Professional Documents
Culture Documents
Properties Relationship Evaluation and Plasticity Analytical Model Approach For Brazilian Tropical Species
Properties Relationship Evaluation and Plasticity Analytical Model Approach For Brazilian Tropical Species
https://doi.org/10.1007/s00107-020-01605-x
ORIGINAL
Abstract
In anisotropic materials such as wood, different responses to bending induced stresses could be expected. This study evalu-
ated the behavior of properties of modulus of rupture (MOR) in bending, the tension strength parallel to grain and the
compression strength parallel to grain of Brazilian tropical hardwood species in accordance with Brazilian, European and
International Standards, and an analytical equation considering plasticity in the compressed region. Experimental data were
obtained using clear wood specimens and tests on ten species. An analysis of the mechanical behavior was conduced for three
strength relationships: bending to compression, bending to tension and tension to compression. Linear regression models
were determined using the 5th percentile and compared to the ones found by the standards cited. The experimental results
were also evaluated to an analytical equation modified from the moment equilibrium of a cross-section. Of the equations
evaluated in this study, the ones suggested by the European and International standards for tension strength from bending
strength were the ones which best estimated the 5th percentile on experimental data. Other relationships did not show a good
agreement with the equations. The analytical equation used had a mean model error 37.4% higher than the experimental
values for the prediction of the ratio of tension to bending strength.
Timber is an anisotropic material that is considered as ortho- In general, standards have approximate formulae for estimating
tropic for simplification in structural design. This means that mechanical properties. The Brazilian code (NBR 7190 1997)
three axes possess different characteristics (longitudinal, is no different, although two of them are contradictory in terms
tangential and radial). It is well-known by researchers that of the values of bending, tension and compression. The first
different results are obtained when a load is applied to the relationship, Eq. 1, shows that there is a significant difference
same axis but in a different direction (i.e., pulling instead of between the characteristic compression and tension characteristic
pushing), which will put the piece in tension instead of com- strength of timber, the latter being higher. The second relation-
pression. The stresses vary in behavior and value. Tension ship, Eq. 2, states that the characteristic bending strength has
tends to have a linear elastic behavior until brittle rupture, the same value as the characteristic tension parallel to the grain.
and compression behavior is mostly linear until a propor-
fc0,k
tional limit, when it becomes elastoplastic until rupture. As = 0.77 (1)
quasi-static bending is a combination of mainly tension and ft0,k
compression on long beams (length/height greater than 21),
among other stresses, it is of particular importance in struc- fm,k = ft0,k (2)
tural engineering to understand how this material reacts.
The method for the estimation of the bending strength
according to this same code considers that timber com-
* Leonardo Araújo Marcolin
pression and tension strength parallel to the grain have the
eng.marcolin@gmail.com
same value, allowing the simplification shown in Eq. 3. No
1
Structural Engineering Department, Escola de Engenharia account is taken of the non-linear behavior of timber in com-
de São Carlos (USP‑EESC), Universidade de Sâo Paulo, pression in the estimation of the bending strength.
Trabalhador Sãocarlense Av. 400 São Carlos, SP, Brazil
13
Vol.:(0123456789)
European Journal of Wood and Wood Products
ISO 16598 ft0,k = 0.5fm,k fc0,k = fm,k (1.7 − 4.6.10−2 fm,k 2.1 Experimental data sampling
(2013) +4.6.10−4 fm,k
2
) (a)
fc0,k ≤ 0.59fm,k (b) The sample used in this study was comprised of ten spe-
EN 384 ft0,k = 0.6fm,k 0.5
fc0,k = 4.3fm,k cies of tropical timber from Brazil: Mezilaurus itauba
(2016) (Itaúba), Erisma sp. (Cambará Rosa), Vataireopsis ara-
NBR 7190 ft0,k = fm,k fc0,k = 0.77.ft0,k roba (Angelim Araroba), Clarisia racemosa (Oiticica),
(1997)
Ocotea sp. (Canelão), Tabebuia sp. (Ipê), Goupia glabra
13
European Journal of Wood and Wood Products
Aubl. (Cupiúba), Hymenaea sp. (Jatobá), Manilkara The strength properties parallel to the grain, related to bend-
sp.(Maçaranduba) and Dinizia excelsa Ducke (Angelim ing strength, were evaluated with estimate formulations sug-
Ferro). The properties of interest were the compression and gested by three normative codes: ISO 16598 (2013), EN
tension strength parallel to the grain and bending strength 338 (2012) and NBR 7190 (1997). A graphical analysis was
( fm ). Each property was estimated according to the Brazil- conducted to verify the Brazilian timber species behavior
ian standard for timber structures design NBR 7190 (1997). compared to relationships suggested by the literature. The
The following numbers of clear wood specimens were used: second evaluation determined the cumulative distribution of
12 for the first 5 species, 34, 36, 35, 24 and 23, respectively the three properties experimentally estimated and calculated
for the other species, making 212 specimens for each prop- the ratios of each property, following the equations used
erty in total. Specimens were cut (cross-section of 50 mm × in codes with a regression analysis and a comparison with
50 mm) from each plank of timber in matched sets of 3, i.e. the references cited. These were plotted to provide a better
one specimen for each of the properties, so that the different understanding of the data.
