Printer Friendly




In recent years, a lot of effort has been devoted to implementing the concept of an integrated forestry-wood chain. Good management of such a complex system requires a clear understanding of the effects that different factors involved may have on the final output of the chain. Raw material databases and simulation software capable of spanning the wood-processing operations have now been developed and are being used as a support tool in understanding and optimizing the forestry-wood chain. This paper reports results of simulation tests using the virtual SawMill software and the Swedish Pine Stem Bank database. The simulated scenario tries to mirror a very customer-oriented production philosophy. The bucking, the sawing, the crosscutting, and the board-grading procedures were simulated for different end-user requirements and a statistical model was built. The purpose of the model was to investigate the theoretical sensitivity of the final product to parameters such as external sawlog features (taper, ovality, bow), harvester measurement accuracy (for length and diameter), sawing pattern optimization, and log positioning in the saw line. Special emphasis was put on evaluating the role of the harvester within this "puzzle." The results show that small improvements of the harvester's measuring performance could lead to considerable improvements in the wood transformation chain.

In Scandinavian countries, the cut-to-length timber harvesting systems have been used for many years, and currently this kind of forest harvesting procedure is the most important one in use. Almost 90 percent of the total volume of harvested timber coming from the Scandinavian forests is obtained with modern harvesters equipped with automatic log measuring and bucking control systems [16]. As the product-oriented concept gains ground worldwide in the forest industry, the short-wood strategy (crosscutting in the forest and sawlog allocation to the sawmills) is becoming more and more common in many European countries and North America.

A cut-to-length system incorporates two machines: a harvester and a forwarder. The harvester consists of a crane, carrier, processing head, and a measuring system. The measuring system allows the operator to buck the stems to diameters and lengths [12,17] that match the product specification made by the sawmills, which in turn try to meet customers' specific needs for grade, wane, width, thickness, or length.

An increased need for timber of certain lengths (along with well-defined wane criteria) is a stringent reality in the timber production chain and occurs for the majority of wood products [11] ranging from blanks to building components. In the case of the short-wood strategy, where the "correct from the start" philosophy is leading, the relationship between the desired length of the timber and the length of the sawlog is particularly straightforward, except for situations where finger-jointing occurs. The sawlog length places a definite upper limit for the length of the sawn timber. Hence, the crosscutting operation performed by the harvester in the forest is of crucial importance for the whole downstream process.

The harvester's length-measuring device has a certain measurement accuracy [13-15], which along with the diameter measurement accuracy [17] will have an influence on the precision of the bucking operation and thus downstream on the final length of the sawn timber.

It is of major importance to find out in which way the harvester's measurement performance affects the final length of the sawn timber. The sawing process also encompasses some procedures/factors that may affect the dimension and the final length of the sawn product, e.g., edging and trimming criteria, trimming after drying, accuracy of log positioning towards the saw blades, sawing pattern allowance, etc.

Thus, the aim of this study was to put together a virtual harvester and a virtual sawmill in a simulation approach that tries to minor the short-wood strategy.

The simulation technique offers the possibility to change the value of the different measurement errors (i.e., accuracy of the harvester length- and diameter-measuring device) and the levels of other implicated factors (i.e., trim allowance, wane tolerance), and consequently to assess the influence of those factors on the final length of the sawn products.

The results should reveal the importance of each factor and its connections to others, thus allowing us to track where improvements have to be made within the chain to optimize the short-wood method.

The Swedish Pine Stem Bank (SPSB) [8] is a large database containing detailed information about 200 Scots pine (Pinus sylvestris L.) trees. The project is based on computed tomography (CT) scanning of these 200 stems, which were carefully chosen from 33 well-documented sample plots all over Sweden. From each sample plot six trees were taken: two small, two medium-sized, and two large. After harvesting the selected trees, the logs obtained were graded by skilled log graders and then CT-scanned in a fourth-generation medical tomograph (Siemens SOMATOM AR.T.).



Once the scanning step was performed, the logs were sawn with a normal sawing pattern (cant sawing), and subsequently all boards produced were dried to 18 percent moisture content. All centerboards were scanned on all four sides using a CCD line camera. Then, two skilled lumber graders independently graded the centerboards according to both the old [1] and new [2] grading system commonly used in Sweden. The complex database represented by the SPSB encompasses all silvicultural and stand data, images from sample plots, all CT and CCD-images, as well as other useful measurements.

