Procedia Engineering 48 ( 2012 ) 409 412

1877-7058 2012 Published by Elsevier Ltd.Selection and/or peer-review under responsibility of the Branch Ofce of Slovak Metallurgical Society at
Faculty of Metallurgy and Faculty of Mechanical Engineering, Technical University of Koice
doi: 10.1016/j.proeng.2012.09.533
MMaMS 2012
Non-destructive tests of modulus of elasticity for the glued laminated
timber beams
Lenka Melzerov
a
*, Petr Kuklk
b
a
CTU in Prague, Faculty of Civil Engineering, Thkurova 7, Prague 6, 16629, Czech Republic
Abstract
Twenty real dimensions beams from the glued laminated timber were tested in our previously works. Twenty advanced FE models were
created precisely according to tested beams. Input files for FE models are lengths of segments and local moduli of elasticity. The segment
is part of lamella between two finger joints. Each local modulus of elasticity was obtained via non-destructive penetration test. The output
for comparison between real beam and FE model is displacement in half span. The quality of input data file from experiments is very
important for the good agreement between real tested beams and FE models. In advanced FE models is described distribution of local
moduli of elasticity via distribution function. The solution is based on the LHS. Accuracy of each distribution function is dependent on
the number of measured local moduli of elasticity. In presented work was used probabilistic approach for determination of corresponding
number of penetration tests as function of segments lengths. Results of this analysis will be used in the latter series of bending tests of
new real dimensions beams and corresponding advanced FE models.
2012 The Authors. Published by Elsevier Ltd.
Selection and/or peer-review under responsibility of the Branch Office of Slovak Metallurgical Society at Faculty of Metallurgy and
Faculty of Mechanical Engineering, Technical University of Koice.
Keywords: Glued laminated timber, Non-destructive tests, FE models
Nomenclature
E modulus of elasticity (MPa)
w beam displacement (mm)
mean of E
standard deviation of E
1. Twenty real tested beams
This article is focused on the twenty beams from the glued laminated timber with the real structure dimensions. An
illustrative example of a glued laminated timber beam appears in Figure 1 (a). In general, such a beam consists of an
arbitrary number of segments glued together over their entire area to form a layered structure. As seen in Figure 1 (b) the
present example assumed eight layers (lamellas). Individual layers in the laminated beam are usually not compact but are
subdivided into segments each having a random length. This arises from cutting out sections of lamellas containing large
knots and other possible flaws to improve the overall quality of the beam. Different parameters were monitored during four-
* Corresponding author. Tel.: +420224355438; fax: +4202-2431-0775.
E-mail address: melzerov@fsv.cvut.cz.
Available online at www.sciencedirect.com
2012 Published by Elsevier Ltd.Selection and/or peer-review under responsibility of the Branch Ofce of Slovak Metallurgical Society at
Faculty of Metallurgy and Faculty of Mechanical Engineering, Technical University of Koice
410 Lenka Melzerov and Petr Kuklk / Procedia Engineering 48 ( 2012 ) 409 412
point bending tests including relation between displacement of beam and its loading. All beams were tested up to their
destruction. Loading was applied incrementally with step equal to 4kN. Maximal forces prior to failure were in the range
from 30 to 60 kN. Uniform maximal loading level for all twenty beams was detected, because for each beam came damage
in different loading level. This uniform loading level is 24 kN for each from two forces in real bending test and 60 kN/m in
2D FEM model, because length of loading is 0,4 m (Fig.2) [1, 2].
(a) (b)
Fig. 1. Illustration of glued laminated timber (a) and cross-section of tested beam (b).
Fig. 2. Illustrative example of FEM model.
2. Twenty advanced FE models
The advanced FEM models employ probabilistic simulations performed in the framework of LHS method. In the light of
this, each segment is assigned Youngs modulus with a corresponding probability density function. In all cases the Gaussian
distribution with the given mean and standard deviation is assumed as seen in Figure 3.
Fig. 3. Illustration of the input data used in the LHS method.
segment
finger joint
411 Lenka Melzerov and Petr Kuklk / Procedia Engineering 48 ( 2012 ) 409 412
The associated distribution function is then utilized to generate individual samples. In the present study the distribution
function was split into 100 intervals to randomly select a single value
k
E as schematically shown in Figure 4 (a). This result
is in accord with the LHS method based on 100 strata. The resulting map of realizations is constructed such as to comply
with a statistical independence of elastic moduli from segment to segment. Note that selecting lamellas to form a beam is
conducted in a totally random manner. Results can be statistically evaluated and fitted to the selected probability density
function as illustrated in Figure 4 (b) for the Gaussian distribution.
(a) (b)
Fig. 4. Principle of selecting the k-th sample in the LHS method (a) Example of the Gaussian probability density function of deflection for the selected
beam (b).
3. Comparing results from FEM and experiments
To compare individual approaches (experiment, deterministic and probabilistic modeling) a single value given by the
averages obtained from 100 samples will be adopted. It might be, however, expected that a better agreement with
experimental results will be obtained with improved probabilistic data of input parameters conditioned by considerably
more measurements in individual segments (recall that only four measurements are presently available for each segment).
Probability of not exceeding a certain limit deflection is even more important than a simple mean, although not examined,
which might provide further insight in the behavior of such structures. The variations of maximal deflections in Figure 5
showing the comparison with the averages delivered by the probabilistic analysis. Clearly, when comparing only averages
the difference between deterministic and probabilistic modeling is almost negligible [3].
Fig. 5. Comparison of measured and calculated deflections.
4. Non-destructive tests of modulus of elasticity
In our previously articles were demonstrated that modulus of elasticity in fibres directions (E
x
) is for beam displacement
significant, but on the distribution of finger joints is displacement practically independent. For our experiments is necessary
the knowledge of distribution of modulus of elasticity in fibre direction as precisely as possible [5]. Measurements of E
x
0
0,1
0,2
0,3
0,4
0,5
0,6
1
5
,
2
9
1
5
,
6
4
1
6
,
0
0
1
6
,
3
5
1
6
,
7
1
1
7
,
0
6
1
7
,
4
2
1
7
,
7
7
1
8
,
1
2
1
8
,
4
8
1
8
,
8
3
1
9
,
1
9
1
9
,
5
4
1
9
,
9
0
2
0
,
2
5
2
0
,
6
0
2
0
,
9
6
2
1
,
3
1
2
1
,
6
7
2
2
,
0
2
2
2
,
3
8
D
e
n
s
i
t
y
w (mm)
16
17
18
19
20
21
22
1 2 3 4 3 6 7 8 9 1011121314131617181920

Measured
ueLermlnlsLlc
LPS
[n]
412 Lenka Melzerov and Petr Kuklk / Procedia Engineering 48 ( 2012 ) 409 412
were realized for twenty beams in each segment of each lamella on four places. Final file of 1448 values of E
x
is for
statistical research sufficient. The rightness of measurement was verified independently via strain gauges and displacements
sensors on the several places of each beam during loading tests. The relation between stress and strain from the stain gauges
during the loading tests is linear [4].
For greater accuracy of the advanced FE models with variable modulus of elasticity is necessary to increase the number
of measurement of local E in segments. Presented methodology is independent on the sample size of E, but in our future
work we will test greater number of local E (1400) in each from small number of GLT beams (3). Measured E will be
distributed regularly in each beam. This approach removes the dependence of measured E accuracy on the segment length.
Both files of E (old and new) will be used together according to Bayesian updating (Fig. 6).
Fig. 5. Bayesian updating of E.
5. Conclusions
The proposed approach is mainly suitable for extreme structures. Extreme structures have large differences between
material parameters of timber. The standard design is more conservative when compared to the present approach.
Acknowledgements
This outcome has been achieved with the financial support of the Ministry of Education, Youth and Sports of the Czech
Republic, project No. LD12023 advanced methods for design, strengthening and evaluation of glued laminated timber.
References
[1] Melzerov, L., Kuklk, P., 2010. "Variability of Strength for Beams from the Glued Laminated Timber", Proceedings of the 48th International
Scientific Conference on Experimentalni Analyza Napeti 2010 Experimental Stress Analysis, pp. 257-260. ISBN 978-80-244-2533-7.
[2] Melzerov, L., Kuklk, P., 2010. Statistical Research of Mechanical Properties of Glued Laminated Timber Beams. Metallurgy 49, p. 376-380. ISSN
0543-5846.
[3] Melzerov, L., Kuklk, P., ejnoha, M. 2011. "Variable Local Modulus of Elasticity as Inputs to the FEM Models of Beams from the Glued Laminated
Timber", 2nd International Conference on Material Modelling, pp. 213. ISBN 978-2-911256-61-5.
[4] Kuklk, P., Melzerov L., 2011. Kompozitn materily na bzi deva, esk vysok uen technick v Praze, Praha p.76. ISBN 978-80-01-04958-7.
[5] Plach, T., Tesrek, P., Wilczynsk, A., Padevt, P. 2010. "Experiment in Real Conditions: Mechanical Properties of Gypsum Block Determined Using
Non-destructive and Destructive Methods", Proceedings of the 3rd WSEAS International Conference on Engineering Mechanics, Structures,
Engineering Geology/International Conference on Geography and Geology, pp. 418-423. ISBN 978-960-474-203-5.
Existing file of E
File of E for the new
beams
Combination of
both sets of E
including
reliability matrix