Docstoc

Bending strength and stress wave grading of tropical hardwoods

Document Sample
Bending strength and stress wave grading of tropical hardwoods Powered By Docstoc
					Bending strength and stress wave grading of
           (tropical) hardwoods

                    Dr.ir. J-W.G. van de Kuilen
          Faculty of Civil Engineering and Geosciences
                 Delft University of Technology


                    ir. G.J.P. Ravenshorst
           TNO Building and Construction Research

                         Workshop
 Probabilistic Modeling in Reliability Analysis of
                Timber Structures

    COST E24 Reliability analysis of timber structures
         Joint Committee on Structural Safety

            10-11 Oktober 2002 - ETH Zurich

                         Switzerland
1. Introduction1

Tropical hardwoods are often used in structures that are exposed to direct weathering or soil
conditions where softwood cannot be applied without the use of chemical treatments. A
hardwood species often used in countries like the Netherlands, Belgium, Germany and the
United Kingdom is azobé, also known as ekki (lophira alata). Azobe is used in structures
such as bridges, road decks and sheet pile walls. The material properties of azobé are
reasonably well documented, Blass and Van De Kuilen, [1991], Van der Linden et al. [1994].
However, with the current environmental pressure on harvesting procedures for azobé, new
species, especially from Brazil, have entered the market. These species are often harvested in
accordance with the FSC (Forest Stewardship Council) regulations. Unfortunately, available
strength data for these species is often limited and if there is strength data available, this data
is often acquired from small specimens or the test procedure used is unknown. Also, grading
rules for these species are often not available.
A major setback for these new species is the European standard EN 384. This standard
requires large numbers of tests for the determination of bending strength, modulus of
elasticity and density, before a species can be assigned to a strength class. The theoretical
background for the number of tests required is not very well documented and sampling
procedures for new species that grow over vast areas are not specified.
From a viewpoint of reliability there is a market mechanism that prevents that new species are
used in heavy structures without sufficient knowledge of that species. Before a new species is
introduced on the market in large quantities, small projects will be carried out and designed
on the basis of a few tests on the material that is available for the specific projects. Timber
trader, contractor, designer and future owner closely work together in these stages of the
project, avoiding large mistakes with severe consequences. Following a standard such as EN
384 would prevent the use of new species because of the costs involved. For these small
projects a limited amount of destructive tests in the laboratory can give sufficient information
about the strength of the timber batch, while non-destructive tests are available to determine
modulus of elasticity and density for instance. Knowledge about the species is increased
during small building projects and more tests can be performed when larger projects with
greater risks involved are scheduled.
In order to obtain some basic knowledge about a number of recently introduced species in the
Netherlands a large project was set-up. The main goal of this project was to develop a general
strength model, primarily based on parameters that can be determined non-destructively, that
can be used to safely estimate the strength of new species. 12 new species were selected and
from each species 40 beams (the minimum required according to EN 384 for strength class
assignment). The data of these 12 species would form a database together with the already
existing database of azobé and is used whenever a new species needs to be judged or when
one of the 12 species will be used in large structures. [Van de Kuilen, 1999]. [Van de Kuilen
& Ravenshorst, 2000].


1

This paper and part of this research work was done during a sabbatical leave of J.W.G. van de Kuilen from Delft
University of Technology in 2002. He gratefully acknowledges the Dipartimento di Ingegneria Strutturale of the
Politecnico di Milano, Italy, for their hospitality and cooperation.




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                                     2/12
2. Strength data of azobé

Over the years about 180 tests on azobé beams and boards have been performed. All tests are
performed on structural sizes. Sizes that are used in structures range from 40 mm thick boards
in sheet pile walls to built up beams with individual member sizes of 180 x 180 mm.