characteristics could be compared in the analysis.
2.3 Validation of the analytical model
2.2 Properties relationship and behavior
The model chosen was derived by the equation of moment
As noted above, the Brazilian standard for the design of equilibrium of a cross-section in Van der Put (2010). As
timber structures uses relationships that contradict them- the equation estimates the values of the maximum bending
selves, which dictate the relation of compression and ten- moment, three experimental properties were used: tension
sion strength parallel to the grain, and bending strength. and compression strength parallel to the grain and maximum
To verify the relationship, the 5th percentiles were cal- bending moment at failure. These values were first analyzed
culated from the experimental data and compared with graphically, on the basis of coordinates ( ft0 , Mmax , fc0/ ft0 ),
the characteristic values from the standards. Firstly, the along with the values estimated by the analytical formula-
5th percentile values were calculated for five samples tion, as explained in the results and discussion.
formed from all species together, ensuring a minimum of
40 specimens per sample. The suggested reduction of 0.5
for the estimation of the percentile was adopted from ISO
16598 (2013). The characteristic values were calculated
3 Results and discussion
by the 5th percentile of the samples. The characteristic
3.1 Experimental data sampling
strength values were calculated using Eq. 8, as suggested
in NBR 7190 (1997), where x1 is the lowest value of the
Table 2 shows the range of values obtained from the experi-
sample, xn is the highest value of the sample, n is the num-
mental tests for each property and species, giving the mean,
ber of specimens in the sample and fk is the characteristic
minimum and maximum value. For the three properties used
strength value.
in this study, the species that showed the lowest values for
[( ) ]
x1 + x2 + ⋯ + xn∕2−1 most properties was Erisma sp. and the one that showed the
fi,k,NBR = 2 n − xn∕2 1.1 (8) highest values for all properties was Hymenaea sp.. The only
−1
2 exception was the lowest value of the tension strength, which
Mezilaurus itauba 61.7 69.0 76.8 4.2 82.5 110.3 154.0 21.0 87.6 116.6 140.8 15.2
Erisma sp. 27.3 34.5 42.6 5.1 30.6 45.4 66.7 13.8 40.3 63.2 81.7 13.4
Vataireopsis araroba 40.4 50.7 63.0 6.9 42.7 73.6 117.0 11.6 54.0 77.2 98.9 16.2
Clarisia racemosa 58.0 70.3 79.0 6.3 53.3 85.8 106.5 23.4 71.6 107.5 134.6 14.6
Ocotea sp. 44.3 52.5 60.2 5.1 50.2 89.4 124.6 18.6 79.3 97.1 111 9.4
Tabebuia sp. 42.0 64.3 82.0 11.9 52.8 96.4 160.8 27.5 76.5 119.3 159.4 21.3
Goupia glabra 30.2 52.8 72.5 11.9 28.9 68.9 115 22.1 44.5 75.8 111.8 16.4
Hymenaea sp. 75.4 93.8 109.0 8.2 103.3 153.4 210.2 30.7 86.0 157.0 211.0 29.2
Manilkara sp. 64.0 84.0 102.0 9.0 103.0 139.5 175.0 24.2 98.0 134.3 182.0 22.0
Dinizia excelsa 53.0 79.3 105.0 11.5 60.0 118.9 197.0 36.4 87.0 132.5 166.0 21.3
13
European Journal of Wood and Wood Products
was for Goupia glabra with a value of 28.9 N/mm2 , although Figure 1 presents the cumulative distribution of the exper-
Erisma sp. was the second lowest with a value of 30.6 N/ imental results for the three properties discussed as solid
mm2 . This shows that it is reasonable to consider the distri- curves, and the three ratios as the black dashed curves. The
