Within-board lumber density variations from digital X-ray images

Transcription

1 Ninth International Symposium on Nondestructive Testing of Wood Within-board lumber density variations from digital X-ray images B. Suryoatmono, S. Cramer, Y. Shi, and K.A. McDonald Abstract Computed radiography (CR) is a potentially practical method to obtain density and internal structure characteristics of lumber for a variety of processing and grading decisions. A CR method is presented to obtain within-board density variations and knot sizes in southern pine lumber. The method will be used in conjunction with grain angle scanning to more fully characterize the unique mesostructure of individual boards. These characterizations will be used in a computer model to simulate the mechanical response of boards. A calibration method has been developed to transform exposure readings to density readings on a 0.9- by 0.9-mm grid. Preliminary comparisons of CR density with gravimetrically determined density for 35- by 35- by 35-mm small clear blocks indicates an average difference of approximately 1 percent. Further refinement and verification of the method is continuing. Examples of density variation in the vicinity of knots are presented and discussed. The authors are, respectively, Research Assistant, As- Soc. Professor, and Research Assistant, Dept. of Civil and Environmental Engineering, Univ. of Wisconsin, Madison, Wis.; and Res. Forest Prod. Technologist, USDA Forest Serv., Forest Prod. Lab., Madison, Wis. The authors gratefully acknowledge the support of USDA Forest Serv., Forest Prod. Lab. Appreciation is extended to W. Peppler and the Dept. of Radiology and Medical Physics at the Univ. of Wisconsin for generously allowing access to facilities and assisting in technical issues related to data collection. Introduction and purpose Computed radiography (CR) offers a means to rapidly assess localized changes in density within individual pieces of lumber. Previous research has shown that specific gravity within a board increases dramatically in and around a knot, and has shown that these increases correspond to localized increases in tensile strength perpendicular to grain (1). The overall research objectives are to nondestructively measure mesostructure characteristics of boards, such as density variations, and to establish the link between mesostructure and structural performance. Mesostructure is the middle level of detail between characterizing cell structure at the microscopic level and traditional treatments of wood as a homogeneous solid. Previous mesostructure characterization has focused on identifying grain angle variations and their relationship to board fracture patterns (3,4). In this paper, we present an additional level of detail by identifying within-board density variations with CR. The equipment used is unique, but likely not prohibitively expensive for applications in the wood products industry. The long-range goal of this research is to provide fundamental methods and information for making edging, trimming, and grading decisions in the production of lumber. Relevant literature Radiation densitometry is a commonly used technique in dendrochronology (2,5,21,22) and in determining density profiles of some wood composites (17,23). The use of this technique in obtaining density distributions within lumber 168 Suryoatmono, Cramer, Shi, and McDonald

2 specimens to aid strength prediction is new. The method provides a detailed picture of density changes within a specimen. The use of X-ray radiation for density analysis was developed in France (22). Other forms of electromagnetic radiation such as gamma rays and beta rays have also been employed in density analysis (6,13 15,24).Different investigators have used different methods of scanning regarding the relative position between object and X-ray source, namely fixed object/moving X-ray tube, moving object/ fixed X-ray tube, and fixed object/fixed X-ray tube (2,7,8,11-14,20). This paper describes a digital X-ray scanning method using a fixed object and fixed X-ray tube. The principle of scanning densitometry is to pass a collimated photon beam through a specimen and into either a radiation sensing device (photoscintillation detection system) or a combination of X-ray film and intensifying screens. Most radiation densitometry utilizes the former because of difficulties associated with film processing and density determination from film. The method described in this paper, however, uses neither a scintillation detector nor X-ray film, but rather a photostimulable phosphor plate. By recalling how X-rays are produced and how they interact with an exposed object, one can explain why there is always some deviation in density determination using X-ray scanning techniques. X-rays are produced in an X-ray tube when electrons accelerated from the cathode striketheanode dueto agivenpotentialdifference. When an electron is stopped by nuclei in the anode, a large amount of the electron energy is converted to heat and the rest appears as X-rays. All electrons striking the anode have the same energy, but the X-rays produced have many different energies. As a result, the X-ray intensity received by an exposed object is not uniform from point to point. X-rays interact with an exposed object in three ways, namely coherent scattering, photoelectric interactions, and Compton scattering. The total interactions govern the mass attenuation factor defined as the fractional reduction of X-ray intensity per material density per unit thickness. Coherent scattering, the change in X-ray photon direction without energy transfer, is most clearly identified with lower energy X-rays. Coherent scattering has little effect on wood densitometry except for a slight blurring of edges. In photoelectric absorption, on the other hand, the X-ray photon is completely absorbed by the exposed atom. Energy transfer, the amount of which depends on the atomic number of the exposed atom, happens in this interaction. In wood material, the main elements are C, O, H, and N with atomic numbers 12, 16, 1, and 14, respectively. Photoelectric absorption is responsible for the major difference in radiographic appearance (contrast) between higher and lower density regions in an exposed board. In the third kind of interaction, Compton scattering, the X-ray photon transfers some energy to the electrons of the exposed material, and proceeds in a different direction with a reduced energy. Because Compton scattering is almost completely independent of atomic number, it is approximately the same for all materials and depends only on the incident X-ray energy (9). The total effect of these three X-ray and material interactions appears in the following equation: = mass attenuation factor [1] Clearly, depending on the compositions of the main elements, the mass attenuation factors are not uniform among species (10). Within a lumber specimen, however, the mass attenuation factor can be assumed constant, in accordance with the homogeneous material assumption. A specially designed filter grid can be used to reduce scatter. It consists of thin flat sheets made of alternating lead and aluminum. Lead strips are designed to attenuate the scattered radiation and the aluminum interspace to allow the passage of the primary X-ray radiation. Another simple way is by using an air-gap scatter reduction scheme where some distance is maintained between the specimen and the X-ray film. The intensity of the collimated beam of monochromatic X-ray radiation after it has passed through some absorber of thickness t may be determined using the following attenuation equation (15): [2] I = intensity of the radiation beam after passing through the wood (counts/unit time) I 0 = intensity of the radiation beam before passing through the wood (counts/unit time) Suryoatmono, Cramer, Shi, and McDonald 169

3 t = thickness of absorber or exposed material (unit length) µ l = linear attenuation factor (unit length -1 ) The linear attenuation factor represents the fractional number of photons absorbed from a beam of radiation per unit length of path of the absorber. As mentioned earlier, the magnitude of attenuation is dependent upon the source of radiation and the atomic numbers of the elements of the absorber. Therefore, transmitted radiation can be defined as: [3] µ m = mass attenuation factor of the absorber (cm 2 /g) = µ l /p p = density of the absorber (g/cm 3 ) Mass attenuation factor is a material property. In general, it depends on the constituent elements in the material and the energy of radiation. Clearly, in wood, it also depends on the moisture content because water has different constituent elements than solid wood. Olson and Arganbright have estimated the mass attenuation factors of several wood species at different energies using elemental mass attenuation factors weighted by their respective wood elemental compositions (1 8). These contributions are based on photoelectric interactions only and neglect possible contributions from coherent and Compton scattering. Computed mass attenuation factors varied among investigated species (pine and larch) at each X-ray energy below 40 kev. However, above this energy the factors were essentially the same at each energy, regardless of difference in constituent elements in wood. Equation [3] can be written in the form of Equation [4] can be used to find the density, p, when I and I 0 or the proportion of I/I 0 is known. These two radiation intensities can be obtained using an X-ray scintillation detection system, for a constant t and small aperture area. This area should be so small that no density variation occurs over the area (16). Computed radiography is a filmless radiography. X-ray images are captured and converted to digital signals, which can be stored or retrieved using disk storage. CR allows customization of contrast and resolution of an X-ray image. Quality images can be obtained even with low radiation levels. Another advantage of CR is that the image plate is reusable. Figure 1 shows the image plate process from exposure to erasure (21). As seen in Figure 1, CR uses a scanning laser stimulated luminescence produced in an image receptor plate exposed to X-ray radiation, and converts the light to digital data. The image plate, coated with photostimulable phosphor crystals, replaces the conventional film/intensifying screens as a medium to detect and record a highquality image. The phosphor crystals act as energy traps when exposed to ionizing radiation. 1 In both conventional X-ray radiography and CR, several factors can affect the quality of X-ray radiographs, such as fluctuation in supply voltage to the X-ray generator, lack of uniformity in sample thickness, and moisture content of the wood (19). The effect of fluctuation in supply voltage to the X-ray generator can be addressed by using a very short exposure duration. Sample thickness and moisture content can be controlled by careful sample preparation. Method, resolution, and accuracy In the present study, X-ray scanning is used to gather within-board density information. The Figure 1. CR image plate process from exposure to erasure (after 21). 1 X-rays and are examples of ionizing radiation. They have enough energy to separate one or more electrons from the atom to produce an ion pair. 170 Suryoatmono, Cramer, Shi, and McDonald

