Professional Documents
Culture Documents
Experiment
0 I J F
N
The scheme of the experiment was projected by Prof.
Shigematsu, 12'13now deceased, for investigating the relation E
between wood quality and growth in Japanese larch trees. 2-2
The experiment on the mechanical properties of Japanese
larch square-sawn timbers were completed at Shinshu Uni-
versity, Forestry and Forest Products Research Institute,
and Nagano Prefectural General Forestry Research Center -4 ~/ J
in Japan. The static destructive bending tests were done
-4 -2 0 2 4
with span lengths of 270 cm by third-point concentrated
loading. The dimensions of the bending specimens were Standard nomal variate
nominally 12.0 cm thick, 12.0 cm wide, and 300 cm long.
The modulus of elasticity (MOE) and modulus of rupture
(MOR) were measured and recorded for each specimen. Fig. 1. Distribution of modulus of elasticity (MOE) by Gauss probabil-
ity paper, y, normalized MOE
After the bending tests, short columns for compression tests
parallel to the grain were cut from the nonfailure portion of
each bending specimen. The length of the column was
54 cm, and the slenderness ratio was 17. The compressive the range from the mean plus the SD to the mean minus the
strength (CS) for each column was obtained using static SD.
compression tests. The details of the experiments are found The relation between MOE and MOR is shown in Fig.
in the previous report; 14the data sets for MOE, MOR, and 2a. Note that the regression line of M O R on MOE was also
CS for each square-sawn timber were used for the following described with 5% and 95% limit lines. The regression line
analysis. We focused on the M O R distribution, and the CS was expressed as M O R = 4.42 MOE + 2.86 (n = 287, r =
data were used for comparing the MOR data. 0.585, SE -- 9.45), where n is the number of specimens, r is
the correlation coefficient, and SE is the standard error. The
additional axes denote normalized MOE (y) and M O R (b),
respectively. The y values was simply defined as y = (MOE
Results and discussion data - mean MOE)/(SD of MOE); b as similar to y. Most of
the plots were within a range of +2 o f y and b. The horizon-
Experimental results tal broken line denotes the 15th percentile level of the
M O R distribution. Foschi et al. 16 used 2P-Weibull fitted to
The mechanical properties of specimens are shown in the test data truncated at the 15th percentile level for reli-
Table 1. The mean and standard deviation (SD) for the ability evaluation of wood structures. It may be presumed
MOE were 9.44 GPa and 1.54 GPa, respectively. Figure 1 that their chief concern about M O R distribution is within
shows the distribution of MOE by Gauss probability paper. the range below the line in Fig. 2a. Similar to MOR, Fig. 2b
It may be assumed that the M O E distribution can be ex- shows the CS distribution; c is the normalized CS; regres-
pressed as a normal distribution, as we reported. 15Thus the sion line, CS = 1.70 MOE + 15.12 (n = 287, r = 0.529,
MOE values of almost 70% of specimens should be within SE = 4.20).
352
90 73,,0 -~
/
--~3
O O ~ 0
e.'~ /
"- 60
n
/ , / ~ / '
-i 0 -.~
30
/ o
/ •-3
0
0 5
[
10
MOE (GPa)
15
I I
20
I ]I ]I[ iV
!
a
Class
Jo
Y
90 ~3, , O, , 3
o6 ~ ----t
.-, 60
12. .----4
v 9 9 9 3 "o )---q
. Ir
P, 30 ffJ
~2
I I I ]
0
0 5 10 15 20 0 2 4 6 8
MOE (GPa) ka: Classified by MOE
b
Fig. 2. Relation between MOE and strength, a Modulus of rupture Fig. 3. Ratio (k~) of MOR and CS to Young's modulus (MOEt of
(M O R ), open circles, h Compressive strength (CS): filled circles, y,b,c, specimens classified by Young's modulus and strength. Class
normalized MOE, MOR, CS, respectively. Dash-dot lines, 95% limit bondaries: p = 0.15.0.5.0.85. respectively; p. cumulative probability, a
lines; dash-dot-dot lines, 5% limit lines; horizontal broken line (a), 15th k~, classified MOE (diamonds) and strength ~squares); open andfilled
percentile level plots, MOR and Cs. respectively, b Effects of classification: blank and
filled circles. MOR and CS. respectively
Formerly, the strength/elasticity ratio 17 was usually ues at each M O R and CS class were obtained similar to the
adopted for evaluating mechanical properties of commer- M O E classes. If the correlation between M O E a n d M O R
cial timbers defined as k, = aa/E'a/1000, where a and E were strong, the difference between two ka values classified
denote strength and elasticity, and the subscript a means by M O E and M O R should be small at each class. Figure 3a
average. We attempted to compare M O R and CS with the shows the ka values at each class. It was clear that k~ values
ratio (ka) at each class of MOE. The four M O E classes were at each M O E class in M O R was almost constant c o m p a r e d
divided by the three boundaries with cumulative probabili- to the ka values in CS. The relations between k, values at
ties of 0.15, 0.50, and 0.85, respectively. In addition, ka val- M O E and strength classes are shown in Fig. 3b. Thediffer-
353
ences b e t w e e n M O R and CS were obvious. In other words, the same M O E value, b o t h were selected; then the selected
the d e p e n d e n c e of strength distribution on Y o u n g ' s m o d u - numbers at E~, E2, E3, and E4 were 10, 11, 12, and 11,
lus levels m a y vary b e t w e e n b e n d i n g and compression due respectively.
