A method for optimizing lumber sorting before kiln-drying.
Drying lumber can be quite challenging when uniformity and consistency are required. Over-drying and under-drying are common causes for lower grade recovery and dimensional stability problems. It is known that both over-drying and underdrying can be reduced by sorting green lumber into different moisture content groups. Additionally, sorting also offers the opportunity to redesign drying schedules. In this study, a new methodology was designed and tested to optimize kiln- drying of lumber by implementing green sorting coupled with modified drying schedules. The methodology was applied to optimize the drying of 114 by 114 [mm.sup.2] hem-fir lumber sorted with an NMI capacitance type meter. It was found that m comparison to unsorted lumber, sorting into three groups reduced the drying time by approximately 7 percent and recovered around three-quarters of the under-dried lumber.
As defined by the Western Wood Products Association (1997), hem-fir is a commercial name for the combination of western hemlock (Tsuga heterophylla) and any of the five different types of true firs. In the coastal area of British Columbia, hem-fir is around 70 to 75 percent western hemlock and 25 to 30 percent amabilis fir (Abies amabilis). Drying hem-fir lumber is difficult because of its large moisture content (MC) variability. According to Nielson et al. (1985), green MC in western hemlock ranges from 55 percent in heartwood to 143 percent in sapwood (with an average of 85%), and green MC in amabilis fir ranges from 55 percent in heartwood to 164 percent in sapwood (with an average of 65%). Additionally, hem-fir lumber with exactly the same green MC can dry at different rates depending on factors such as species, basic density, percentage of sapwood and heartwood, presence of juvenile and reaction wood, lumber dimensions, annual rings orientation, and non-uniform drying conditions inside the kiln. For example, it was proven experimentally that for 114 by 114 mm hem-fir lumber, pieces with vertical annual ring orientation take 8 days longer to dry than pieces with horizontal annual ring orientation (Avramidis and Hao 2003).
Regardless of the reason, hem-fir lumber always has a large MC variability after drying. Dried lumber with a final MC higher than the certain limit of between 19 percent and 21 percent is considered wet. There is not a specific MC limit that defines over-dried lumber, but lumber with a final MC below 10 to 12 percent is more likely to downgrade due to excessive shrinkage and distortion. In general, final MC variability affects the lumber dimensional stability depending on the environmental conditions (Avramidis et al. 2005). Reducing final MC variability is, therefore, a priority in kiln-drying, and it can be partially achieved by implementing lumber sorting. Lumber sorting before drying is a very common practice in the lumber manufacturing industry, since it creates groups of lumber with similar "dry-ability" characteristics.
At the laboratory level, Kozlik and Ward (1981) studied the dry-ability characteristics of sorted young-growth hemlock. Holmes and Arganbright (1984) measured air-drying times for different sorts of incense-cedar. Kozlik (1987) tested schedules to dry sorted incense-cedar squares. Yichun et al. (1996) dried thick hem-fir lumber sorted into groups based on species and basic density. Wallace et al. (2003) studied the effect of sorting on the stability of western hemlock exposed to Japanese winter conditions. Avramidis et al. (2004) performed a study to characterize how combinations of lumber sorting, drying, and redrying of Pacific coast hemlock affect yield and drying times. Sugimori et al. (2006) studied the effect of initial MC, green weight, heartwood ratio, and heartwood color as sorting parameters for drying of sugi (Cryptomeria japonica) lumber.
In the area of modeling, drying of sorted lumber was simulated with mathematical models that take into consideration the lumber MC variability. In particular, Cronin simulated MC distribution in lumber drying by using Markov Chains (Cronin et al. 1997) and the theory of random variables (Cronin et al. 2002, 2003). Kayihan (1984, 1985) used Monte Carlo simulation to predict MC variability in lumber drying, and Elustondo et al. (2004, 2005) and Elustondo and Avramidis (2003, 2005) simulated lumber MC distribution in drying by using discrete probability analysis.