The images achieved through CT-scanning of logs were analyzed automatically by using image analysis algorithms [6)]. These images describing the logs consist, however, of a large amount of information. In order to reduce the amount of data, a method for parameterization of the log has been developed [10]. The parameter files issued as a result describe the outer shape of the log and the heartwood border using polar coordinates having the pith as origin. One radius at every degree, every 10 mm along the log describes the outer shape while a mean radius for 12 degree sectors every 10 mm along the log describes the heartwood border. The location of the pith is given every 10 mm along the log by using an X to Y reference system. The description of every knot (location, size, and type) is made by using 11 parameters acquired from CT-images by semiautomatic image-processing algorithms. Table 1 shows a synthesis profile of external features such as diameter, length, taper, ovality, and bow of the 625 sawlogs from t he SPSB.


The virtual SawMill is a sawing simulation software that is able to utilize the digitized logs acquired from CT-scanning and stored in the SPSB. The program is capable of reconstructing a 3-dimensional representation of the outer shape and the internal structure of the log and of generating boards through a sawing procedure that is easily controlled by the operator. When generating boards, vSM can identify internal and external defects such as knots and wane. The grading procedure is based on explicit grading rules [2] and the value for each generated board is determined. Once the sawing procedure is completed, a detailed sawing report is available. A previous study has successfully dealt with the validation of the vSM at the single-log level [7], and even a large-scale validation approach of the vSM along with the SPSB has been carried out versus a real sawmill yield [4].

The vSM reads the logs from a SPSB CD-ROM and automatically creates an internal database from which different log selections can be made. The user interface allows not only the 3-dimensional visualization of the log, but also allows the visualization of diverse sorts of sawing parameters. A main feature of the code is its open architecture: the vSM is endowed with a particular programming aid (WoodScript) that allows users to adjust the program according to their specific needs. If, for instance, only the outer shape of the log is in focus, the knot structure can be easily removed and different hypotheses can be tested in a nondestructive environment through sawing simulation. Wane criteria adjustment or automatic batch sawing mode are some other facilities that vSM offers.


The study was based on the simulation technique, as the SPSB and the vSM were employed. With these tools in hand we have tried to simulate an appropriate scenario upon the short-wood strategy, namely when the harvester is to supply the sawmill with logs having a certain length. Figure 1 schematically illustrates the bucking operation. The desired length for a given sawlog is influenced by three elements:

* the length covering the maximum number of accommodated cutting modules, in this case 300 mm, (a x M);

* the desired sawlog offset/trim allowance (Dlo), calculated as half of the cutting module (150 mm), which is added to the (a x M) length;

* the harvester's length measurement accuracy, which causes the final length of the sawlog to become shorter or longer (with the quantity X) than the desired length.

In a condensed mathematical form, the final length of a log can be expressed as follows:

Final length = Target length +/-

X = (a x M + Dlo) +/- X, (mm)

This formula describes, in fact, exactly what happens in the real situation in the forest when the harvester performs the bucking operation.

Figure 2 shows the structure of the SPSB concerning the log offset parameter. The length measurement accuracy parameter was based on this criterion. It means that, for example, a log with 80 mm offset is 70 mm shorter than the desired length (which is expected to have 150 mm offset), and in the same way a log with 220 mm offset is 70 mm longer than the desired length. Hence the measures -70 mm and +70 mm express, in fact, deviations from a target length that can be attributed to the harvester's length measuring device. The log offset of the 625 logs from the SPSB permitted us to simulate the length measurement accuracy with a standard deviation (SD) of 4.3 cm.

An important feature of the vSM is that it not only has great flexibility in setting the sawing parameters, but the parameters for the log diameter measurement can also be modified. It is possible to include in the simulations different measurement errors for diameter, and thereby a harvester diameter measuring procedure can be simulated.

In this study, an SD of 6 mm was used to simulate the diameter measurement accuracy for the harvester. For both length and diameter deviations, it was assumed that it was possible to compensate for systematic errors, i.e., mean errors after compensation equal to zero, and that these measurement errors are characterized by a normal distribution, N(0, [[sigma].sup.2]).

Figure 3 shows a quite uniform distribution of the sawlogs from the SPSB when plotting them in a diameter-length measurement accuracy coordinate system. There are some moderate outliers and at least five strong outlier logs for which the diameter and length measurements deviate more than three times the SD. These logs were, however, kept in the study as they might illustrate a part of reality where quite large measurement deviations occur [13].