bution of these two species as the furthest apart from each lower solid curve shown is formed by the values of the com-
other among the samples used in this study for this property. pression strength, the middle one by the tension strength and
The lowest values for compression, tension and bend- the highest being the bending strength. Until approximately
ing strength, were respectively: 27.3, 28.9 and 40.3, and the the 70th percentile value, both tension and bending strength
highest values for compression, tension and bending follow a similar behavior, above which tension strength
strength, were respectively: 109, 210.2 and 211. The lowest becomes equal and then higher at the 90th percentile. Ten-
mean values for compression, tension and bending strength, sion and compression start at approximately the same value,
were respectively: 34.5, 45.4 and 63.2, and the highest mean although at the 90th percentile, tension strength is higher
values for compression, tension and bending strength, were than bending strength, showing that compression strength
respectively: 93.8, 153.4 and 157. The lowest standard has a lower variability than both the other two properties.
deviation (SD) values for compression, tension and bend- The curves formed by the cumulative distribution of
ing strength, were 4.2, 11.6 and 9.4 for the species Mezilau- the ratios of the properties shown in Fig. 1, from the lower
rus itauba, Clarisia racemose and Ocotea sp., respectively. one to the higher, are as follows: compression to bending
The highest standard deviation (SD) values for compression, strength, compression to tension strength and tension to
tension and bending strength, were 11.9, 36.4 and 29.2 for bending strength. It is important to emphasize that the value
the species Tabebuia sp. and Goupia glabra Aubl., Dinizia of the ith percentile of one property does not correspond to
excelsa Ducker and Hymenaea sp., respectively.
2.5
3.2 Properties relationship and behavior 200 fc0 ft0 fm
2
Strength Value (N/mm²)
Ratio
the population, for each species, was calculated by using 100
the Brazilian standard cited and the experimental 5th per- 1
centile, shown in Table 3. The Brazilian code showed for
all three cases higher values for the strength properties than 50
0.5
the experimental data. The minimum value of the ratio of
fc0/ft0 ft0/fm fc0/fm
the estimated to the experimental tension strength was 0.965
0 0
for Manilkara sp. and the maximum was 1.308 for Goupia 0 0.2 0.4 0.6 0.8 1
glabra. Of the 36 ratios calculated, only three were below Percentile
1.0.
Fig. 1 Cumulative distributions of experimental data
Table 3 Characteristic and 5th Species fc,k (N/mm 2 ) ft,k (N/mm 2 ) fm,k (N/mm 2 )
percentile values by species
fc,05 fc,k,NBR ft,05 ft,k,NBR fm,05 fm,k,NBR
Mezilaurus itauba 62 68 83 89 89 95
Erisma sp. 27 27 31 32 40 44
Vataireopsis araroba 58 62 56 74 74 90
Clarisia racemosa 41 44 44 52 54 63
Ocotea sp. 45 47 53 72 80 86
Tabebuia sp. 44 45 59 68 86 93
Goupia glabra 34 38 33 48 52 55
Hymenaea sp. 78 88 111 111 97 118
Manilkara sp. 68 77 107 105 99 107
Dinizia excelsa 61 66 66 83 90 109
5th percentile 61 47 36
Mean 114 104 69
13
European Journal of Wood and Wood Products
percentiles the slopes are constant. Within this interval, the 150 Goupia glabra Aubl.