In figure 2.1 the relationship between static modulus of elasticity and bending strength (MoR)
is shown, while in figure 2.2 the dynamic modulus of elasticity and bending strength is


                                 Relationship between static modulus of elasticity and bending                                                Relationship between dynamic modulus of elasticity and bending
                                                      strength of azobe                                                                                             strength of azobe
   Bending strength (N/mm )




                                                                                                                Bending strength (N/mm )
                              120.00                                                                                                       120.00
 2




                                                                                                                2
                              100.00                                                                                                       100.00
                               80.00                                                                                                        80.00
                               60.00                                                                                                        60.00
                               40.00                                                                                                        40.00
                               20.00                                                                                                        20.00
                                0.00                                                                                                         0.00
                                   10000 12000         14000   16000    18000   20000      22000     24000                                      10000   12000   14000    16000    18000   20000      22000     24000
                                                                                       2
                                                    Static modulus of elasticity (N/mm )                                                                     Dynamic modulus of elasticity (N/mm2)




Figure 2.1 Static MoE versus MoR                      Figure 2.2 Dynamic MoE versus MoR
shown. In both cases the relationship is not very strong.

The relation between density and bending strength as is shown in figure 2.3

                                       Relationship between density and bending strength of azobe
                              120.00
  Bending strength (N/mm 2)




                              100.00

                               80.00

                               60.00

                               40.00

                               20.00

                                0.00
                                       800       900       1000        1100         1200      1300       1400
                                                                                3
                                                                  Density (kg/m )




Figure 2.3 Density versus MoR

Some beams had severe grain deviations and this was found to be the main strength
determining parameter. On the basis of the test series, azobé has been assigned to strength
class D60 of EN 338 in conformity with EN 384. EC5 allows the use of the following depth
factor:




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                                                                                                                                        3/12
                              0 .2
      150 
kh =                                                                                            (1)
      h 

for depths lower than 150 mm with a maximum value of 1.3. In the Netherlands azobé is
assigned to a strength class with a characteristic bending strength of 70 N/mm2 , (Strength
class K70 according to NEN 5498) but the use of a depth factor is not allowed for strength
classes with a characteristic bending strength higher than 40 N/mm2 .

In figure 2.4 strength data of azobé is shown, as a function of depth and the relationship
between depth and strength is much smaller than is assumed in Eurocode 5. This, naturally, is
caused by the fact that grain deviation does not increase with beam size. It must be recognized
that bridge manufacturers in particular suffer from the depth effect and the corresponding
assignment to strength class D60 it causes. The average strength of azobé was found to be 101
N/mm2 with a standard deviation of 15.1 N/mm2 .
                    1.8

                                                                    Eurocode 5/EN384

                    1.6                                             Azobé



                    1.4
 Depth factor [-]




                    1.2



                     1



                    0.8



                    0.6



                    0.4
                          0          50    100                150            200              250
                                                 Depth [mm]

Figure 2.4 Depth effect in azobé as compared to Eurocode 5.

3. New species

A total of 11 new hardwood species were selected and tested in conformity with EN 384, nine
from (sub)tropical regions and two from Europe. From each species 40 beams were randomly
selected from trade stocks. The species are summarized in table 1.

Based on visual grading of the pieces a general requirement for grain deviation was
determined at 1:10. In addition a maximum knot area of 0.2 was defined, where knot area is
determined as the ratio between the sum of diameters of the knots, measured perpendicular to




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                            4/12
the longitudinal beam axis, and the circumference of the beam. Although most species do not
have knots at all, a value had to be chosen to avoid the use of beams with large growth
disturbances, which sometimes are knots, but sometimes just appear to be knots. Grain
deviations around these areas often cause very low strength values.

Table 3.1 "New" species
Species                 Origin                                                  Latin name

Angelim vermelho                                   Brazil                       Dinizia excelsa
Basralocus                                         Surinam                      Dicorynia guianensis
Cumaru                                             Brazil                       Dipteryx odorata
Denya                                              Ghana                        Cylicodiscus gabunensis
European oak                                       Poland                       Quercus robur
Karri                                              South-Africa                 Eucalyptus diversicolor
Massaranduba                                       Brazil                       Manilkara bidentata
Nargusta                                           Bolivia                      Terminalia amazonia
Piquia                                             Brazil                       Caryocar villosum
Robinia                                            Hungary                      Robinia pseudoacacia
Vitex                                              Solomon islands              Vitex cofassus (spp)

