth-percentile
bending strength characteristic value of the bending strength is relatively
Abb. 8 Abhängigkeit der Biegefestigkeit vom statischen E-modul small. This coincides with the observation made for
spruce and pine by Ranta-Maunus (2001), and for reli-
Only a few boards from sample N4 had moisture ability analysis a lognormal distribution might be used.
contents below fibre saturation point. The fibre satu- To determine the characteristic value, the traditional
ration point for azobé varies between 25 for interme- statistical distribution can be used, and also the method
diate wood (the layer between sapwood and heartwood) given in ISO 2394, which is based on Bayesian statistics.
and 30% for heartwood. The strength values of these The depth effect as it is observed in softwoods, is con-
boards were treated similarly as the values for the sat- siderably less pronounced in a species like azobé in
urated boards, since in the range between 17 and 25% which the absence of knots results in a different failure
moisture content, the influence on the mechanical behaviour. Values for compression and shear strength
properties is considered to be small with respect the are considerably higher than the values given in strength
whole sample. From the test results, it can be concluded class D60 in EN 338 to which azobé is assigned. To a
that for all strength properties the wet characteristic lesser extent, this is also the case for the bending
strength is higher than the dry characteristic strength strength, Taking into account that all data were deter-
from strength class D60. For bending strength, the mined using wet specimens.
Wood Handbook (1999) gives a ratio between the green
state and 12% moisture content of 1.45 for both References
bending and shear and of 1.32 for compression. These
values apply for defect free material. Most studies show Barrett JD, Griffin H (1989) Size effects for Canadian dimension
an increase in bending strength with a decrease in lumber, CIB-W18 Timber Structures, Paper 22-6-1, Berlin,
moisture content below the fibre saturation point. German Democratic Republic, September
However, for species or grades with large defects such Barrett JD, Lau W (1989) Moisture content adjustments for in-
grade data, CIB-W18 Timber Structures, Paper 22-6-2, Berlin,
an increase in strength is sometimes not found at the German Democratic Republic, September
lower fifth-percentile level and the increase in strength Barrett JD, Fewell AR (1990) Size factors for the bending and
depends on the location in the cumulative density tension strength of structural timber, CIB-W18 Timber Struc-
function (Barrett and Lau 1989). Because the lower tures, Paper 23-10-3, Lisbon, Portugal, September
Benjamin JR, Cornell AC (1971) Probability statistics and decision
fifth-percentile is mainly determined by the presence of for civil engineers. McGraw-Hill
knots in these studies, it might be argued that for azobé Blass HJ, Gard WF (1994) Machine strength grading of timber,
this will not be similar. For engineering purposes, the Pacific Timber Engineering Conference, vol 2. Gold Coast,
following procedure could be adopted for structural Australia, pp 598–603
Bohannan W (1966) Effect of size on bending strength of wood
tests on wet tropical hardwoods that are difficult to dry beams. US Forest Products Laboratory, Bulletin No. 56
to a moisture content of about 12%. The procedure is NIST (2001) Dataplot. National Institute of Standards and Tech-
based on the fact that in design codes such as Eurocode nology, USA
5 (2002) the average ratio of the kmod values between Eurocode 5 (2002) Design of timber structures, Draft version
service class 1/2 and 3 is 1.2 for all strengths. Applying European Standard EN 338. Structural timber—strength classes
European Standard EN 384. Structural timber—determination of
this ratio leads to a dry characteristic bending strength characteristic values of mechanical properties and density
of 78.0 N/mm2, a compression strength of 70.7 N/mm2 European Standard EN 1193. Timber structures—structural timber
and shear strength of 16.7 N/mm2. However, since no and glued laminated timber—determination of shear strength
tests were performed with dry timber, it is not known and mechanical properties perpendicular to the grain
European Standard EN 1912. Structural timber—strength clas-
whether the strength really increases as it does for ses—assignment of visual grades and species
softwood. Therefore, the dry characteristic strength Frühwald E, Thelandersson S (2001) Nordic wood: reliability of
values that were calculated may be used for structures timber structures, Reliability analysis, COST Action E24 Reli-
in service class 3; it remains questionable, however, to ability Analysis of Timber Structures, Copenhagen Meeting,
19–20 October 2001
assume these dry characteristic values for bending, Fewell AR, Curry WJ (1983) Depth factor adjustments in the
compression and shear, respectively, for structures as- determination of characteristic bending stresses for visually
signed to service classes 1 or 2. graded timber. Structural Engineer, vol 61B, No.2
10

Fewell AR, Glos P (1988) The determination of characteristic Schickhofer G (2001) Determination of shear strength values for
strength values for stress grades of structural timber. Part 1. GLT using visual and machine graded spruce laminations CIB
CIB-W18 Timber Structures, Paper 21-6-2, Parksville, Canada W18 Timber Structures, Paper 34-12-6, Venice
Houtinstituut TNO (1961) Buig- en drukproeven bij azobé van Van der Linden MLR, Van de Kuilen JWG, Stolle P, Katsma H
handelskwaliteit en afmetingen en bijbehorend klein foutvrij (1996) Full scale testing of timber sheet pile walls. World timber
hout (in Dutch). Report H-61-78 engineering conference, vol III, New Orleans, pp 557–564
Houtinstituut TNO (1977) Sterkte eigenschappen van gebruikt en Van Douwen AA, Kuipers J, Loof HW (1958) Corrections to the
ongebruikt azobé (in Dutch). Report H-77-93 mean value and the standard deviation from test series with
ISO 2394 (1998) General principles on reliability for structures symmetrical test specimens, Stevin laboratory report. 4-82-9/oe-
Leicester DE, Breitinger HO (1992) Measurement of shear 5, Delft University of Technology
strength. In: IUFRO S5.02 Proceedings, Nancy Weibull W (1939) A statistical theory of the strength of materials,
NEN 5480, Quality requirements for timber—Species Azobé Royal Swedish Institute for Engineering Research. In: Pro-
Rammer DR, McLean DI (1996) Recent research on the shear ceedings N. 141:45
strength of wood beams. In: Proceedings of the international Wood Handbook (1999) Forest Products Laboratory, Madison
wood engineering conference, vol 2, New Orleans, pp 96–103,
28–31 October 1996
Ranta-Maunus A (2001) Summary report on existing strength data.
COST Action E24 Reliability Analysis of Timber Structures,
Copenhagen Meeting, 19–20 October 2001