The next "scene" in the scenario reproduces the next step in the short-wood strategy where the harvested sawlogs (supposed to have the requested length) arrive at the sawmill and the sawing procedure occurs. In Table 2 the sawing pattern for each log's top-end diameter interval is listed. The chosen diameter classes correspond to the ones in use at a Swedish sawmill. In the sawing process, there are obviously factors that may directly affect the final length of the sawn product, e.g., wane exclusion, trimming module, inaccurate positioning, or centering of the log face to saw blades.

This study considered the case of a very customer-oriented production where the focus is put only on the length of the sawn timber. In such a situation, the traditional grades (based on the Nordic Timber Grading Rules) are no longer relevant. Only the permitted length deviation and the wane requirements are now leading criteria for the timber production. In other words, the external features of the log (e.g., length, diameter, bow, ovality, taper), correctly chosen sawing pattern, and correct sawing procedure (log orientation and centering) dominate over the knot structure of the log. The sawing simulations completed in this study used the logs without the knot structure and a curve sawing was performed. A standard trimming (for log-end cracks; damage of logs during harvesting, handling, or storage; and board drying checks) of 50 mm was included in the simulations.

As shown previously, the conceived scenario was adjusted to keep as far as possible the same attributes as a real situation. One simulation consisted of sawing all 625 logs from SPSB. Finally, the way factors involved in the scenario influence the number of off-grade boards was investigated, i.e., the boards shorter than the target length, (a X M), as shown in Figure 1. At this stage of the study, only the centerboards were considered.


Three different simulations were completed, the differentiation criteria being the wane requirements. Boards not fulfilling the respective wane criteria are automatically trimmed to the nearest length module, and thus the risk of becoming off-grade product occurs. In Simulation 1, the most tolerant regarding accepted wane, the produced boards had to fulfill the normal wane requirements (A, B, and C grade) as defined by the Nordic Timber Grading Rules (1). In Simulation 2, only boards fulfilling the wane requirements for grade A were allowed. In Simulation 3, the strictest, only boards with less than 3 mm wane passed the "off-grade" test.


The results from the three simulations (Fig. 4) are presented in connection with

Figure 2. Within each simulation, the percentage of off-grade boards was calculated for each log offset class. A sawlog was considered to give off-grade boards if at least one of the resulting centerboards was an off-grade product.

For each simulation, the same measurement accuracy for diameter (SD =6 mm) and length (SD = 4.3 cm), as shown in Figure 3, were used, as well as the same material (the 625 logs from SPSB). The same sawing pattern and the same sawing procedure were applied. Thus the results are compatible for comparison.

For all three simulations, the diameter measurement accuracy (SD = 6 mm) resulted in almost 29 percent of the logs being incorrectly sorted with respect to their diameter. Approximately half (45%) of these incorrectly sorted sawlogs fell into a greater diameter class, while the other part (55%) was underestimated and sorted into the lesser diameter class.

In the conditions of Simulation 1, it turns out that 18.54 percent of all the sawlogs gave off-grade boards. The share of the off-grade boards in the total number of centerboards was 8.94 percent.

For Simulations 2 and 3, the proportion of the sawlogs giving off-grade products was 21.95 and 37.2 percent, respectively. The share of off-grade boards in the total number of centerboards rose to 11.6 and 17.85 percent, respectively.

As Figure 4 shows, the standard trimming (ST), in this case 50 mm (Fig.1), seems to be a sharp boundary when considering the off-grade boards. It means that in order to get as few off-grades as possible, the difference between the ST and the desired log offset/trim allowance (Dlo), in this case 150 mm (Fig.1), should accommodate the harvester's error magnitude for length measurement.

Figure 5 illustrates the risk (as % probability) of obtaining off-grade boards (for normal wane criteria requirements) from logs with different target log offsets. The graphic is based on the probability density function of the normal distribution (3) assumed for the harvester's length measurement deviations. Three different length measurement accuracies for the harvester have been simulated:

* the first, with an SD = 3.3 cm, calculated under the assumption that the difference of 100 mm between the ST (50 mm) and Dlo (150 mm) would be able to accommodate the harvester's length measuring error, i.e., 100 [greater than or equal to] 3 x SD or SD [less than or equal to] 3.3 cm;

* the second one, with SD = 2.5 cm, which seems to be a quite good approximation of the precision that machines have today in the forest [13-15];

* and the third one, with an improved measurement accuracy (with an SD as low as 1.5 cm), which can be seen as an ambition for the constructors.


