ASSESSMENT OF LOCAL TIMBER DEFECTS DURING TESTING AND GRADING AS INFLUENCED BY MACHINE APPROVAL PROCEDURE Andreas Rais1, Peter Stapel2, Jan-Willem van de Kuilen3
ABSTRACT: Depending on the utilised principle, many grading machines are only able to measure an average property of the board, for example density or longitudinal eigenfrequency. Other grading machines on the market are able to detect local weaknesses such as knots and therefore try to locate the weakest cross section. The basis for the derivation of settings is a comparison between data of the laboratory and data from the grading machine. A regression model is created containing the machine parameters, which provide the best fit between indicating properties and grade determining properties. If the grading machine detects knots, the maximum knot value between the inner load points (bending test) is used for developing a model. This maximum knot value is also used for deriving settings. It is obvious, that it is not always possible to choose the maximum knot value of the entire board. Consequently, the very same board is assigned to a lower indicating property during testing than during grading in practise. KEYWORDS: Machine grading, local weakness, knots, bending test, length effect
1 INTRODUCTION 123 The production of strength-graded timber in Europe as well as the number of approved strength grading machines listed in EN 14081-4 [1] is increasing. The method for determining settings for these according to EN 14081-2 [2] can be illustrated as follows: All pieces pass the grading machine for recording the machine data. These specimens are tested in laboratory in order to receive the characteristic values for strength, modulus of elasticity, and density. Testing of the critical section between the loading points in a four-point bending test is required according to EN 384 [3]. The standard EN 14081-2 [2] emphasizes that this section is the position where failure is expected to occur. The grade determining properties (strength, modulus of elasticity, density) are representative for the original board. These data are combined with the indicating 1 Andreas Rais, Holzforschung München, Technische Universität München, Winzererstr. 45. 80797 München, Germany. Email:
[email protected] 2 Peter Stapel, Holzforschung München, Technische Universität München, Winzererstr. 45, 80797 München, Germany. Email:
[email protected] 3 Jan-Willem van de Kuilen, Holzforschung München, Technische Universität München, Winzererstr. 45, 80797 München, Germany. Email:
[email protected], TU Delft, the Netherlands
properties (IP, condensed machine values) to calculate a model, which is used to predict the properties of a board. Finally, the settings are determined. There are models that use knot measurements to estimate the strength class. As is generally known, the most important strength-reducing characteristics in softwoods are the knots. Investigations have shown that more than 90 percent of failures in Norway spruce occur in the vicinity of knots. This paper deals with different assessments of local defects in boards such as knots and slope of grain, during approval testing of grading machines and grading operations in practise. When deriving settings, the maximum knot value of the entire board is applied only if it is located in the middle part of the board; otherwise the knot value from the middle part is used. But how does the grading machine evaluate these boards in practice? Most grading machines using knot values in the model are able to detect knots over the entire length of the board. The knot value varies over the entire board length. The software calculates the largest knot value. This value is plugged into the model and predicts the strength class of the complete board. Comparing the indicating properties of the very same boards during testing and grading there is a systematic decrease in yield in practise for grading machines using knots to determine strength classes. Figure 1 shows,
which indicating property can be calculated from the testing and the grading range. 60
n = 660
IP at testing
50 40 30 20 10 0 0
10
20
30
40
50
60
IP at grading
Figure 1: Discrepancy between the indicating properties IP during testing and grading
The objective of this paper is to describe the discrepancy of different maximum knot values, which have to be taken into account when deriving machine settings compared to the maximum knot value at the time of grading for the same board. The original length of the tested board influences this discrepancy.
2 Material and method 2.1 Material The dataset comprised a total of 146 Norway spruce (Picea abies) specimens from Poland. The cross section is 40 x 100 mm², and the mean length is 4002.5 mm. All the specimens pass the grading machine GoldenEye-702, so it is possible to record the distribution of the knots. For each of the 146 boards the machine determines the ten largest knot values. Additionally, the position of the knots over the length of the board is known. The knot value used is the same value, which is applied in the models for predicting the bending strength using GoldenEye-702. Table 1 shows the mean value and the coefficient of variation (cov) of the bending strength fm, the local modulus of elasticity Em and the density r12. Figure 2: Distribution of the three largest knot values over the entire board length
Table 1: Description of the sample
n
fm
Em
r12
2.2 Method
mean
cov
mean
cov
mean
cov
-
N/mm²
%
N/mm²
%
kg/m³
%
146
34.2
34.4
10200
20.5
436
9.3
Figure 2 shows the distribution of the three largest knot values up to 4 meters. The maximum knot (largest knot) is identified both between the inner load points (testing) and over the complete length (grading).
In Figure 3 the grey-coloured ends of the board represent the areas, which cannot be considered for physical testing due to the bending test set-up of EN 408 [4]. For boards with 100 mm depth, it means that as a maximum about 70 centimetre from one of the board ends falls outside the testing range. The maximum knot value during testing can only be searched in the remaining middle part of the tested board. It is always attempted to choose the largest knot. If it is not possible to test the largest knot, the next largest knot is chosen (alternative knot).
3 Results testing
grading
Figure 3: Discrepancy between knots during testing and grading
The maximum knot value of each board, which is detected and calculated by the grading machine, is compared to the maximum knot value taken when deriving the settings. Knot values are determined based on an X-ray picture of the board and are evaluated according to a certain procedure, leading to numerical values in the range of 500 to 3000. At the time of deriving the settings, the maximum knot value depends on the original length of the tested board. As the original length of the board seems to be important, five different length categories of tested boards are created by theoretically cutting boards: The first length used to analyse possible effects on the tested boards is exactly 19 times its height, the second length is 24 times the height, the third length is 29 times the height, the fourth length is 34 times the height and the fifth length is 39 times the height. This investigation focuses on the maximum knot values of the shortest (1) and longest (5) boards. The five different original length classes of testing are numbered from 1 to 5, starting with the shortest length. long
The length effect described in this paper reveals that the knot value during testing which is used for deriving settings is not regarded as being representative for the knot value during grading. By decreasing the original board length the discrepancy increases. In contrast to other parameters used in machine grading like density and eigenfrequency, the knot values vary over the entire board. The broken line as well as the continuous line in Figure 5 shows that the mean of the maximum knot value increases with the length of the board. The mean of the maximum knot value found over the entire board length increases from 1500 to 1650 (length class 1 to 5). This is not surprising, because the probability of large knots to occur increases with increasing length. This is a wellknown effect. Recently that kind of length effect is described by [5] with respect to visual grading rules. The continuous line indicates the situation of bending tests according to the European standard EN 408 [4] with increasing length. The mean of the maximum knot value is always below the broken line. The distance between the two lines decreases with increasing length. If the original length of the tested board is that small, that there is no other option than to test the middle part of the board, the discrepancy between knot value during testing and grading reaches its maximum value. The mean of the maximum knot value increases from just over 1300 for short boards to just over 1600 for long boards.
short beginning of the board
1
2
3
4
5
Figure 4: One of 146 specimens considering 5 different length categories
1600
1500
complete board (grading) between load points (testing)
1400
1300 1 short
Finally, the indicating properties are calculated according to Equation (1) below – during deriving settings and during grading in practise. IP = a + b ⋅ knot + c ⋅ density
maximum knot value
1700
2
3 board length
4
5 long
Figure 5: Influence of the length on the knot value
(1)
IP = indicating property, a, b, c = regression coefficients, knot = maximum knot value, density = mean density of the board. Hence, all boards are graded twice. First the maximum knot value is entered in the given model when deriving settings. Second the maximum value is entered in the given model when grading in practise. The density is assumed constant.
Figure 6 shows the rank of the knot size, which can be placed between the load points for every single board. For short lengths (1) the maximum knot value can be tested for 70 boards. Looking at the remaining 76 boards, the maximum knot is located outside of the load points. From a statistical point of view, the amount of 70 boards is explained by the knot distribution of the maximum knot (see Figure 2); actually it is expected that at one third of all boards the maximum knot value of the entire board can be selected.
140
number of boards
120
board length
100
short (1)
80
4000 knot value of short (1) boards at testing
If the tested board is long (5), the maximum knot value can be selected in most cases. Only in 29 of the cases an alternatively smaller knot value has to be chosen.
long (5)
60
n = 146 3000
2000
1000
0 0
40
1000
2000
3000
4000
knot value of short (1) boards at grading
20 0 1
2
3
4
5
6
no knot
ranking of knot size
Figure 8: Discrepancy of knot values between testing and grading of short boards
Figure 6: Rank of the knot size value, which can be placed within the limits of the test range
relative alternative knot value
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0
n = 29 n = 76
short (1)
long (5) board length
Figure 7: Largest knot value at testing in relation to the largest knot value at grading depending on the original tested length
Figure 8 and 9 summarize the effect of different original lengths during testing. The longer the board, the smaller the difference between knot values used in testing and knot value used in grading.
knot value of long (5) boards at testing
Now, the alternative knot value is considered. This knot value is plugged in the model during the settings derivation process, when the primary largest knot size cannot be tested. The length of the original board tested, also influences the dimension of the alternative knot value. Figure 7 shows that with increasing length the alternative knot values become higher. The median of the boxplot on the right is 0.86 (long boards); the median of the boxplot on the left is only 0.71 (short boards). The whiskers are limited by the maximum and minimum value.