to failure types: brittle or ductile fracture. Next, these data were fitted to 3P-Weibull, and the ob-
tained p a r a m e t e r s are shown in Table 2. The relation be-
tween M O E and the p a r a m e t e r s of 3P-Weibull was
A p p l i c a t i o n of target-E m e t h o d
expressed a p p r o x i m a t e l y as exponential equations for each
parameter.
T h e concept of the t a r g e t - E m e t h o d is simple. W h e n you
Finally, simple simulation was done as following: E a c h
know the a p p r o x i m a t e distribution of Y o u n g ' s modulus in
replication was r e p e a t e d 5000 times, and we o b t a i n e d 5000
your objective materials, sampling for destructive tests lim-
data sets for M O E . M O R .
its materials at some particular E-levels. W e expected that
the whole strength distribution estimated by this m e t h o d 1. A s t a n d a r d n o r m a l r a n d o m n u m b e r was chosen to calcu-
might reflect the d e p e n d e n c e of strength distribution on late M O E .
Y o u n g ' s modulus. Figure 4 illustrates the difference in 2. A uniform r a n d o m n u m b e r (rl) from 0 to i was chosen to
sampling a m o n g the regression model, the p r o o f loading calculate strength.
approach, and the target-E method. T h e p r o o f loading ap- 3. In the regression model, rl was t r a n s f o r m e d to the stan-
p r o a c h and the target-E m e t h o d n e e d smaller sampling dard n o r m a l r a n d o m n u m b e r (r~), and the M O R value
sizes than the regression model. W e examined the efficiency was o b t a i n e d by linear regression with the above M O E
of the t a r g e t - E m e t h o d for estimating strength distribution value and SE • r2 using the a b o v e m e n t i o n e d experi-
through simple simulation. mental regression equation.
A t first, four t a r g e t - E levels were set as y = - 1 . 5 , - 0 . 5 , 4. F o r the target-E m e t h o d , the three p a r a m e t e r s of 3P-
0.5, and 1.5; they were i n t e n d e d to r e p r e s e n t the four above- Weibull were calculated with the a p p r o x i m a t e equations
m e n t i o n e d M O E classes. W e selected 10 specimens with as shown in T a b l e 3 using the M O E value in item 1
M O E values nearest to each target-E: E 1 = 7.1, E2 -- 8.7, (above). The M O R value was o b t a i n e d by the 3P-
E 3 = 10.2, and E 4 = 11.8 GPa. W h e n two specimens h a d Weibull with rl in item 2 (above).
MOR
EI 0.296 24.2 13.8 7.2
E2 0.284 27.6 16.8 8.7
E3 0.352 32.6 15.4 10.2
E4 0.256. 33.8 22.9 11.7
CS
E1 0.196 14.9 12.9 7.2
E2 0.150 15.2 15.3 8.7
E3 0.263 21.1 14.4 10.2
E4 0.176 19.6 18.3 11.7
k, m, x0, shape, scale, and location parameters of 3P-Weibull,
Fig. 4. Concept of target-E method comparing to other models respectively; CS, compressive strength
CI Ca r2 Cl C2 r~ C1 C2 r2
MOR 0.340 -0.015 0.048 14,1 0.078 0.948 6.9 0.094 0.721
CS 0.166 0.016 0.017 8.6 0.075 0.693 8.1 0.066 0.768
Each parameter of 3P-Weibull was expressed as I / k , m, or x0 = C~eC2X;x, MOE; r 2, determination
coefficients
354
Table 4. Basic statistics of simulated results
Propert No. Mean SD CV
..!.. J o
MOE (GPa) 5000 9.44 1.50 15.9
MOR (MPa)
Reg. 5000 44.3 11.6 26.3
Tar. 5000 43.4 10.4 23.9
CS (MPa)
Reg. 5000 31.0 5.0 16.0
Tar. 5000 30.8 6.2 20.0 y y d
Reg. and Tar. denote simulation results by regression model and
target-E method, respectively m i
. i . i~~J ~. .i. .
CS (MPa) i| |l
50th 30.4 31.0 (1.02) 30.5 (1.01)
10th 25.5 24.6 (0.96) 23.0 (0.90)
5th 23.4 22.9 (0.98) 21.3 (0.91)
Reg. and Tar. denote simulation results by regression model and
target-E method, respectively. Values in parentheses are the ratios of
simulated values/experimental data
c y y f
0.15 60
>, 50
0.1 o>' 40
O
g
t,.
Q.. g
O 30
>
...I_1
0.05 20
E
O lO
0 0
0 10 20 30 40 -4 -2 0 2 4
MOle? (MPa) b
40 60
50
30
40
13,..