It was of interest to find out which factors (from the measurement process, from the sawing procedure, or physical parameters associated with the sawlog) might explain the behavior of the offgrade curves after the 50-mm limit in Figure 4. Hence, a model based on the PLS (Projection to Latent Structures based on Partial Least Square) regression technique was built. This kind of approach was preferred instead of the conventionally used Multiple Linear Regression (MLR) because the PLS method can cope with nonindependent variables and with noise in the data, which is frequently the case when working with biological material like wood, with measurement processes, and with large data sets [5]. For carrying out the PLS analysis, the software program SIMCA-P-7.01 was used.

The model was built for the data and conditions specified in Simulation 2. The y-variable (response) was the presence or absence of off-grade boards after sawing. The x-variables (predictors/factors), questionable at this stage for the behavior of the off-grade curve, were as presented in Table 3.

The objective with this model was to extract the possible interesting information contained in the set of previously mentioned factors that are able to explain the variance of the y-variable (the off-grade response). In the PLS technique, [[r.sup.2].sub.y] (called coefficient of determination or goodness of fit) is a parameter that quantifies the model's ability to explain the variance of the y-variable. When the modeled variable is of a dummy (binary) type, a new parameter, the goodness of classification (C), can be defined. It measures the accurateness of the model when classifying the observations with respect to the binary response (zero or one).

Another important parameter in the context of a PLS model is [[Q.sup.2].sub.y] (called goodness of prediction), which is the result of the cross-validation technique [5] and indicates the model's ability to predict future observations.

The trade-off between [[r.sup.2].sub.y] and [[Q.sup.2].sub.y] is an efficient and robust tool for assessing multivariate models. Large gaps between these two parameters (differences larger than 0.2 to 0.3) reveal that the model explains random variations in the data while close values for [[r.sup.2].sub.y] and attest that the model describes true connections between x- and y-variables.

As a result of the PLS analysis, a statistical model was fitted to the data from Simulation 2. The variation in the data could be explained by one principal component (PC). The goodness of fit for this model was [[r.sup.2].sub.y] = 0.622, the goodness of prediction was [[Q.sup.2].sub.y] = 0.610, and the goodness of classification was C = 0.934.

The coefficients of the fitted linear model are graphically displayed in Figure 6. They show how each of the nine factors (x-variables) influences the off-grade response (y-variable). The graph is based on the coefficients centered (mean equal to zero) and scaled. SIMCA uses the unit variance (UV) scaling technique, which ensures that each scaled variable is given equal variance.


With the help of the SPSB and the vSM software, the simulation approach used to conduct this study made two important things workable: 1) it was possible to simulate the log measurement procedure (for both length and diameter) performed by a harvester in the forest; and 2) the downstream steps of the process followed by these sawlogs (diameter sorting, log breakdown, board grading) were all simulated.

Figure 4 (where three different wane requirements and trim allowances were simulated) distinctly shows that logs with a trim allowance less than 50 mm will always give off-grade boards. This is understandable as the ST parameter applies by default to all sawlogs. However, the trend and the form of the curves after this limit of 50 mm are quite unusual. It would be rather expected that the proportion of off-grade boards decreases progressively as the log offset increases.

A careful analysis of Figure 4 reveals a common pattern that exists in the appearance of all three curves. The only thing that distinctly differentiates them is the overall share of off-grade boards obtained when sorting with respect to different wane requirements. This value went from 8.94 percent in Simulation 1 up to 11.6 and 17.85 percent in Simulations 2 and 3, respectively. This step up of the curves from Simulations 2 and 3 is quite evident and explicable: the stricter the wane criterion is, the greater the probability of getting off-grade products.

Finally, these curves would further suggest that some factor(s) other than the sawlog offset are responsible for the off-grade occurrence after the 50-mm limit. As the simulations have been run through the same material (SPSB), with the same simulator (vSM), and based on the same sawing procedure, the most probable variables that could influence the results would be the external features of the sawlogs, some particular sawing parameters, or a combination of both.