Hymenaea sp.
f t0 (N/mm²)
Manilkara sp. 8
slope for the compression to bending strength ratio was the Dinizia excelsa Ducke
EN
33
Exp Data
lowest, with a value of 0.315, followed by 0.504 for the com-
100
pression to tension strength ratio and 0.538 for the tension to 165
98
ISO
bending strength ratio. This shows that within this interval,
which comprises 50% of each sample, the first ratio cited has 50 90
71
the lowest variability. The changes in the values for the ratios R
NB
within this interval are as follows: 0.158, 0.252 and 0.269.
These slopes were obtained by linear regression within the 0
0 50 100 150 200
specified intervals. Table 4 shows the values of significant
f m (N/mm²)
percentiles of the cumulative distributions of the ratios.
Figure 2 shows the experimental results of the tension
parallel to the grain against the bending strength. Note that Fig. 2 Relationship of tension to bending strength
each species tends to be located within a particular region of
the graph. Confirming earlier observations above, the spe- Figure 4 shows the relationship between the tension and
cies that have the lowest values are Erisma sp., followed by compression strength. This figure shows the same behavior
Goupia glabra. The ones that have the highest values are of the distribution of the values as Figs. 2 and 3, regarding
Hymenaea sp, followed by Manilkara sp. and Dinizia excels. the species and magnitude of result. Of the three standards
The three approximating equations are also illustrated. used in this study, the only one that provides and estimated
The one suggested by the Brazilian standard shows a bet- value is the Brazilian standard. It results in higher values
ter prediction of the values above the mean rather than the than the 5th percentile, which cannot be considered as
lower tail of the distribution. The filled black circles indicate appropriate.
the characteristic values of the sample formed by the whole
sample, considering a minimum sample size of 40 speci- 3.3 Regression analysis
mens. The data showed better agreement with the EN 338
(2012) derived values than with those of ISO 16598 (2013), An analysis of the 5th percentiles values was conducted for
which was conservative by comparison. each pair of variables discussed for the regression. Firstly,
The behavior of the compression strength compared with for the three cases in this study, the analysis was conducted
the bending strength is shown in Fig. 3. A similar distribu- for paired variables alone, i.e. one property against the other.
tion of the species can be observed in Fig. 2, where Erisma The regression was defined to intercept the origin of the
sp. corresponds to the lowest values and Hymenaea sp. to the axis (0,0) with a linear behavior. The second analysis was to
highest. The approximate formulae found in the standards
did not represent appropriately the characteristic values of
598 (a)
0
25th 0.747 0.525 0.557 0 50 100 150 200
50th 0.874 0.595 0.694 f m (N/mm²)
75th 1.044 0.684 0.821
95th 1.492 0.893 1.099
Fig. 3 Relationship of compression to bending strength
13
European Journal of Wood and Wood Products
Mezilaurus itauba
with the highest slope (tension/bending strength ratio against
200 Erisma sp.
Vataireopsis araroba
tension strength). As already pointed out, the formula that
Clarisia racemosa
Ocotea sp.
best predicted the values for these two properties was the one
Tabebuia sp. in EN 338 (2012), which was the closest to the regression
150 Goupia glabra Aubl.
Hymenaea sp. defined by the experimental 5th percentile values. The equa-
f t0 (N/mm²)
Manilkara sp.
Dinizia excelsa Ducke
Exp Data
tion obtained using the regression with the dependency gave
100
a conservative estimate of the values of the lower half, while
being unsafe for the upper half, relative to the characteristic
values. The expression found by Burger and Glos (1997) did
50 not fit well for the purpose of this study.
R7
19
0
The equations suggested for the data analyzed in this
NB
study are shown in Eqs. 9, 10 and 11. It should be empha-
0 sized that these equations cannot be used concurrently. The
0 20 40 60 80 100
f c0 (N/mm²)
estimated 5th percentile can only be one of the terms on the
left, derived from known variables on the right side of the
equations; the process is not reversible.