20 30
>o 20
10
#: lO
0 I _ _ I I 0
0 5 10 15 20 -4 -2 0 2 4
C
MOE (GPa) b
b
Fig. 7. Strength distribution at each target-E level, a Normalized MOR
Fig. 6. MOR distributions truncated at the 15th percentile (a) (y). b Filled plots, normalized CS (c); squares, E~; diamonds, Ea;
and regression lines of MOR on MOE in the lower tail (b). Circles, triangles, E3; circles, E 4 ; s e e Table 2
experimental; heavy curves, regression model; thin curves, target-E
model
Conclusions
sibly with a more effective method than the target-E
method for each species,
We propose the target-E method for estimating M O R dis-
The target-E method is successful for estimating M O R
tribution in Japanese larch timber as an alternative method
distribution in Japanese larch timber, although some prob-
to the conventional regression model. The comparison of
lems remain for its practical use, for example social machin-
simple application of a regression model and the target-E
ery for collecting more M O E data, a more effective setting
method to experimental data led to the following results.
of target-E, a more accurate interpolation method of M O R
We also compared M O R and CS for contrast.
distributions between each target-E, and so on. It should be
noted that only 15% of the data from all specimens was 1. The M O E - M O R or M O E - C S regression lines obtained
needed for the target-E method. by simulation with a regression model and the target-E
356
method were near the regression lines obtained from 6. Hayashi T ,1989) Performance prediction of wood laminates (in
Japanese) Mokuzai Gakkaishi 35:1144-1153
experimental results. 7. Ang AH-S. Tang WH 11977l Probability concepts in engineer-
2. The lower tail of MOR distribution estimated by the ing planning and design (in Japanese) {translated by Ito M.
target-E method was a better approximation of the ex- Kameda H). Maruzen. Tokyo, pp 282-293
8. Takeda T. Hashizume T t1999} Differences of tensile strength
perimental data than that obtained by the regression
distributions between mechanically high and low grade Japanese
model, although only 15% of the data of all specimens larch lumbers I. J Wood Sci 45:200-206
was needed for the target-E method estimation. 9. Takeda T. Hashizume T 41999) Differences of tensile strength
3. The regression model, on the other hand, expressed distributions between mechanically high and low grade Japanese
larch lumbers II. J Wood Sci 45:207-212
CS distribution more accurately than the target-E 10. Yakeda T. Hashizume T (2000) Differences of tensile streng:b
model. distributions between mechanically high and low grade Japanese
larch lumbers III. J Wood Sci 46:95-101
11. Madsen B (1992) Structural behaviour of timber. Timber Engi-
Aeknowledgments We thank Dr. Takashi Nakai, Mr. Toshinari
neering, North Vancouver, pp 37-49
Tanaka, Mr. Hirofumi Nagao, Dr. Fukio Takei, Mr. Takahisa Yoshida,
12. Shigematsu Y (1990) Wood quality of larch in relation to its
Prof. Tatsuo Nakano, and Prof. Morihiko Tokumoto for their advice
growth. 1 (in Japanese). Mokuzai Kogyo 45:445-451
and suggestions. 13. Shigematsu Y 11991) Wood quality of larch in relation co its
growth. 2 (in Japanese) Mokuzai Kogyo 46:9-16
14. Tokumoto M. Fakeda T. Nakano T. Hashizume T, Yoshida T.
Takei F. Nagao H, Tanaka T. Nakai T (1997) Mechanical proper-
References ties of full-sized square lumbers of karamatsu (m Japanese). Bull
Shinshu Univ For 33:75-145
15. Takeda T. Tokumoto M. Nakano T. Hashizume T. Nagao H (1998)
1. Sato H, Nakahara Y (1967) The complete Japanese joinery (trans- The fifth percentile estimate of bending, longitudinal compressive.
lated into English by KP Nit). Hartley & Marks, Vancouver, pp and tensile strengths of square sawn larch timbers from Shinshu t in
221-251 Japanese). Mokuzai Gakkaishi 44:170-177
2. Iijima Y, Nakai T (1990) Species effect on MOE-MOR relation of 16. Foschi RO. Folz BR. Yao FZ {1989) Reliability-based design of
square sawn lumber. Presented at the 1990 international timber wood structures. Structural research series report No. 34. First
engineering conference, Tokyo, pp 842-846 Folio Printing. Canada, pp 20-22
3. Hayashi T (1998) Engineered wood (in Japanese). Nihon Mokuzai 17. Tsuiii S, Sugiyama H (1961 I On the strength properties and the
Shinbunsha, Tokyo, pp 10-13 allowable unit stresses of commercial timbers (in Japanese). Trans
4. Harada T, Suzuki S (1998) Rapid learning the revised Japan archi- Architect Inst Jpn 67:67-74
tectural standard (in Japanese). Ohmusha, Tokyo, pp 1-15 18. Natrella MG (1963) Experimenta] statistics. Handbook 91.
5. Sobue N (1993) Nondestructive testing of wood (in Japanese). US Department of Commerce Technology Administration.
Mokuzai Gakkaishi 39:973-979 Washington, DC, pp 5.1-5.46