In figure 3.1 the test results for static modulus of elasticity and bending strength are shown. In
some cases a static modulus of elasticity of more than 30.000 N/mm2 was found. In figure 3.2
and 3.3 the same is done for the dynamic modulus of elasticity and density respectively is
done. The dynamic modulus of elasticity was determined using stress wave analysis
equipment developed at TNO Building and Construction Research.
                                                      Strength data of 11 "new" species


                                160.0
                                                                                                     karrie
                                140.0
                                                                                                     European oak
   Bending strength (N/mm 2 )




                                120.0                                                                cumaru
                                                                                                     angelim vermelho
                                100.0
                                                                                                     vitex
                                 80.0                                                                massaranduba
                                                                                                     nargusta
                                 60.0
                                                                                                     piquia
                                 40.0                                                                denya
                                                                                                     basralocus
                                 20.0
                                                                                                     robinia
                                  0.0
                                        0   5000    10000    15000      20000        25000   30000
                                                                                 2
                                               Static modulus of elasticity (N/mm )

Figure 3.1 Relationship between the static modulus and the bending strength.




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                                                5/12
                                                              Strength data of 11 "new" species


                                     160.0                                                                karrie
                                                                                                          European oak
                                     140.0
         Bending strength (N/mm 2)



                                                                                                          cumaru
                                     120.0                                                                angelim vermelho
                                     100.0                                                                vitex
                                                                                                          massaranduba
                                      80.0
                                                                                                          nargusta
                                      60.0
                                                                                                          piquia
                                      40.0                                                                denya
                                      20.0                                                                basralocus
                                                                                                          robinia
                                       0.0
                                             0   5000     10000       15000    20000   25000      30000
                                                                                           2
                                                 Dynamic modulus of elasticity (N/mm )

Figure 3.2 Relationship between the dynamic modulus of elasticity and the bending strength.


                                                              Strength data of 11 "new" species


                                     160.0                                                                karrie
                                                                                                          European oak
                                     140.0
         Bending strength (N/mm 2)




                                                                                                          cumaru
                                     120.0                                                                angelim vermelho
                                     100.0                                                                vitex

                                      80.0                                                                massaranduba
                                                                                                          nargusta
                                      60.0
                                                                                                          piquia
                                      40.0                                                                denya
                                      20.0                                                                basralocus
                                                                                                          robinia
                                       0.0
                                          500           700            900          1100          1300
                                                                 Density (N/mm2)



Figure 3.3 Relationship between the density and the bending strength.

The average values for the species are summarized in table 3.2. The values given for the
bending strength are the values as obtained from the tests. Thus, no correction factors for
depth, moisture content or sample size as specified by EN 384 have been applied to the given




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                                                     6/12
data. The depth of the beams in the test was always around 150 mm and the tests were
performed as four-point bending in accordance with EN 408.


Table 3.2 Average values of 11 “new” species.
Species              Bending strength    Static modulus of         Density
                                         elasticity
                     [N/mm2 ] (cov)      [N/mm2 ] (cov)            [kg/m3 ] (cov)
Angelim vermelho       78.8   (21%)          16000   (11%)         1086 (5%)
Basralocus             58.0   (33%)          17200   (30%)          939 (11%)
Cumaru                 102    (21%)          18300   (16%)         1078 (5%)
Denya                  75.7   (16%)          17000   (30%)          991 (8%)
European oak           42.0   (16%)           9300   (18%)          885 (7%)
Karri                  62.0   (20%)          15500   (17%)          924 (8%)
Massaranduba           110    (14%)          24700   (35%)         1100 (4%)
Nargusta               73.9   (22%)          19900   (22%)          742 (9%)
Piquia                 63.0   (14%)          18600   (31%)          940 (5%)
Robinia                66.0   (24%)          15200   (13%)          740 (6%)
Vitex                  58.0   (18%)          13100   (16%)          908 (10%)

A typical failure mode for a tropical hardwood is shown in figure 3.4. This is failure of a
massaranduba beam on the left and a piquia beam on the right. The piquia beam shows a
combination of compression wrinkles and tensile failure.




Figure 3.4 Failure of a massaranduba (right) and a piquia (left) beam.




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                      7/12
Figure 3.5 Failure in compression of a karri beam.