The fitted model based on the PLS algorithm, which took these factors into consideration, was able to describe 62.2 percent of the variability within the behavior of the off-grade curve from Simulation 2. In other words, there is still variation that is not described by the model. This variation is caused by some other factors that were not included in the analysis. Some of them can be linked to the outer shape of the logs. Small local variations/irregularities of the outer shape of the logs, i.e., at every cross-section level, could be very significant in this case. The taper, the bow, and the ovality as used at this stage were calculated with an averaging procedure along the log. It would be an advantage if a more accurate description of a log's shape irregularities could be achieved, e.g., taper on different parts of the log, ovality on the top-part of the log, and bumpiness. The vSM's open architecture and the log features extent in the SPSB are elements that make this kind of achievement possible.

Another cause for the unexplained variation could be the distribution (number and evenness) of the logs in different categories. Despite the large number of SPSB sawlogs used in the simulations, the very tight separation into classes with different log offsets (Fig. 2) or with different sawing set-ups (Table 2) meant that in fact few logs were present in each of these classes. This would further suggest that an eventual enlargement of the SPSB would be a justifiable and useful prospect.

All these factors just discussed, which could be the causes for the unexplained variation, set certain limitations for the accurateness of the fitted PLS model. However, the fact that [[r.sup.2].sub.y] (0.622) and [[Q.sup.2].sub.y] (0.610) have very close values indicates that the model describes a true relationship between the considered x-variables and the off-grade response.

As the coefficients plot shows (Fig. 6), the most important variables of the model are the performance in positioning the log in the saw line during the second saw (Dlp-Ss) and the matching between the log diameter and the set-up diagonal (Dlog-Diag). The Dlp-Ss and Dip-Fs variables are positively correlated with the response; i.e., better log positioning will appreciably decrease the risk of getting off-grade products. The variables Dlog-Diag and Dmin-Diag are negatively correlated with the response, which suggests that an optimization of the sawing classes and of the sawing set-ups within each class should contribute to the attenuation of the off-grade phenomenon.

The Bow and the Ovality of the log are next in importance. Both are positively correlated with the level of the response, which means that logs with great ovality and/or bow present a high risk of producing off-grade boards. It turns out that the Taper of the log has nearly no effect at all on the occurrence of off-grade boards. This judgment is reasonable, as the sawing procedure completed by the vSM included automatic length-wise centering and only the centerboards were considered in the study.

The harvester's performances regarding diameter and length measurement accuracy, L-ma and D-ma, are important for the model and are negatively correlated with the response, i.e., the higher the measurement accuracy is, the less the probability of getting off-grade products. It appears that for the studied scenario, the diameter measurement has greater importance than the length measurement performed by the harvester. Good diameter measurement accuracy will increase the proportion of correctly sorted logs, which along with a judicious sawing set-up optimization will significantly contribute to the reduction of the off-grade proportion.

Nevertheless, when judging the importance of L-ma and D-ma one has to keep in mind that the model was built just for the observations for which the magnitude of the length measurement error made the log offset longer than ST (50 mm) (Fig. 4). Thus the importance of the L-ma variable is to some extent linked to the chosen trimming scenario.

Therefore, an independent assessment of the L-ma variable's influence on the off-grade proportion is displayed in Figure 5, where a harvester with three different L-ma variables was simulated. It turns out that when practicing bucking with a large log offset (120 to 150 mm), the improvement of the harvester's length measurement accuracy (from 3.3 to 2.5 and 1.5 cm) has a negligible, but positive, influence on off-grade occurrence. This positive influence rapidly becomes considerable when the log offset used is reduced to less than 120 mm.

The majority of the harvester types currently used in the forest have an L-ma close to 2.5 cm (14,15). At the same time, many Swedish sawmills use a sawing procedure where the cutting module is fixed to 300 mm and thus requires sawlogs with 150-mm offset from the raw material suppliers. This practice vis-a-vis the log offset is quite common, but it does not mean that it is the best. Figure 5 shows that for a regular harvester (L-ma = 2.5 cm), the reduction of the log offset from 150 to 100 mm will increase the risk of getting off-grade products by only 2.5 percent. Nevertheless, in most of the contracts where the products' length is in focus, 2.5 to 3 percent is a commonly tolerated allowance for boards exceeding the permitted length deviations.

As Figure 5 also indicates, an improvement of the harvester's length measurement accuracy with just 1 cm (L-ma = from 2.5 to 1.5 cm) should make it possible to decrease the necessary log offset to 80 mm for the same off-grade risk. For the same log offset (100 mm), this improvement will cause the off-grade risk to drop from 2.5 percent to off-grade risk-free production. Hence, when assessing the economic potential of such an improvement, at least two elements should be considered: 1) the sawlogs will have lengths closer to the target lengths of the sawn timber, which means smaller volume and thus lower investments for the raw material and transportation; and 2) fewer products will fall into the off-grade category, and consequently higher sales profits will be achieved through increasing the volume of such profitable sawn products.