4 Department of Radiology and Medical Physics at the University of Wisconsin, Madison, Wis. provided access and technical assistance in using the X-ray tube and digital scanning equipment. To obtain high-quality radiographs of a wood specimen, the X-ray source was positioned mm from the image plate to reduce parallax distortion (19). Each plate was exposed for 0.5 seconds and the electrical current was 200 ma. Both the air gap technique (152 mm) and a filter grid with a 3.2- by 3.2-mm grid layout were used to reduce X-ray scattering. The energy level of the X-ray source was 50 kev. These parameters were kept constant throughout the study. The image plate was digitally scanned after X-ray exposure using the Philips Computed Radiography (PCR) System to produce readings on a by 0.18-mm grid over a 319- by 388-mm (14- by 17-in.) area resulting in almost 19 megabytes of ASCII data. For simplicity, data was reduced to a 0.9- by 0.9-mm grid. The readings were machine-dependent numbers ranging from 1 to 1024 and were influenced by parallax and scatter. A reading of 1 indicated highest radiation exposure and a reading of 1024 indicated lowest radiation exposure. We denote the digital representations as PCR values. According to the specification of the PCR system, the correlation between the PCR value and the X-ray radiation is: [5] E = radiation exposure in milliroentgen (mr) PCR = PCR value L, S = laser reader parameters The value of L controls the reading range of the PCR values. It can be preset via the control panel of the system. A value of L=1, which lets the laser reader read the entire range of the image, was chosen. The value of S controls the reading sensitivity. It cannot be entered directly by the user and is calculated by the PCR system. AS mentioned earlier, density can be obtained by using the standard attenuation equation, i.e., Equation [4]. The ratio of I/I 0 in Equation [4] is equal to the ratio of E/E 0 where E 0 is defined as the X-ray radiation in mr without any absorber. Therefore, the radiation exposure at any point, if there is no absorber, can be expressed as: PCR 0 = PCR value, if there is no absorber [6] PCR 0 values vary from point to point, substantiating the preceding discussion that X-ray radiation is not uniform over an area. PCR 0 and PCR i values should theoretically be used at the same location in calculating density at any arbitrary point i. However, it is impossible to obtain PCR values at the same location from the same exposed image plate; the absorbing material either does or does not cover the point. As a result, PCR 0 is established for each point by gathering a blank exposure at the beginning of each scanning session to establish the scatter characteristics of the experimental setup. It was established that the scatter characteristics remain constant during a scanning session of several hours. Thus, by this procedure, PCR 0 and PCR i at any point i can be obtained from a nonsampled exposure and a sampled exposure, respectively. The value of S of a nonsampled image plate and that of sampled image plate are different. This is why S 0 appears in Equation [6]. By substituting E/E 0 = I/I 0, into Equation [4] and using Equations [5] and [6], one finds that density at point i is shown in Equation [7]: [7] Equation [7] follows traditional radiology the- ory. Due to all uncertainties in using X-rays ex- plained earlier, however, it appeared that some adjustments were needed. We introduced two experimental constants, A and B, as shown in Equation [8]: [8] To evaluate these constants, we employed two small specimens called calibration samples. Gravimetric densities of both specimens were substituted into Equation [8] to yield Equation [9] containing unknowns A and B. Constants A and B are the solution of the resulting linear simultaneous equations. A unique solution will result from two distinct values of thicknesses, t 1 and t 2, of the calibration samples 1 and 2, respectively in the path of the X-ray. Suryoatmono, Cramer, Shi, and McDonald 171

5 [9] As discussed earlier, mass attenuation factor is a material property that depends on the X-ray energy. In the case of wood, it depends not only on solid wood content, but also on moisture content. To investigate the effect of moisture content on mass attenuation factor, we conducted X-ray exposures of southern pine under moisture content conditions of 12 percent and ovendry. Twenty small wood blocks at 12 percent moisture content were exposed to X-rays. The gravimetric density of each block was measured and substituted into the theoretical equation (Eq. [7]) to obtain its mass attenuation factor. The average mass attenuation factor was cm 2 /g with a coefficient of variation of 1.1 percent. Similarly, ten small blocks were ovendried for 39 hours at 103 C and exposed to X-rays. The mass attenuation factor for dry wood was only 1 percent lower than wood at 12 percent moisture content. As a result, we adopted a constant mass attenuation factor of cm 2 /g for the entire range of moisture contents studied. That compares to a value of computed by Olson and Arganbright considering the photoelectric effect only (1 8). Densities of full-width lumber specimens (38 by 140 mm and 30 by 89 mm, length 160 to 400 mm) and small blocks (35 by 35 by 35 mm) were gathered using the complete procedure outlined in Figure 2. CR densities were compared to gravimetric densities. Gravimetric densities relied on caliper measurements to the nearest 0.1 mm and weights to the nearest g. Even exercising the utmost care, some small errors are inherent in the gravimetric technique. It was found that the maximum difference from the two measurement techniques for small clear blocks was 6.1 percent and that for actual size lumber specimens was 3 percent. Figure 3 shows a comparison between gravimetrically measured densities and CR densities. Using the method described earlier, data were collected from several exposures on different days and different moisture content conditions (12% and ovendry) using 123 samples. The correlation coefficient (r 2 ) between the two types of readings was 0.97 and the average error was 1.3 percent. Density patterns around knots Using the above method, we conducted a small preliminary investigation of density patterns around Figure 2. Schematic of the CR density determination method. 172 Suryoatmono, Cramer, Shi, and McDonald

6 Table 1. Density data for southern pine boards containing knots. Average Maximum Ratio of Ratio of width of Ratio of length of sample sample maximum density high-density zone to high-density zone to Sample density density to average density transverse knot diameter longitudinal knot zone (kg/m 3 ) knots. Five specimens, consisting of 520-mm lengths of 38- by 140-mm (2- by 6-in.) southern pine each containing a single wide-face knot ranging in size from 30 to 45 mm in diameter, were subject to X-ray and data analysis. The results are summarized in Table 1. On average, the density within the knot increases by a factor of two above the average specimen density. High-density zones were defined as regions that possessed a density 20 percent higher than the specimen average. These high-density zones, in all cases, included the whole knot-wood region and also a transition zone of deviated grain in the clear wood. As indicated by Bethge et al. (1), this increase in density in the transition zone is tied to increases in perpendicular-to-grain tensile strength. This suggests that the natural growth process attempts to compensate for the high grain angles and weakening effect of a knot by increasing density in the highdensity transition zone. Table 1 reveals that the width of the high density zone varies considerably from 1.0 to 1.7 times the transverse knot diameter. The length of the high-density transition zone varies from 1.1 to 1.4 times the longitudinal knot diameter. It is postulated that the absence of this transition zone in edge-knot situations reduces the strength that would otherwise be achieved. Figures 4 and 5 show the variation in density along transverse lines (Fig. 4) and a longitudinal line (Fig. 5) running near or through the knot in sample 5. The visual knot size and location is superimposed over the density lines in these figures. The transition zone clearly lies outside the visual boundaries of the knot, especially in the transverse direction. In all cases, this transition zone of increased density is within one knot diameter of the knot center point. It has been known for some time that visual knot sizing does not capture the grain deviation associated with a knot. Figures 4 and 5 combined with the data in Table 1 indicate that visual knot sizing Figure 3. Comparison of CR and gravimetrically measured densities. does not fully identify the region of high density associated with a knot. Summary and conclusions The primary objective of this study was to gather within-board density variations of lumber. This was achieved through the development of a unique but simple method utilizing digital X-ray scanning and the PCR system to provide density maps on a 0.9- by 0.9-mm grid. A calibration technique utilizing a nonsampled exposure and two small calibration specimens was developed to improve accuracy. Although many factors, such as nonuniformity of X-ray exposure and the prediction of mass attenuation, played important roles in the discrepancy of the radiographic densities, comparison with gravimetrically determined densities for small clear blocks indicates an average difference of only 1 percent. This method will ultimately be combined with grain angle data and utilized in new strength prediction models. Suryoatmono, Cramer, Shi, and McDonald 173

7 Application of the CR method has revealed knot densities that are twice the average board density in southern pine samples. Most importantly, a transition zone between knot and clear wood ex- ists along the outer boundaries of a knot where density increases dramatically. Previous research suggests the increased density in the transition zone is tied to an increase in perpendicular-to- Figure 4. Density variation along the transverse direction (width) of a 38- by 140-mm (2 by 6-in.) sample of southern pine containing knots. Figure 5. Density variation along the longitudinal direction (length) of a 38- by 140-mm (2 by 6-in.) sample of southern pine containing knots. 174 Suryoatmono, Cramer, Shi, and McDonald

8 grain tensile strength, suggesting that the transition zone attempts to compensate for the weakening influence of the knot. Literature cited 1. Bethge, K., C. Mattheck, K.A. McDonald, and S.M. Cramer Failure behavior of knot-containing boards under tensile and compressive loads. Primärbericht, Karlsruhe Nuclear Res. Center, Inst. of Material Res. II, Karlsruhe, Germany. 29 pp. 2. Cown, D.J. and B.G. Clement A wood densitometer using direct scanning with X-rays. Wood Sci. Tech. 17: Cramer, S.M. and K.A. McDonald Predicting lumber stiffness and strength with local grain angle measurements and failure analysis. Wood and Fiber Sci. 21(4): et al Exploring the relationship between local grain angle and initial fracture in lumber subject to tensile load. In: Proc Inter. Conf. on Timber Engineering. Vol. 2. Forest Prod. Res. Soc., Madison, Wis. pp Fletcher, J.M. and J.F. Hughes Uses of X-rays for density determinations and dendrochronology. In: Tree analysis with special reference to Northwest America. J.H.G. Smith and J. Worrall, eds. B.C. Faculty Bull. No. 7. Vancover, B.C., Canada. pp Harris, J.M The use of beta rays in determining wood properties. Parts 1-5. New Zealand J. Sci. 12(2): Henkel, M Measurement and density profiles in particleboard and fiberboard by X-ray. Holztechnologie 10(2): Hoag, M. and M.D. McKimmy Direct scanning X-ray densitometry of thin wood sections. Forest Prod. J. 38(1): Kelsey, C.A Essentials of Radiology Physics. W.H. Green, St. Louis, Mo. 10. Kouris, K.R.E. et al Effect of constituent elements on wood in X-ray densitometry measurements. Archaeometry 23(1): Lindgren, L.O On the relationship between density/moisture content in wood and X-ray attenuation in computer tomography. In: Proc. 5th Nondestructive Testing of Wood Symp., Washington State Univ., Pullman, Wash. pp The accuracy of medical CAT-scan images for nondestructive density measurements in small volume elements within solid wood. Wood Sci. and Tech. 25: Medical CAT-scanning: X-ray absorption coefficients, CT-numbers and their relation to wood density. Wood Sci. and Tech. 25: Malan, F.S. and P.G. Marais Some notes on the direct Gamma ray densitometry of wood. Holzforschung 46: Moschler, W.W., Jr. and E.F. Dougal Calibration procedure for direct scanning densitometer using gamma radiation. Wood and Fiber Sci. 20(3): and P.M. Winistorfer Direct scanning densitometry, An effect of sample heterogeneity and aperture area. Wood and Fiber Sci. 22(1): Nearn, W.T. and K. Bassett X-ray determination and use of surface-to-surface density profile in fiberboard. Forest Prod. J. 18(1): Olson, J.R. and D.G. Arganbright Prediction of mass attenuation coefficients of wood. Wood Sci. 14(2): Parker, M.L. et al A computerized scanning densitometer for automatic recording of tree ring width and density data from X-ray negatives. Wood Fiber 5 (3): and LA. Jozsa X-ray scanning machine for tree-ring width and density analyses. Wood Fiber 5(3): PCR/PACS, A Radiology Evolution AT&T and Philips. 22. Polge, H. and P. Lutz Possibilities of density measurement of particleboard perpendicular to the plane by X-rays, Holtechnologie 10(2): Steiner, P.R. et al Application of X-ray densitometry to determine density profile in waferboard relationship of density to thickness expansion and internal bond strength under various cycles. Wood Sci. 11(1): Winistorfer, P.M. et al A direct scanning densitometer to measure density profile in wood composite products. Forest Prod. J. 36(11/12): In: Proceedings of 9th International symposium on nondestructive testing of wood; 1993 September 22-24; Madison, WI. Madison, WI: Forest Products Society; 1994: on recycled paper Suryoatmono, Cramer, Shi, and McDonald 175