5th World Conference on Timber Engineering 1998 August 17-20, Montreaux, Switzerland. Reproduced with the publisher's autorization. ©1998, Presses polytechniques et universitaires romandes, Lausanne, Suisse.

TABLE OF CONTENTS

ABSTRACT INTRODUCTION PRINCIPLES OF THE NON DESTRUCTIVE TESTS ROUND WOOD X-RAYS GRADING MACHINE RESULTS AND ANALYSIS CONCLUSIONS BIBLIOGRAPHY

ABSTRACT

Some new NDE techniques have been used and developed in CTBA for grading small round wood of French Douglas specie. The methodology is used to evaluate their mechanical characteristics by both X-rays and Sylva-test. The results are discussed. They show that it is possible to define directly geometric dimensions, local density of clear wood, Knot Volume Ratio (KVR) along the piece and young modulus values. The KVR curve allows to find the higher value for nodosity, to assess the distance between whorls and to measure the mean quantity of knots on various zones. The clear wood density curve gives an estimation of the mean value of clear wood density and the general density curve give the whole density of the piece. The predicted critical defect has been found in the piece for any considered test, and a better correlation between destructive evaluation and NDE techniques is obtained.

1. INTRODUCTION

In many European countries there is surplus of small diameter (8-15 cm) round timber, which is harvested when thinning forest. A joint European (FAIR) project was started 1996 in order to develop the use of small diameter round timber in construction. The objective of the program is to develop structural systems in which small diameter round timber can be used and thereby create new market for them. In this project, CTBA aims to test the quality of small round Douglas fir from France and develop a strength grading method for roundwood.

2. PRINCIPLES OF THE NON DESTRUCTIVE TESTS

2.1 Description of CTBA X-rays densimeter
Two Non Destructive Tests have been used in order to control roundwood quality, X-rays and sylva-test techniques. As Sylvatest [1] [2] is a well-known method, we just describe the X-rays technique. The principle of a density measurement by means of X-rays (GAMMA, micro-wave) consists in measuring the lightening of photons with a material of continuous thickness [3]. The specimen goes through a X rays radiation fan beam (20 kV, Tungstene) with an initial number of counts I₀. The number of count / resulting from the specimen absorption is detected by means of 24 Nal cells disposed around a circle of 520 mm. The difference between I₀ and I is correlated with the specimen density. This lightening is worded as follows :

I / I₀ = e ^{µ r e}         (1)
Where: I₀ is the Number of impact per time unit on the material and is known after a calibration phase without samples ; I is the Number of impacts through the material and is calculated during the trial ; is the Coefficient of massive absorption and varies according to the material nature ; r is the surface density (gr/cm²) and is calculated after the measurement for a continous µ. The distance between the piece of wood and the source should be adjusted according to the thickness in order to obtain the best accuracy of density. The X-rays generator has the advantage of a low supplying, which allows us to limit the radioprotection strengths.

2.2 Calculation of non-destructive parameters of grading
The data base software of C.E.A. can give us a date file ASCII of 93 columns on 400 lines. Each information corresponds to a measurement of surface density which is easy to put into density. The knots may be thus defined with X-rays technics since the knot density has an average value of 2,5 higher than a clear wood. With this software, we have the possibility to define local density. 86 variables (location and gradient) were selected to find out the best correlations between the physical and mechanical characteristics of wood specimens submitted to a bending test. The software can generate a bi-dimensional signal r(x,y) of the specimen density, from an initial datafile of 400 rows and 93 columns (38 400 values):

X-resolution:	The speed of the conveyors has been adjusted at 50 m/min.
Y-resolution:	Each line contains information recorded from the 24*4 cells detectors according to the combined position of the piece and the X ray source.

The end-resolution is ~ 5 mm² for a single data value. The previous density r(x,y) is transformed into two signals r(x) and r(y) according to the following formulae:

     (2)

2.3 Experimental protocol
The processing of structural wood grading developed in CTBA is decribed in the following figure and allow us to set up a wood grading.