4000 n = 146 3000
2000
1000
0 0
1000
2000
3000
4000
knot value of long (5) boards at grading
Figure 9: Discrepancy of knot values between testing and grading of long boards
The above figures make the difference between short (Figure 8) and long (Figure 9) boards obvious. The difference between the knot values is high for short boards. Many points are located below the bisecting line. This shows that even the alternatively tested knot value is a bad compensation for the highest knot value on the short specimen. For long (5) boards the discrepancy is smaller and less frequent. On the one hand the biggest knot can be placed between the supports in most of the cases, and on the other hand the alternative knot value is more likely of the same dimension. In a next step the effect on the yield of a grading machine using knots is shown when grading the strength class C 30. During deriving settings the maximum knot value is only taken from the middle part of the boards. In contrast, when grading in practise the grading machine is sensitive to knots of the entire board. Using long boards for deriving settings, the indicating property of the same board during testing and grading is almost identical (Figure 9 and right hand side of Figure 10). In relation to the knot value and the indicating properties, the tested range between the load points can identify the entire tested board correctly. This means that for each board the maximum knot value used for testing is equal to the one used for grading. There is a slight difference of four boards in the yield for strength class C 30.
If the length of the tested board is short (left hand side of Figure 10) and nothing but the middle part can be tested, the maximum knot value of this section must be taken. Consequently, the indicating property of the same board varies significantly between testing and grading. The yield drops by 39 % or 24 boards. Furthermore, Figure 10 shows an additional length effect, which is already described for visual grading by [5]: During grading the yield goes down from 38 boards at a short length to 29 boards at a long length. Since grade varies longitudinally in timber, grade yield decreases if the length of timber increases. The variation in grade is mostly caused by longitudinal variation in knot properties. 70
n = 62
boards graded in C30
60 50 n = 38
40
n = 33
n = 29
30 20 10 0 testing
grading short
testing
grading long
Figure 10: Different yields during testing and grading of short and long boards
4 Conclusions Knots are valid predictors for grading determining properties. Some machines utilize the maximum knot value for assessing strength classes. The machine can cover the entire length of each board to seek for the largest knot. But when the settings are derived, it is not always possible to test this largest knot, because it might be located at the ends of the board. Then, the next largest knot is applied and also plugged in the model. In these cases, the machine treats the same board differently when graded for testing and when graded in practise. During grading in practise the indicating properties tend to be lower. This investigation shows that this effect is systematic and is affected by the original length of the tested board. In practise, the different assessment of knot values of the very same board during grading and testing leads to a decrease in yield. It means that the length effect in timber as reported for instance by [6] is automatically included. As a consequence, grading machines that detect local weaknesses are able to grade correctly also longer boards than used in the approval testing. However, this higher performance is not considered when machines are approved according to the European standard EN 14081.
ACKNOWLEDGEMENT We acknowledge Martin Bacher from MiCROTEC for providing the data and for his advice and support.
REFERENCES [1] EN 14081-4:2008-08 Timber structures – Strength graded structural timber with rectangular cross section – Part 4: Machine grading – Grading machine settings for machine controlled systems, European Committee for Standardization, Brussels, 2008. [2] EN 14081-2:2008 Timber structures –Strength graded structural timber with rectangular cross section – Part 2: Machine grading; additional requirements for initial type testing, European Committee for Standardization, Brussels, 2008. [3] EN 384 Structural timber - Determination of characteristic values of mechanical properties and density, European Committee for Standardization, Brussels, 2003. [4] EN 408 Timber structures - Structural timber and glued laminated timber - Determination of some physical and mechanical properties, European Committee for Standardization, Brussels, 2003. [5] Øvrum A, Vestøl GI, Høibø OA (2008) Modeling the longitudinal variation of sawn timber grades in Norway spruce (Picea abies (L.) Karst.). Holz als Roh- und Werkstoff 66(3): 219–227. [6] Isaksson T, Thelandersson S (1995) Effect of test standard, length and load configuration on bending strength of structural timber. CIB-W18, Copenhagen, Denmark.