4. Prediction of bending strength using stress wave analysis

Multiple regression analysis was used to determine a “general” strength model for new
species. In the following table the correlation coefficients are given for the "all data" dataset.

Parameter                    Edyn                    ?                   Estat                 f stat
Edyn                       1.0000000
rho                        0.4347751             1.0000000
Estat                      0.6558195             0.2241474             1.0000000
f stat                     0.7438121             0.4111260             0.5913608             1.0000000

The linear multiple regression equation reads:
 f m = 0.038 Edyn + 0.0008 Estat + 0.0223ρ − 28.2675 with R2 = 0.582 (r = 0.764)                         (2)

A slightly better prediction is obtained when also the logarithm of Estat and Edyn is used as well
as ?2 :

f m = 0.0079Edyn − 84.2096log( Edyn ) − 0.0008Estat + 36.6253log( Estat) − 0.3038ρ + 0.0002ρ 2 + 538.0231
with R2 = 0.615 (r = 0.784)                                                                       (3)

The abovementioned species were used to derive a general regression equation. Regression
equations were derived with the database minus the data of one species and the bending
strength of this species was then predicted. In this way the model was optimized using all
datasets and the sensitivity could be determined to judge the strength of a new unknown
species. Predicted strength is calculated on the basis of a regression equation including static
and dynamic modulus of elasticity, as obtained with the TNO apparatus and density. In
addition, all the data of the species azobé was added to the database. It is outside the scope of
this paper but the system in practice uses a larger database, including some softwood data, as
well as several modification factors for the determination of the actual strength class. The
status of the overall model before it was applied for the first time in practice was even better
than presented here (R = 0.85). The relationship between regressed strength and actual
strength is shown in figure 4.1.




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                                  8/12
Figure 4.1. The regressed bending strength on the basis of the general strength model.

In figure 4.2 the relative mistake is shown on the average bending strength. The ratio given is
the predicted strength divided by the measured strength. Consequently, a value above 1.0 is
an overestimation, while a value lower than one is an underestimation of the strength. The
same is done in figure 4.3 for the characteristic value. These modification factors included
moisture content, depth effect, samples size. The values presented here are just the rough data
from which the characteristic value is determined as m-1.64s, where m is the sample mean
and s the sample standard deviation.

                                1.4

                                1.2
   Ratio predicted / measured




                                 1

                                0.8

                                0.6

                                0.4

                                0.2

                                 0
                                        rri




                                                                                 tex




                                                                                                      ia


                                                                                                               a
                                                         u




                                                                                                                                 ia
                                                                                            sta




                                                                                                             ny
                                                       ar
                                      Ka




                                                                                                    qu




                                                                                                                              bin
                                                                                                                       us
                                                                 o
                                                k




                                                                               Vi


                                                                                a
                                                      m




                                                                                         rgu




                                                                                                           De
                                              oa




                                                              elh




                                                                              ub




                                                                                                                    loc
                                                                                                  Pi




                                                                                                                            Ro
                                                    Cu




                                                                            nd


                                                                                       Na
                                             an




                                                                                                                 sra
                                                            rm




                                                                         ara
                                          rop




                                                          ve




                                                                                                               Ba
                                        Eu




                                                                       ss
                                                        lim




                                                                     Ma
                                                      ge
                                                    An




Figure 4.2 Prediction of the average bending strength




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                                                              9/12
                                2.2
                                 2
                                1.8
   Ratio predicted / measured




                                1.6
                                1.4
                                1.2
                                 1
                                0.8
                                0.6
                                0.4
                                0.2
                                 0
                                        rri




                                                                         u




                                                                                                                          a
                                                                                                                 ia
                                                                                            tex




                                                                                                                                            ia
                                                                                                        ta




                                                                                                                        ny
                                                                       ar
                                      Ka




                                                                                                               qu




                                                                                                                                         bin
                                                                                                                                  us
                                                                                          ba
                                                  k




                                                                      o




                                                                                                      us
                                                                                          Vi
                                                                    m
                                                oa




                                                                                                                      De
                                                                   elh




                                                                                                                               loc
                                                                                                             Pi
                                                                                        du