Fig 1:Schema of the grading machine available in CTBA. [1]

For sawn timber, the X-rays densimeter sets up a bi-dimensional cartography which allows us to calculate its density point by point. With this cartography a program calculates densities and numbers of knots. In a first stage, a nodosity parameter allows us to assess the location of the more likely bending failure point [5]. With this value, the software forecasts the location of the center of the bending trial. At this moment it calculates several densities and knot ratios on the 20 H (20 times the heigth which corresponds to the piece height after being cut for a bending test).

3. ROUND WOOD X-RAYS GRADING MACHINE

The applied X-rays grading method to sawns was adapted to logs. This processus allows to test round woods with the same material than that used for sawnwood. On the other hand, the software of data processing was adapted to take into account the modification of data (circular section and no more rectangular section) and new grading criteria.

3.1 Specimen description
Specimen for this paper are calibrated Douglas fir from trees of 10 to 17 years. The diameter is 120 mm in four meters length. Logs were barked then turned into a cylindrical shape. Because of their shape the characterization of pieces only requires one dimension (diameter) instead of two (thickness and length) for sawn. Furthermore there is always heart wood in pith and near the middle. So the mechanical characteristics are more homogenous than for sawns taken from several areas of pith. But some cracks may appear along pieces while drying which may reduce the bending characteristics but have no effect on tensile test.

3.2 Measurements
The X-rays machine performes a numerical radiography of wood. It provides a cartography in which each point is linked to wood mass for this coordinate.

Fig 2: Map of the round density

For a first treatment, it is possible to measure pieces section in the whole length. Once this characteristic is known, the cartography allows to produce a curve of wood density centimeter by centimeter, as if we cut the log into round of 1 cm and as if we measured density of each slice.

Fig 3: Mean density along the beam for a round wood.

3.3 Lineous profile of density
The previous curve is then studied in order to provide on one hand the graph of wood density without knot and on the other hand the KVR graph (Knot Volume Ratio) which corresponds to the knots density split up by the whole density.

Fig 4: Variation of the density of clear wood and KVIR parameter along the round wood log.

The X-rays grading have some advantages comparing to the visual grading :

• Measurements of density are more precise (the error mean value is around 4kg/m³).
• The whole calculation is quicker than the measurement and location of KAR.
• The location of main defect is reliable (the location of bending failure is predicted for 96% of cases in a maximal nodosity location).
• The values of wood without knots and nodosity allows a possibility to set up a simulation of piece in finite elements method.
The X-rays grading allows us to automate the grading and to increase the productivity. Besides more important pieces of information allows to refine grading. The measurement of KAR and X-rays grading could consequently be used as a basis to create a grading standard for round woods. Furthermore, we have to talk about our ressources in round woods simply barked, which cost is still weaker. We should not neglect it. Despite their esthetical aspect by comparison with that of calibrated wood, their general characteristics are as good as those of calibrated woods. So we can easily plan their use in forestry and country areas. The main obstacle to their use is their shape which is not a cylindrical shape. This deals with the problem of chain of non destructive data and their treatment. However, with a low numerical treatment, it is possible to obtain the same type of information for non-calibrated logs. Finally, the technics developped in an European contract arouse the interest of the decision-makers for other species (Larch ... ), justify the work performed by CTBA in this area and confirm the advantages wood chain may benefit from.

4. RESULTS AND ANALYSIS

4.1 Sample description
The set of sampling contains 180 beams of douglas fir. All considered pieces are calibrated in a 120 mm diameter. All specimen have been broken in a 4-point Bending machine in order to mesure MOE and MOR. The following table describes the global characteristics of the sample.

	MOR MPa	MOEL 12 GPa	MOEG 12 GPa	KAR %	moisture content %	mv 12% Kg/m3
mean	52.5	11.1	9.4	24.0	12	442
5% value	36.8	7.7	7.3	12.0	10.3	367
st.dev.	9.9	2.5	1.3	8.7	1.1	51

Table 1: Sample description

4.2 Methodology
To evaluate the ability to grade of a parameter P, we use the following methodology:

1-	The beams are sorted out by P.
2-	The sampling set is sub-divided in 3 sub-groups of same size A lower grou which contains the 60 beams whose the parameter P has the lowest values A medium grou which contains the 60 beams whose the parameter P has the medium values A higher group which contains the 60 beams whose the parameter P has a higher values
3-	For each sub-group we mesure the whole mean characteristics.