In an industrial operation drying of sorted lumber can be also complemented with drying schedules that are optimized for each lumber sort. The optimum dry-bulb [mm.sup.2] wet-bulb temperatures at different stages of the drying process depend on species, products, and initial MC. Lumber with high MC and low dry-ability require more conservative drying schedules to prevent stress and checks, and lumber with low MC and high dry-ability require shorter drying times to prevent over-drying and distortion. Therefore, the main objective of this study is to develop a methodology for optimizing kiln-drying of 114 by 114 [mm.sup.2] hem-fir timber by utilizing presorting coupled with modified drying schedules.
Materials and methods
Timber sorting was performed at a local sawmill in Port Alberni, BC, with the help of in-line NMI sorting equipment. NMI (Northern Milltech Inc., MC Pro 1500, DOS version) is a commercial capacitance type MC meter designed to sort green lumber before drying. The NMI uses four sets of two metallic plates placed above and below the wood to measure electric capacitance as the wood passes transversally at line speeds. Since electric capacity is affected by both MC and density, the NMI value is similar, but not necessarily equal to the lumber MC.
The sorting strategy consisted of rejecting all of the lumber with NMI > 75 percent, and then dividing the remaining lumber into three sorts referred to as low (L), medium (M), and high (H). NMI sorting points for creating the L, M, and H sorts were: minimum-NMI = 0 percent, low-NMI = 40 percent, high-NMI = 53 percent, and maximum-NMI = 75 percent. Out of 155,880 pieces, the NMI meter reported 129,807 pieces with NMI < 53 percent, 25,811 pieces with 53 percent < NMI < 75 percent, and 262 pieces with NMI > 75 percent. By assuming normal distribution, these numbers correspond to 42.2 percent average and 11.2 percent standard deviation (SD). The corresponding NMI normal distribution is shown in Figure 1. As is indicated in the figure, the L, M, and H sorts contain 40 percent, 40 percent, and 20 percent, respectively, of the total dried lumber.
Another three sorts were created at FPInnovations--Forintek in Vancouver, BC, by combining different amounts of L, M, and H lumber. These new sorts were referred as unsorted (L + M + H), wet-free (L + M), and dry-free (M + H). In total, six different lumber sorts were dried and six different drying times and final SDs were obtained. The measured results were analyzed mathematically using three empirical equations that were specifically developed for this study:
[R.sub.(low-sort)] = [R.sub.(mixed)] + A x [[X.sup.N.sub.low-MNI[left and right arrow]75%]]
[R.sub.(high-sort)] = [R.sub.(mixed)] + B x [[X.sup.N.sub.0%[left and right arrow]high-MNI]]
[R.sub.(medium-sort)] = [R.sub.(mixed)] + A x [[X.sup.N.sub.high-NMI[left and right arrow]75%]]
+ B x [[X.sup.N.sub.0%[[left and right arrow]low-NMI]] + C
x [[X.sup.N.sub.high-NMI[left and right arrow]75%]] x [[X.sup.N.sub.0%[left and right arrow]low-NMI]]
X = the fraction of pieces that are contained in each sort, and
A, B, C, and N = constant parameters that are adjusted to fit the experimental drying time and SD by the method of least square error.
[FIGURE 1 OMITTED]
For simplicity, both drying time and SD are represented by the same letter R (standing for result), but they are treated as two different sets of experimental results. After adjusting the parameters, the equations were used to predict R for other hypothetic sorting strategies where low-NMI and high-NMI are the arbitrary sorting points.
Drying runs were conducted in a laboratory 0,7 [m.sup.3] conventional kiln. Each kiln charge contained 42 pieces of 0.9-m-long hem-fir lumber arranged in a single six layers by seven pieces package. The numbers of L, M, and H pieces used for each sort, as well as the corresponding sorting points, average MC and SD, are reported in Table 1. The lumber layers were spaced with standard 19-mm stickers, and the air velocity throughout the layers was approximately 3 m/sec.