                                                                                                                                       Ro
                                                                                                    rg
                                                                  Cu
                                                an




                                                                 rm




                                                                                                                            sra
                                                                                                  Na
                                                                                     ran
                                             rop




                                                               ve




                                                                                                                          Ba
                                                                                   sa
                                           Eu




                                                           lim




                                                                                 as
                                                                                M
                                                         ge
                                                       An




Figure 4.3 Prediction of the characteristic value of the bending strength

From figures 4.2 and 4.3 it can be concluded that the characteristic bending strength for
nearly all species can be determined with an accuracy of +/- 20%. At the strength class levels
of most of these hardwoods, generally around D40, D50 of D60, this 20% is about one step in
strength class. Consequently, when the characteristic bending strength class is determined on
the basis of this model, the species could be assigned to a strength class one lower than
predicted and this would be a safe approximation for most species. The one species that
stands out from a negative point if view is basralocus. The sample tested had a very large
scatter as can be seen from table 3.2 and figure 3.3. The cause of this scatter is unknown,
since no specific growth details could be determined visually other than knots. In figure 4.4
the influence of the knot area on the bending strength is shown. Although the bending
strength decreases due to the knots, the knot area is most cases was still relatively small.
                                  Knot area and bending strength of Basralocus
    Bending strength (N/mm )
   2




                                100
                                80
                                60
                                40
                                20
                                  0
                                       0        0.05     0.1       0.15        0.2     0.25       0.3
                                                               Knot area [-]

Figure 4.4 Influence of knot area on the bending strength of basralocus




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                                                                         10/12
As a result of this, the general strength model takes into account model and prediction
uncertainties, sample sizes and the required safety level in structural applications.

5. Conclusions and practical application

Based on a large dataset of 12 species a prediction model for the strength of new "unkown"
species has been developed. The model is able to predict the strength class using static and
dynamic modulus of elasticity and density. The system is currently being used for analysis of
batches of timber from one of the 12 species that are to be used in heavy structural
applications. The first trial was for a batch of cumaru to be used in a sluice door. In figure 5.1
the large beams are analysed on the timberyard. In figure 5.2 the beams already form part of
the frame of the sluice door.




Figure 5.1 Testing of large dimension cumaru          Figure 5.2 Beams in a sluice door frame
beams.


In figure 5.3 and 5.4 the sluice doors are placed in position. The doors are 4.5 meters wide
and 6.6 meters high. The two largest member sizes that were machine graded were 7 meter
long with a cross section of 0.42 x 0.42 meters. The second largest members were seven
beams of 5 meters long with a cross section of 0.4 by 0.4 meters. Although the whole batch
was acquired as 'cumaru', two distinct colors could be distinguished, one yellowish and one
brownish and questions were raised whether they were indeed all of the species 'cumaru'. The
stress wave grading, however, showed no differences in strength between these members.




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                             11/12
Figure 5.3. Sluice door hoisted into position.      Figure 5.4 Sluice door in its final position


The system is in trial state to be used whenever a batch of a new species enters the market.
The non-destructive testing method can be used to specify a safe characteristic bending
strength value for structural applications of a batch of that particular "unknown" species.

Literature

Blass, H.J., Van de Kuilen, J-W.G., The characteristic strength of azobe, TNO Report B-91-
1124, 1991.

Van der Linden, M.L.R., Van de Kuilen, J-W.G., Blass, H.J., Application of the Hoffman
Yield Criterion for load sharing in timber sheet pile walls. Pacific Timber Engineering
Conference pp. 1994

Van de Kuilen, J-W.G. Development of a general determination method for the strength of
timber. Phase 1. TNO Reports 1999-LBC-R7027 / R7028/2 / R7029 / R7030/2. (in Dutch)

Van de Kuilen, J-W.G., Ravenshorst, G.J.P. Development of a general determination method
for the strength of timber Phase 2. TNO Reports 2000-CHT-R101 / R102 / R156 / R157 /
R158 / R172 / R173 / R174 / R178 / R179 (in Dutch).




J-W.G. van de Kuilen, G.J.P. Ravenshorst                                                           12/12

				
DOCUMENT INFO