With this methodology, we can see if a parameteter has some influence on another. And, if it is the case, we can in addition see if this correlation is linear or not.

4.3 Results and Analysis

4.3.1 Optimal grading.: sample graded by MOR
MOR is the most important characteristic in building. We can learn by the optimal grading [6] that the best mean value of a sub-group is equal to 63 MPa and the lower mean value 42 MPa. Then, all others type of grading can be compared to the optimal grading to see if it is possible to do better. Table 2 also shows the influence of the increasing MOR on all parameters.

	MOR Mpa	MOEL 12% GPa	MOEG 12% MPa	MV 12% Kg/m3	KAR %	US Speed m/s	US modulus GPa	KVR %	predicted MOR
low	42	10.0	8.4	422	26	4746	9.6	12.6	49.8
medium	52	10.8	9.4	439	25	4832	10.3	13.4	51.4
high	63	12.5	10.5	466	23	5046	11.9	13.6	56.3

Table 2: Optimal grading: sample graded by MOR

4.3.2 Sample graded by MV 12%
The increasing of the MV 12 % value has an influence on the global modulus (MOEG 12 %) but not on the local modulus (MOEL 12 %). This is due to the fact that the local modulus depends in fact on both knots and density. Table 3 shows that the increasing of the density lead to the increasing of Knot Area Ratio (KAR). Then, the two opposit effects are canceled. So we can say that the parameter MV 12% seems to be a good parameter to find the higher sub-group (MOR=58).

	MOR MPa	MOEL 12% GPa	MOEG 12% MPa	MV 12% Kg/m3	KAR %	US Speed m/s	US modulus GPa	KVR %	predicted MOR
low	49	11.1	87	385	20	4896	9.3	10.7	48.8
medium	51	10.7	9.3	446	26	4830	10.5	13.7	51.6
high	58	11.6	10.3	495	27	4898	12.0	15.1	57.1

Table 3: Sample graded by MV 12%

4.3.3 Sample graded by KAR
The increasing of KAR has not a significance influence on modulus of elasticity (MOE local or global). Table 4 also shows that KAR parameter has a very low correlation with the MOR.

	MOR MPa	MOEL 12% GPa	MOEG 12% MPa	MV 12% Kg/m3	KAR %	US Speed m/s	US modulus GPa	KVR %	predicted MOR
low	54	11.5	9.4	420	15	4924	10.2	11.0	52.3
medium	53	11.5	9 5	443	24	4820	10.4	13.0	52.5
high	51	10.4	9.4	464	34	4881	11.1	15.5	52.7

Table 4: Sample graded by KAR

4.3.4 Sample graded by US speed
The ultrasonic wave method (Sylvatest) is used because of is good correlation with the modulus of elasticity. And, we can see in table 5 that a grading with the US speed parameter gives approximatively the same results as the optimal grading for both MOE_Local and MOE_Global. The US speed parameter is quite good to predict MOR, as MOR and MOE_Local are correlated.

	MOR MPa	MOEL 12% GPa	MOEG 12% Mpa	MV 12 Kg/m3%	KAR %	US Speed m/s	US modulus GPa	KVR %	predicted MOR
low	48	10.1	8.7	444	24	4433	8.8	13.2	51.8
medium	51	11.3	9.4	439	26	4939	10.7	13.2	51.2
high	58	12.0	10.2	443	24	5252	12.2	13.2	54.5

Table 5: Sample graded by US speed

4.3.5 Sample graded by US Modulus
UltraSonic modulus is predicted with the US method and using the theory of elasticity :

US modulus = (MV 12%)*(US speed)² (3)
The grading is better than with the US speed. But the mean value of the modulus predicted by the US modulus isn't equal to the real two modulus global and local. So we are in the case of the result of a good grading by a wrong predictor.

	MOR MPa	MOEL 12% GPa	MOEG 12% MPa	MV 12% Kg/m3 3	KAR %	US Speed m/s	US modulus GPa	KVR	predicted % MOR
low	47	10.1	8.4	409	23	4529	8.4	11.7	49.1
medium	50	11.0	9.1	440	24	4932	10.6	12.9	52.2
high	60	12.2	10.8	478	27	5163	12.7	15.0	56.2

Table 6: Sample graded by US Modulus

4.3.6 Sample graded by KVR
The KVR gives nearly the same results as KAR and there is a good correlation between the two parameters. But with the X-rays grading machine we can mesure both KVR and density. So the X-ray grading machine give more various parameters and are quite interesting for this reason.

	MOR MPa	MOEL 12% GPa	MOEG 12% MPa	MV 12% Kg/m3	KAR %	US Speed m/s	US modulus GPa	KVR %	predicted MOR
low	51	11.2	9.0	405	19	4961	10.0	9.7	50.6
medium	52	11.1	9.4	446	24	4773	10.3	12.9	53.3
high	54	11.1	9.9	476	30	4890	11.4	16.9	53.6

Table 7: Sample graded by KVR

4.3.7 Sample graded by Predicted MOR
The predicted MOR is the MOR estimated by linear regression from the parameters given by the X ray grading machine. We can observe that the mean of the predicted MOR value is nearly equal to the mean value of the real MOR. This means that the X-ray machine has a good capacity to predict the mean value of a graded group. On the other hand, the X-ray grading machine is the best method with the US wave to grade round woods. These two machines have been developed for two different uses which are (i) To evaluate a stucture in a finished building which cannot be done by the X-rays machine (too heavy !) ; and (ii) To realise a fast industrial grading where UltraSonic waves (Sylvatest) can't be used (too slow!). So the two technics are complementary.

	MOR MPa	MOEL 12% GPa	MOEG 12% MPa	MV 12% Kg/m3	KAR %	US Speed m/s	US modulus GPa	KVR %	predicted MOR
low	47	10.5	8.6	411	24	4771	9.4	12.0	46.7
medium	53	11.1	9.5	437	25	4912	10.6	13.5	52.3
high	58	11.8	10.2	480	24	4942	11.8	14.0	58.5

Table 8: Sample graded by Predicted MOR

5.CONCLUSIONS

The strength grading allows us to guarantee the fact that wood properties are sufficient for its use, particularly when strength and stiffness properties are reliable. Building design have to be based upon pieces grading and mechanical strength values defined by standards recognized at the European level. Strength can only be indirectly defined by parameters which can be visually set up or with other non destructive methods. So the choice of the grading method is important for at least an economical point of view. In this way, the X-rays densimeter from CTBA allows a machine grading to take into account the following aspects :
==> output :
This parameter aim is to test the grading capacity to extract the highest number of pieces of acceptable mechanical characteristics.

==> Simplicity:
This criteria aim is to assess the adaptation facility of grading in firms. It is defined as the number of criteria which allow the grading. The throughtspeed speed of sawns is about 200 m/min which meets the industrial requirements.

==> Reliability:
For a given class, 95% of graded pieces show values higher than class values (in accordance with the quality control of product).

Bibliography:

1. Sandoz J.L. "Triage et fiabilité dus bois de construction validé par la méthode Ultrasons." Ecole Polytechnique Fédérale de Lausanne, These 1990.
2. Guillot J.L., Lanvin J.D., Sandoz J.L. "Classement structure du sapin/Epicea par réseaux de Neurones à partir de mesures SylvaTest" 40 Colloque des Sciences et Industries du Bois, Nancy, Septembre 1996
3. Chambellan D. ; Pascal G. ; Reverchon P. "Imageur radiometrique pour le controle par défilement ou la tomodensimetrie à l'aide de photons X ou gamma." 6 eme Conférence COFREND, Nice 94
4. Program TIMGRAD "Improving garding methods for structural timber by non destructive techniques" BRE CT 94 1567 CR 1840-91
5. EN 338 : Structural Timber -Strength classes
6. Rouger F. "Application of a modified statistical segmentation method to timber machine strength grading." Wood & Fiber Science, 28(4), 1996