Fig. 4 Relationship of tension to compression strength
ft0,k = 0.64867fm,k (9)
identify whether the ratios had any dependency on the other
properties. Figure 5a shows the trend analysis of the ten- fc0,k = 0.47939fm,k (10)
sion strength to bending strength ratio ( ft0 ∕fm ) to the tension
strength parallel to the grain strength ( ft0 ). To determine ft0,k = 1.015fc0,k (11)
the dependency, a linear regression was calculated of the
5th percentile of the ratio against the property for which the
trend has been evaluated, and the slope of the regression is 3.4 Validation of the analytical model
given in Table 5. The highest slope was 0.0041, as shown
in Fig. 5a. The validation was carried out by comparing the estimated
Figure 5b shows the values of the experimental data values obtained using the equation and the experimental val-
together with the equations suggested in the standards cited ues for each specimen ( fc0 , ft0 , Mmax ). For the tension and
previously, the equation obtained by Burger and Glos (1997) compression stresses parallel to the grain in the formulae,
and the two calculated regressions, one without considera- the range of the strength property values ( fc0 , ft0 ) of the
tion of the dependency (WO Tendency) and the other con- experimental data was considered.
sidering the dependency (W Tendency) for the regression
2.5
Exp. Data
200 WO Tendency
W Tendency
2
NBR 7190
ISO 16598
150 EN 338
f t0 (N/mm²)
100
1
0.5 50
0 0
0 50 100 150 200 0 50 100 150 200
f t0 (N/mm²) f m (N/mm²)
13
European Journal of Wood and Wood Products
Fig. 6 Analysis of the influence of the plasticized depth on the formulae a minimum and maximum surface values, b surface for the maximum
tension strength value
Table 5 Trend analysis Trend Ratio Slope (mm/mm) Ratio Slope (mm/mm) Ratio Slope (mm/mm)
fc0 ft0 ∕fm 0.0028 fc0 ∕fm 0.00194 fc0 ∕ft0 0.00119
ft0 0.0041 − 5.4 (e−5) − 0.00128
fm 0.0012 0.00029 − 0.00014
The height of the specimen used for the experimental for the tension strength property were analyzed, resulting in
estimation of the maximum bending moment was of 50 mm, two surfaces as shown in Fig. 6a, where the lower surface,
so the possible plasticized height is between 0 mm and 50 was obtained from the maximum experimental value for
mm, since there is no physical meaning for a different value the tension strength, and the upper surface, for its minimum
in this case. value. The range for the bending moment used in the formu-
To better visually depict the results, it was found appro- lae to create both graphs in Fig. 6 was the same as the one
priate to evaluate the behavior of the formulae. All values obtained on the experimental data (40.3–211 N/mm2 ) and
13
European Journal of Wood and Wood Products
Model Error
4
0
RTM RCM RCT
Fig. 8 Experimental and analytical model analysis: compression to
tension Fig. 9 Experimental and analytical model error
for the plasticized depth x, 0–25 mm. The results obtained The model error for each equation was calculated as the
for the upper surfaces converge closer to those for the lower ratio of the estimated value of the formulation to the experi-
surface when both moment and x-dimension values are con- mental value. The results were plotted as a box and whisker
sidered. However, when the x-dimension value exceeds 40 plot (Fig. 9), where RCT is the ratio of the compression
mm, which is 80% of the cross-section, the ratio value rises strength to tension strength parallel to the grain, RTM is the
exponentially, which is inconsistent with the experimental ratio of the tension strength parallel to the grain to the bend-
data, so it will not be considered in the analysis. ing strength and RCM is the ratio of the compression strength
As the values of the plasticized depth decrease (x value parallel to the grain to the bending strength. As indicated in
of 25–12.5 mm), the surface plot becomes concave, which Figs. 8, 7 and 9, the analytical model gave the best estimate of
happens in the lower half of the height, as shown in Fig. 6b. the ratio of tension to bending strength, with a mean value of
Note that the function estimates the lowest values at a quar- 1.374, median of 1.334 and standard deviation of 0.486, fol-
ter of the total height for x. lowed by the ratio of compression to bending strength and the
The ranges considered for the bending moment and ratio of compression to tension strength, with a mean value
tension strength were 1–5.106 N mm, and 1–220 N/mm2 , of 4.502, median of 4.470 and standard deviation of 0.863.
respectively. The surface shown in Fig. 8 was the one that
resulted in the lowest values, giving an estimate, close to the
data, for an x-dimension of a quarter of the height. This was 4 Conclusion
chosen according to the analysis of the equation behavior.