To minimize random variability, the laboratory lumber sorts were prepared with matched specimens. Lumber sorted at the sawmill was 4 m long, thus it was possible to cut three matched 0.9-m pieces from each 4-m lumber piece. Additionally, four 25-mm-wide samples were cut at the ends of each 0.9-m section for determination of initial MC. All of the 0.9-m pieces were weighed with an electronic balance, labeled, end-sealed with glue, and put into plastic bags to prevent MC losses before drying. The pieces were stored in a cold room at around 0[degrees]C temperature, and they were allowed to warm-up inside the plastic bags before drying. The weight, width, and thickness of each piece was measured before and after drying to calculate final MC, basic density, and shrinkage.
MC after drying was also determined by using the ovendry method. After drying, two adjacent 25-mm samples were cut from the center of each dried piece. One sample was used to measure the average MC and the other sample was used to measure the core MC. The core was defined as the lumber zone below 25 mm from the external surfaces, and the zone between the core and the external surfaces was defined as the shell. To reduce the experimental error, the shell MC was calculated mathematically on the basis of the measured average and core MC. Lumber quality after drying was estimated on the basis of a visual inspection. Surface checks were classified as "none" for clean surface, "low" for hairline checks, and "high" for checks wider than 1 mm. Degrade was then defined as the percentage of pieces with collapse or a "high" level of checks.
[FIGURE 2 OMITTED]
The unsorted lumber (L + M + H) was dried with a time-basis schedule consisting of an initial 6 hours warm-up ramp, seven successive steps in which dry-bulb and wet-bulb gradually increased every 24 hours, and a final step at a constant 78[degrees]C dry-bulb and 65[degrees]C wet-bulb temperature that ended when the average lumber MC was approximately 15 percent. The experimental dry-bulb, wet-bulb, and MC curves measured during drying of unsorted lumber are shown in Figure 2. The other five sorts were dried with a normalized MC-basis schedule developed on the basis of the drying curve measured for unsorted lumber. The normalized MC-basis steps valid for all of the sorts were determined from the average MC of unsorted lumber measured at the end of each time-based step. The definition of normalized MC is shown in the Equation below, where MC = 33 percent was the average MC of unsorted lumber at the end of step 7.
Normalized MC = MC - 33%/ Initial MC - 33%
Results and discussion
The generic MC-basis schedule is reported in Table 2. Similar to the time-basis schedule, the MC-basis schedule also includes 6 hours of warm-up and a final step that begins when the MC = 33 percent. It ends when the MC = 15 percent. The experimental drying time and final MC and SD measured for the six drying runs are reported in Table 3. Additionally, the experimental drying curves measured for the six laboratory lumber sorts are shown in Figure 3; the black lines are the experimental MC curves and the grey lines show the experimental time and MC measured at the end of each step. As can be observed, drying time ranges from 17.2 days (for the H sort) to 9.1 days (for the L sort),
Average thickness shrinkage was between 2.5 percent and 2.8 percent depending on the experiments, and the results suggested that shrinkage is not considerably affected by the sort initial MC. Experimental shell and core MC after drying are shown in Figure 4. The data suggest that the shell and core MC difference tends to zero when the average MC tends to the equilibrium moisture content (EMC). Figure 4 also includes two discontinuous lines indicating that the core MC is lower than the theoretical fiber saturation point (FSP = 30%) when the average MC is lower than approximately 20 percent. Therefore, MC = 20% was used as the wet-limit for which it can be assumed that the lumber core is below FSP.
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
Degrade is shown in Figure 5 as a function of the sort initial MC. As can be observed, degrade increases proportionally to the amount of L pieces contained in the sort. One possible explanation for the higher degrade associated with L lumber is that the NMI capacitance meter tends to concentrate low-density lumber in the L sort. Since low-density lumber has a higher concentration of juvenile wood, then the L sort is more susceptible to checks. The experimental results showed that the average basic density was 395 kg [m.sup.-3] for the L sort, 423 kg [m.sup.-3] for the M sort, and 433 kg [m.sup.-3] for the H sort, and the percentage of pieces containing the pit (juvenile wood) was approximately 20 percent in the L sort, 5 percent in the M sort, and 2 percent in the H sort.
[FIGURE 5 OMITTED]
[FIGURE 6 OMITTED]
[FIGURE 7 OMITTED]
For the purpose of this study, only the total drying time and the final SD were analyzed to optimize the sorting strategy. Regression parameters for total drying time and final SD are reported in Table 3. In addition, Figures 6 and 7 show the comparison between experimental and calculated drying time and SD as a function of the fraction of total lumber contained in each sort. To facilitate visual comparison, the six experimental data points were indicated with dots and the six calculated data points were connected with straight lines.
Final SD was not used directly to optimize sorting, but it was used to calculate the percentage of wets. Because of the relatively small number of pieces involved in the experiments (42 pieces per run), the actual number of wets after drying ranged from 1 in the L sort to 6 in the H sort. This was not enough to find a meaningful relationship between sorts and wets. Therefore, the percentage of wets was estimated theoretically from the SD by assuming normal distribution.
To calculate the overall drying time and percentage of wets, the results for individual sorts were weighed on the basis of the fraction of timbers contained in each sort. The results of the calculations for overall drying time and percentage of wets are shown in Figure 8 and 9, respectively. These show calculated results for unsorted lumber and four hypothetic sorting strategies. Since the sawmill only dries lumber with NMI < 75 percent, the grey curve represents the case of two sorts with a cutting point indicated by the low-NMI. The other three curves represent sorting strategies using three sorts with cutting points indicated by the low-NMI and high-NMI values.
[FIGURE 8 OMITTED]
[FIGURE 9 OMITTED]
The optimum low-NMI and high-NMI sorting points that minimize overall drying time and percentage of wets are listed in Table 4. Figure 8 shows that by using only two sorts (high-NMI = 75%), sorting at low-NMI = 49 percent reduces the overall drying time from 13.6 to 12.6 days with respect to unsorted lumber. If three sorts are used, then the overall drying time can be potentially reduced to around 12.4 days by using a low-NMI = 45 percent and a high-NMI = 54 percent. Figure 9 shows that by using only two sorts (high-NMI = 75%) sorting at low-NMI = 42 percent reduces the percentage of wets from 19 to 8 percent with respect to unsorted lumber. If three sorts are used, then the overall percentage of wets can be potentially reduced to approximately 2 percent by using a low-NMI = 41 percent and a high-NMI -- 51 percent.
The average price of kiln-dried (KD) hemlock squares in September 2008 was approximately $760/Mfbm (1800 $/[m.sup.3]) (Madison's Canadian Lumber Reporter 2008), and wet lumber after kiln-drying is sold at a lower price depending on market demand. For this example, it is assumed that lumber price for wet lumber is around 20 percent of the average KD lumber price. With these prices, a basic economic analysis was performed for a typical British Columbia sawmill with four 250 Mfbm kilns and a kiln time utilization of around 80 percent (due to maintenance, loading, and unloading). Total lumber production was calculated on the basis of the total kiln volume, kiln utilization, and theoretical drying time. The production of KD and wet lumber was determined from the total lumber production by using the theoretical percentage of wets. Then, the annual lumber sales was calculated by multiplying the production of KD and wet lumber by the corresponding lumber prices.
The results of this basic economic analysis are reported in Table 5. As can be observed, annual lumber sales can potentially increase from M$15.7 to M$17.8 to $18.9 (13% to 20%) by implementing one of the optimized sorting strategies. This represents an annual sales increase of M$2.1 to $2.2 with respect to drying of unsorted lumber. In comparison with the present sawmill sorting strategy (two sorts), annual sales could potentially increase in M$0.7 to $0.8 by implementing three optimized sorts. The economic effect of MC-based presorting in lumber quality was not analyzed in this study, but it is expected that this practice will also reduce the amount of over-dry lumber, and this in turn will reduce shrinkage, warp, and cross-sectional distortion (that are causes of lumber downgrade through the planer).
In this study a method for optimizing lumber sorting was developed and tested for drying 114 by 114 [mm.sup.2] hem-fir lumber. The method consisted of sorting the entire lumber population into three sorts and then combining the sorted lumber into six subgroups that show the effect of the sorting parameters on the drying results. The 114 by 114 [mm.sup.2] hem-fir lumber was sorted at a local sawmill in British Columbia using the NMI capacitance-type technology. It was found that in comparison with unsorted lumber, sorting into three lumber groups can potentially reduce drying time by approximately 7 percent and recover approximately three-quarters of the lumber that would still be wet after drying unsorted lumber. This would represent a potential annual sales increase in the order of the million dollars depending on the kiln capacity and market conditions.
Another important result of this study is that by sorting 114 by 114 [mm.sup.2] hem-fir lumber with NMI technology, lumber with low-density tends to accumulate in the sorts having low NMI values. For this reason, lumber sorts having lower NMI values are more susceptible to checks and twist. On the positive side, lumber sorts having lower NMI values dry faster and with a lower MC dispersion at the end of drying. This result suggests that sorting 114 by 114 [mm.sup.2] hem-fir lumber with NMI technology must be coupled with optimized drying schedules that take into consideration that the low sort is easier to dry, but more susceptible to degrade and distortion.
Avramidis, S. and B. Hao. 2003. Automated lumber sorting for stability (ALSOS): Effect of slope of grain and annual ring orientation on the drying characteristics of hemlock baby squares. Rept. prepared by the Univ. of British Columbia for the Coastal Forest and Lumber Assoc. and the ZAIRAI Lumber Partnership, Vancouver, Canada. 25 pp.
--, J. Aune, and L. Oliveira. 2004. Exploring pre-sorting and re-drying strategies for Pacific coast hemlock square timbers. J. of the Inst. of Wood Sci. 16(4): 189-198.
--, S. Shida, A. Kounoutsakos, and D. Kobayashi. 2005. Stability study of square timbers in simulated house frame service in Japan. Forest Prod. J. 55(10):84-91.
Cronin, K.P., K. Abodayeh, and J. Caro-Corrales. 2002. Probabilistic analysis and design of the industrial timber drying process. Drying Tech. 20(2):307-324.
--, B. Norton, and J. Taylor. 1997. Drying lumber in kilns: A methodology for stochastic analysis using Markov modelling. Drying Tech. 15(3-4):765-790.
--, P. Baucour, K. Abodayeh, and A. Barbot Da Silva. 2003. Probabilistic analysis of timber drying schedules. Drying Tech. 21 (8): 1435-1458.
Elustondo, D.M. and S. Avramidis. 2003. Simulated comparative analysis of sorting strategies for RFV drying. Wood and Fiber Sci. 35(1):49-55.
-- and --. 2005. Comparative analysis of three methods for stochastic lumber drying simulation. Drying Tech. 23:131-142.
--, --, and L. Oliveira. 2004. Estimation of green moisture content distribution in hemfir timber by stochastic modeling. Holzforschung 58:413-417.
--, --, and R. Zwick. 2005. The demonstration of increased fiber utilization using optimized lumber sorting and radio frequency vacuum drying. Forest Prod. J. 55(1):76-83.
Holmes, S. and D.G. Arganbright. 1984. Green sorting incense-cedar for increased air-drying yard throughput. Forest Prod. J. 34(3):57-63. Kayihan, F. 1984. The process dynamics and the stochastic behavior of batch kilns. AIChE Symp. Series 246(81):104-116.
--. 1985. Stochastic modeling of lumber drying in batch kilns. In: Proc. of Drying '85, Washington, DC. Kozlik, C.J. 1987. Kiln-drying incense-cedar squares for pencil stock. Forest Prod. J. 37(5):21-25.
-- and J.C. Ward. 1981. Properties and kiln-drying characteristics of young-growth western hemlock dimension lumber. Forest Prod. J. 31(6):45-53.
Madison's Canadian Lumber Reporter. 2008. The source for Canadian and U.S. lumber and panel prices. Madison's Canadian Lumber Reporter 58(47): 14.
Nielson, R.W., J. Dobie, and D.M. Wright. 1985. Conversion Factors for the Forest Products Industry in Western Canada. Forintek Canada Corp., Vancouver.
Sugimori, M., K. Hayashi, and M. Takechi. 2006. Sorting sugi lumber by criteria determined with cluster analysis to improve drying. Forest Prod. J. 56(2):25-29.
Wallace, J.W., I.D. Hartley, S. Avramidis, and L.C. Oliveira. 2003. Conventional kiln drying and equalization of Western hemlock (Tsuga heterophylla (Raf.)[Sarg]) to Japanese equilibrium moisture content. Holz als Roh- und Werkstoff 61(4):257-263.
Western Wood Products Assoc. (WWPA). 1997. Hem-fir species facts. www.wwpa.org/hemfir.htm.
Yichun, Z., L. Oliveira, and S. Avramidis. 1996. Drying characteristics of hem-fir squares as affected by species and basic density presorting. Forest Prod. J. 46(2):44-50.
The authors are, respectively, Research Scientist and Group Leader, FPInnovations--Forintek, Vancouver, British Columbia, Canada (email@example.com). This paper was received for publication in March 2009. Article No. 10595.
Table 1.--Pieces, sorting points, initial MC, and SD for the six sorts prepared at the laboratory. High Initial Initial Sort L M H Low NMI NMI MC SD (no. of pieces) (%) H 0 0 42 53 75 93.2 19.9 M + H 0 29 13 40 75 80.5 23.4 M 0 42 0 40 53 70.5 17.8 L + M + H 17 17 8 0 75 70.7 28.1 L + M 21 21 0 0 53 62.0 22.2 L 42 0 0 0 40 50.1 11.4 Table 2.--Normalized MC-basis schedule. Wet-bulb Dry-bulb Normalized Step temperature temperature Time MC Actual MC ([degrees]C) (h) (%) Warm-up 49 49 6 -- -- 1 51 52 -- 0.9 -- 2 53 55 -- 0.8 -- 3 55 58 -- 0.7 -- 4 57 62 -- 0.6 -- 5 59 66 -- 0.4 -- 6 61 70 -- 0.2 -- 7 b3 74 -- 0.0 33 8 65 78 -- -- 15 Cool-down -- -- 6 -- Table 3.--Experimental drying time, and final MC and SD measured for the six drying runs. Drying Final MC Final SD time Final MC (oven-dried (oven-dried Sort days (load cells) samples) samples) (%) H 17.2 15.4 15.0 3.3 M + H 15.8 15.3 15.1 5.1 M 14.0 15.3 14.8 3.0 L + M + H 13.6 15.2 14.2 5.8 L + M 11.8 15.2 15.3 5.0 L 9.1 15.4 15.0 1.5 Table 4.--Equation coefficients for drying time and final SD. A B C N Drying time -6.66 4.13 0.35 0.73 Final SD -9.86 -2.91 -67.35 1.49 Table 5.--Theoretical results for different sorting strategies. Low High Drying Strategy (a) NMI NMI time Wets Production Sales (%) (days) (%) (MMfbm year) (M$ year) 1 0 75 13.6 19.2 21471 15.7 2 53 75 12.6 14.8 23175 17.1 3 45 54 12.3 5.1 23740 17.9 4 41 51 12.4 2.1 23548 17.8 (a) 1 = unsorted lumber; 2 = current sawmill strategy; 3 = minimum drying time; and 4 = minimum percentage of wets.
|Printer friendly Cite/link Email Feedback|
|Author:||Elustondo, Diego Miguel; Oliveira, Luiz|
|Publication:||Forest Products Journal|
|Date:||Sep 1, 2009|
|Previous Article:||Introducing lesser-known wood species from certified forests in Bolivia to the U.S. market.|
|Next Article:||Long-term drying behavior, dimension, and weight changes due to moisture cycling in wood-polypropylene composites.|