Finally, from the point of view of the wood-conversion industry, this will mark a step forward towards avoiding the "sawmill paradox" [9] through the synergistic effects of a better utilization of the wood raw material and an efficient transformation process. Of course, the economic gain hitherto stated could be modified if sawlogs of specific lengths and with "length warranty" will be more expensive to purchase. Still, the experience from some Swedish sawmills shows that paying more for the right raw material even increases the profitability of the process if a product-oriented philosophy is leading the production.


By using the simulation technique based on the SPSB and the vSM, it was possible to tackle the question regarding the role of the harvester within the forestry-wood chain. Within the framework of the simulated scenario (the short-wood method), the harvester's intrinsic performances regarding measurement accuracy appear to be very important. Rather small improvements of the measurement devices would lead to an appreciable improvement of the wood transformation chain and inevitably to economic gains.

An overall consideration of the results suggests some directions for future work. Further studies should take into consideration some other factors that might play a substantial role in the wood chain, e.g., variation of the trimming module or even more flexible grading criteria for both logs and boards. The potential gains of using a more accurate description of the outer shape of the sawlogs in the industrial process should be investigated as well. Finally, it appears that an enlargement of the SPSB could make this research tool more powerful and increase its usefulness for further nondestructive studies.

The authors are, respectively, Post-Graduate Student, and Professor, Lulea Univ. of Technology, Dept. of Wood Technology, Skelleftea Campus, Skeria 3, S-93187 Skelleftea, Sweden. This paper was received for publication in April 2000. Reprint No. 9109.

(*.) Forest Products Society Member.


(1.) Anonymous. 1982. Guiding Principles for Grading of Swedish Sawn Timber. 4th ed. Assoc. of Swedish Sawmill Men, Stockholm, Sweden. 36 pp.

(2.) _____. 1997. Nordic Timber Grading Rules. 2nd ed. Arbor Pub., Stockholm, Sweden. ISBN 91-7322-227-5. 64 pp.

(3.) Box G., W. Hunter, and J.S. Hunter. 1978. Statistics for Experimenters. An Introduction to Design, Data Analysis, and Model Building. John Wiley & Sons, Inc., New York. ISBN 0-471-09315-7. pp. 47-50.

(4.) Chiorescu, S. and A. Gronlund. 1999. Validation of a CT-based simulator against a sawmill yield. Forest Prod. J. 50(6):69-76.

(5.) Eriksson L., E. Johansson, N. Kettaneh-Wold, and S. Wold. 1999. Introduction to Multi-and Megavariate Data Analysis using Projection Methods (PCA & PLS). Umetrics AB, Umea, Sweden.

(6.) Grundberg, S. 1994. Scanning for internal defects in logs. Licentiate thesis, Lulea Univ. of Technology, Skelleftea, Sweden. 1994:14 L. ISSN 0280-8242.

(7.) _____ and A. Gronlund. 200_. Validation of a virtual sawmill. Accepted to Holz als Rob- und Werkstoff.

(8.) _____, _____, and U. Gronlund. 1995. The Swedish Stem Bank - a data base for different silvicultural and wood properties. Res. Rept. TULEA[BRI] 1995:31. Lulea Univ. of Technology, Skelleftea, Sweden. ISSN 0347-0881. 14 pp.

(9.) Gronlund, A. 1992. Sawmill Technology. Sveriges Skogindustriforbund. Yrkesbok X-724. Swedish Union of Forest Industry, Stockholm, Sweden. ISBN 91-7322-7242 (in Swedish).

(10.) _____, L. Bjorklund, S. Grundberg, and G. Berggren. 1995. Manual for Pine Stem Bank. Tech. Rept. 1995: 19 T. Lulea Univ. of Technology, Skelleftea, Sweden. ISSN 0349-3571 (in Swedish, English summary). 25 pp.

(11.) Lindgren, J.-E. 1987. Length considerations when selecting timber for different purposes. Res. Rept. The Swedish Univ. of Agri. Sci., Uppsala, Sweden. ISBN 91576-3031-3 (in Swedish, English summary). 204 pp.

(12.) Linehan, P.E. and T.J. Corcoran. 1994. An expert system for timber harvesting decision making on industrial forest lands. Forest Prod. J. 44(6):65-70.