All the experimentally obtained ratios of the compression Several analyses were conducted in this study regarding the
to tension strength properties had a lower value than the bending, tension and compression strength parallel to the
surface estimated by the analytical model. grain estimate formulae suggested in standards and an analyti-
The same procedure of analysis was applied to both cal equation based on solid mechanics. This was done using
Eqs. 6 and 7, obtaining Fig. 7a, b. The first figure shows the experimental data of 10 Brazilian tropical hardwood species.
relationship of the tension strength parallel to the grain to For the determination of the strength properties by test,
the bending strength. The experimental data shows a better the species that showed the lowest values for most properties
agreement with the analytical estimate in Fig. 7a, with a was Erisma sp. and the one that showed the highest values
mean value of 0.925 for the experimental sample and 1.154 for all properties was Hymenaea sp.. It was noted that the
for the analytical formulation. Figure 8 shows the relation- results for each species tended to fall within a distinct interval,
ship of the compression strength parallel to the grain to the where Erisma sp. was in the lower range for all properties, and
bending strength. In this case, the experimental values did Hymenaea sp. in the higher range. Regarding the Brazilian
not show as good agreement, to the analytical estimate as code method for the calculation of the characteristic values of
in Fig. 7a, with a mean value of 0.624 for the experimental the properties, most of the values obtained were higher than
sample and 2.703 for the analytical formulation.
13
European Journal of Wood and Wood Products
the 5th percentile values from the experimental data. The ratio References
of the value calculated by code formula and the values derived
from the experimental data had a range of 0.965–1.308. Burger N, Glos P (1997) Strength relationships in structural timber
Of the three pairs of relationships studied, the equation subjected to bending and tension. In: International council for
research and innovation in building and construction working
suggested by EN 338 (2012) of the tension strength paral- commission W18—timber structures CIB-W18, Vancouver
lel to the grain to the bending strength was the closest to EN 338 (2012) Structural timber: strength classes (Unpublished Draft).
the data. ISO 16598 (2013) for this same relationship had Standard, European Committee for Standardization—CEN
lower estimate values than the data, and for the compression EN 384 (2016) Structural timber: determination of characteristic values
of mechanical properties and density. Standard, European Com-
strength to the bending strength it did not show a good esti- mittee for Standardization—CEN
mate. In general, the Brazilian code showed a better agree- Galligan WL, Gerhards CC, Ethington RL (1979) Evolution of tension
ment with mean or above the mean estimations than the lower design stress for lumber. Report FOL 28, Forest Service, Forest
tail. Three equations for the properties were suggested for the Products Laboratory
ISO 16598 (2013) Timber structures: structural classification for sawn
data of this study, for Brazilian hardwood species. timber (Unpublished Draft). Standard, International Organization
The results of the analysis of the analytical model for Standardization, Geneva
derived from the moment equilibrium of the cross-section NBR 7190 (1997) Design of timber structures. Standard, Associação
equation shows that the best estimation was for the ratio Brasileira de Normas Técnicas—ABNT, Rio de Janeiro
Steiger R, Arnold M (2009) Strength grading of Norway spruce struc-
of tension to bending strength. The model error for this tural timber: revisiting property relationships used in EN 338
ratio had a mean value of 1.374 and a standard deviation classification system. Wood Sci Technol 43:259–278. https://doi.
of 0.486. For both the other formulae, ratio of compression org/10.1007/s00226-008-0221-6
to bending strength and ratio of compression to tension Van der Put TACM (2010) Failure criterion for timber beams loaded in
bending, compression and shear. Wood Mater Sci Eng 5(1):41–49.
strength, a mean value of 4.502 and standard deviation of https://doi.org/10.1080/17480270903582163
0.863 were calculated.
Publisher’s Note Springer Nature remains neutral with regard to
Acknowledgements This research was supported by Brazilian fund- jurisdictional claims in published maps and institutional affiliations.
ing agencies: Conselho Nacional de Desenvolvimento Científico e
Tecnológico–Brasil (CNPq) and Coordenação de Aperfeiçoamento de
Pessoal de Nível Superior–Brasil (CAPES).
13