(13.) Plamondon, J.A. 1999. Log-length measurement errors with six single-grip harvester heads. Wood Harvesting. Tech. Note 297. Forest Engineering Res. Inst. of Canada, Vancouver, B.C., Canada.

(14.) Sondell, J. 1995. Evaluation of five bucking-to-value systems for harvesters marketed in Sweden. IUFRO XX Congress, P3.07-session 2. IUFRO, Vienna, Austria.

(15.) Tufts, R.A. and R.W. Brinker. 1993. Productivity of a Scandinavian cut-to-length system while second thinning pine plantation. Forest Prod. J. 43(11/12):24-32.

(16.) Uusitalo, J. 1999. Predicting wood quality with harvester data. In: Proc. of the 3rd JUFRO Workshop on Connection Between Siviculture and Wood Quality Through Modeling Approaches and Simulation Software. G. Nepveu, ed. September 5-12, 1999, La Londes-Les-Maures, France. IUFRO, Vienna, Austria. pp. 549-554.

(17.) Wilhelmsson L. and J. Arlinger. 1997. How worthy is it to measure the "correct" diameter with the harvester? Res. Rept. No. 336. The Forestry Res. Inst. of Sweden, Uppsala, Sweden. 11 pp. (in Swedish, English summary).
 External features of the 625 sawlogs from the
 Swedish Pine Stem Bank.
Variable Diameter Length Bow Ovality Taper
 (mm) (mm/m)
Min. 122 3210 0 0 2
Max. 374 5530 42 32 24
Min./Max. 0.32 0.58 0 0 0.08
Mean 196 4460 9 4.49 9.18
Median 191 4490 7 4 8
Standard deviation 50 435 6 3 3
Skewness 0.58 -0.24 1.34 1.6 1.07
 Sawing pattern (cant sawing)
 and the Diff parameter for
 each log top diameter interval. [a]
Log top diameter Sawing pattern Diff
 Min. Max. First saw Second saw Dmin - Diag
 120 139 19, 75, 19 19, 38, 38, 19 5.94
 140 154 19, 100, 19 19, 38, 38, 19 7.23
 155 175 19, 100, 19 19, 50, 50, 19 5.43
 176 193 19, 125, 19 25, 50, 50, 25 7.83
 194 205 19, 125, 19 19, 63, 63, 19 7.42
 206 225 19, 150, 19 19, 25, 50, 50, 25, 19 17.63
 226 239 19, 175, 19 25, 25, 50, 50, 25, 25 16.29
 240 253 19, 175, 19 25, 25, 63, 63, 25, 25 15.26
 254 267 19, 200, 19 25, 25, 63, 63, 25, 25 8.45
 268 288 19, 200, 19 19, 25, 75, 75, 25, 19 5.54
 289 319 19, 225, 19 19, 25, 75, 75, 25, 19 8.49
 320 400 19, 225, 19 19, 25, 50, 50, 50, 25, 19 1.39
(a.)Diff is a parameter associated with each diameter interval
and for eash set-up and is the difference between
the minimal diameter of the interval (Dmin) and the diagonal
of the set-up (Diag) with respect to the centerboards.
 Abbreviation and description of the
 variables from the fitted PLS model.
Abbreviation Variable description
Taper Decrease of the diameter along the sawlog (mm/m).
 The diameter was measured each 10 mm.
Ovality Average difference between the largest and the
 smallest diameter of the oval cross section along
 the sawlog (mm).
Bow The distance between the highest point of the log
 curvature and the line joining the centers of the
 log ends (mm).
H - Dma The harvester's diametr measurement accuracy (mm).
Dmin - Diag The difference between the minimal diameter of the
 sawing class (Dmin) and the diagonal of the chosen
 set-up (Diag).
Diog - Diag The difference between the diameter of each log
 (Dlog) and the diagonal of the diagonal of the
 matching set-up (Diag).
H - Lma The harvester's length measurement accuracy (mm).
Dlp - Fs Deviation in positioning of log face to the saw
 blades in the first saw (mm).
Dlp - Ss Deviation in positioning of log face to the saw
 blades in the second saw (mm).
COPYRIGHT 2001 Forest Products Society
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2001 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Swedish Pine Stem Bank database; simulated sawmill
Publication:Forest Products Journal
Geographic Code:4E
Date:Feb 1, 2001

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters