Recent advances in scanning technology have enabled the detection of surface defects in green boards. This report is the second phase of a project that was initiated to investigate the benefits of surface defect scanning (grade scanning) on the value of lumber that can be recovered from optimized edgers. The first study focused on BC Interior mills and found that little or no potential increase was available. In this study, the benefits for mills processing large logs into high value products (e.g. BC Coastal sawmills) were investigated. Sixty-six hemlock cants were evaluated, with cant widths from 12 to 27 inches - large enough to allow multiple sawing solutions. After a sawing solution was generated for each cant by an optimized shifting-saw gang edger, using only profile scanned data, a grader, taking visible defects into account, proposed alternate sawing solutions. The values of all sawing solutions were calculated and for 76% of the cants, a sawing solution proposed by the grader was more valuable than the one generated by the optimizer.
The study found that an average increase of 29% in the value of large cants was available when edging decisions took surface defects into account, in comparison with the value resulting from optimization based on cant profile alone. However, it should be noted that the accurancy of vision systems will likely be less than that of a human grader so the actual gains will likely be somewhat lower.
Les tests réalisés dans le cadre de ce projet ont généré des résultats sur la performance d’une nouvelle génération d’optimiseurs fabriqués par la compagnie Comact pour automatiser les opérations d’éboutage et de classification des sciages après rabotage. Ce système automatisé de classification communément appelé « GradExpert » est muni de caméras lasers pour la détection de défauts géométriques et de caméras couleurs pour la détection de défauts visuels.
Deux échantillons de sciages de ÉPS séchés rabotés en provenance de l’est du Canada et un troisième contenant des sciages séchés rabotés de pin lodgepole affectés par le dendroctone du pin ont été utilisés pour évaluer l’efficacité de ce système automatisé de classification.
Nos résultats indiquent que la technologie utilisée par Comact pour traiter de l’aspect géométrique des sciages permet de classifier avec une précision étonnante les défauts tels que la flache, l’omission, le grain arraché et les courbures. Nos résultats stipulent une efficacité de 92,7 % lorsque le système automatisé traite ce type de défauts. La technologie de vision utilisée pour identifier et traiter les défauts visuels comme les nœuds, les fentes les roulures, la carie, les trous de vers et certains défauts de fabrication a donné des résultats inférieurs, mais tout de même satisfaisants. Seuls les résultats obtenus lors de la détection de la coloration, des cassures transversales, de la déviation du fil et du bois de compression ont été moins probants. L’efficacité de détection des défauts visuels a été évaluée à 64,2 %. Des résultats similaires ont été obtenus avec les sciages affectés par le dendroctone du pin, ce qui laisse présumer que ce type de matériel peut être traité adéquatement avec ce système automatisé.
Des tests ont également été réalisés afin de comparer la performance d’une opération de classification semi-optimisée équipée d’un classeur linéaire en interaction avec un poste de classification manuelle et le système automatisé de la compagnie Comact. Selon nos résultats, le système automatisé GradExpert a commis près de quatre fois moins d’erreurs que le système semi-optimisé. Pour ce qui est de l’efficacité en volume et valeur, le système automatisé GradExpert a obtenu une note presque parfaite, soit une efficacité de 99,9 % en volume et 99,5 % en valeur, alors que le système semi-optimisé a généré respectivement 104,1 % et 96,4 % du volume et de la valeur optimale de l’échantillon.
Les pertes en valeur ont été réduites de façon considérable. Elles sont passées 13,30 $/Mpmp pour le système semi-optimisé à seulement 1,90 $/Mpmp pour le système automatisé, soit un gain de 11,40 $/Mpmp. Basé sur une production au rabotage de 100 MMpmp et sans tenir compte d’une réduction anticipée de personnel, ce gain représente un surplus annuel de 1,14 millions de dollars.
Forintek, Tembec, Comact et l’Université McMaster, ont réalisé le développement du Smart Mill Assistant (SMA). Le SMA est un système fonctionnant en temps réel qui détecte rapidement le procédé hors contrôle. Ce système aidera les gestionnaires des scieries et leur personnel de supervision à détecter, identifier et corriger les déviations du procédé.
Le SMA est basé sur l’acquisition et l’analyse de données en temps réel. L’analyse statistique multivariée et des méthodes d’analyse simple et de visualisation de données, ont été utilisées pour la construction des modèles de monitorage et de diagnostic. Le système SMA surveille le procédé à deux niveaux. Dans un premier temps, le procédé a été observé dans son ensemble. Le modèle développé permet de détecter lorsque le procédé est hors contrôle et d’identifier la variable ou le centre de transformation en cause, sans toutefois identifier les causes du problème. Dans un deuxième temps, un système expert a été développé autour d’un centre de transformation, la débiteuse à scies multiples. Les modèles développés pour ce système permettent de détecter le problème, d’en identifier les causes et de fournir des solutions aux usagers.
Le SMA a été implanté dans une scierie pilote de Tembec. Les modèles implantés permettent d’estimer théoriquement un gain potentiel de 5 à 10 $/Mpmp. Ces gains varieront selon le type d’équipement en place, le temps de réaction du personnel de la scierie et la capacité à corriger les problèmes. L’évaluation des gains réels n’était pas complétée au moment de la rédaction du rapport.
Forintek, Tembec, Comact and McMaster University have developed the Smart Mill Assistant (SMA), a real-time system that rapidly detects when the process is out of control. The SMA will assist sawmill managers and supervisors with detecting, identifying and correcting process deviations.
The SMA relies on the acquisition and analysis of process data in real time. Multivariate statistical analysis, simple data analysis and visualization methods have served as a basis for the development of monitoring and diagnostic models. The SMA monitors the process at two levels. The first model monitors the overall process. It can detect when the process is out of control, and identify the variable or machine centre responsible for a problem, but it does not, however, identify the factors responsible for the problem. At the second level, we have developed an expert system for one machine centre, the bull edger, which is capable of detecting problems, identifying the causes for them, and suggesting solutions to mill users.
The SMA has been implemented in a Tembec pilot sawmill. Based on the models in place, we estimate that gains of $5 to $10/Mbf are theoretically achievable. Actual gains will vary with the type of equipment, personnel reaction time and its ability to correct problems. At the time of report preparation, the evaluation of actual gains was not totally completed.
Today's machine centres are being increasingly automated but often operate as a collection of isolated machines run by a variety of computer systems. Clearly, such heterogeneous computing and control environments present a formidable barrier to the problem of interoperability. Already there are vendors that provide a partial solution to the problem, since they provide methods of interoperability only between machines that they supply. Vendor-specific methodologies are in general proprietary, and do not inter-operate with any other vendor's equipment. What's needed to facilitate widespread machine-to-machine data exchange is a universal methodology to connect to optimizer data, or any data for that matter, with plug-and-play simplicity. In order to enable enhanced data availability and also to lay the foundations for the evolution of process monitoring and control in the sawmilling industry, this project was undertaken to create a common methodology for vendor-neutral data exchange between machine centres, process monitoring and control systems, and business systems.
A task forceª, with members drawn from sawmilling and equipment vendor companies, selected the well-established specifications for data exchange published by the OPC Foundation, a consortium of companies committed to universal data exchange in industry. While these specifications specify standards-based methods for data exchange, the task force recognized that there was an additional layer required to create standard plug-and-play access to sawmill optimizers. This additional standardization layer specifies exactly what data is made available per optimizer type. After testing these ideas for primary breakdown optimizers and PLCs in a sawmill-based pilot project, the task force unanimously adopted the OPC specification and our per-optimizer layer as a practical standard for data exchange in the sawmilling industry.
Given this initial success, however, there needs to be a continuing effort to ensure that the evolving sawmill standards eventually are applied to all optimizer types, and that sawmill managers and executives are aware of their benefits. Continuing effort must ensure that multi-vendor support per optimizer type does not result in tag list fragmentation which would undermine the benefits of standards. The methodologies adopted during this project will never become standard in the sawmilling industry unless the majority of sawmillers demand the standard OPC optimizer interfaces defined by this project.
ª In this document, “task force” is used interchangeably with “working group”. On 21 March 2002, a standards committee was struck from task force members, but soon lost its meaning when the task force adopted an email list approach to collaboration. The email list was much more inclusive and therefore much larger, and became the defacto “working group”. By project end, the working group consisted of 40 members.
In this study, extensive veneer compression tests were conducted to examine the transverse compression behaviour of veneer at both ambient and controlled temperature and moisture content (MC) environments. Based on the results, a novel method was developed to characterize overall surface quality of veneer and other wood materials in terms of their bondability and compression behaviour.
The method would have significant implication in both theory and practice. In theory, the general wood compression theory would need to be modified. The revised wood compression theory would include four stages instead of commonly defined three. The first stage, which has long and so far been overlooked but is critically important, could be named as “non-linear conformation”. During this stage, the contact area increases nonlinearly with the load applied. It is this stage that directly reveals the interfacial bonding behaviour of wood materials such as veneer-to-veneer and strand-to-strand and their minimum compression required for achieving adequate contact (bonding). In practice, the method provides a fast and objective way of evaluating surface roughness/quality of veneer and other wood materials. The new method also establishes the maximum compression allowable for achieving the best panel performance in terms of bonding strength, stiffness and dimensional stability. Based on the concept of this method, it was further found that both minimum compression required and maximum compression allowable are independent of temperature and MC, which provides a direct benchmark to the material recovery during panel hot-pressing.
In a case study with Trembling aspen veneer, the variation of veneer surface roughness/quality and its effect on resulting material recovery were first revealed. Then, the optimum panel densification was identified for performance plywood and LVL products based on the frequency distribution of the minimum compression required and the maximum compression allowable. Finally, an overall veneer quality index was established to compare veneer overall quality for different species/thickness. The method shows good potential in practical applications for increased material recovery, reduced glue consumption and improved panel performance.
Over the past thirty years the Canadian forest industry has implemented optimizing technologies at most machine centres within its sawmills. Studies have shown that the impact of this variety of hardware and software has been an approximate drop of 20% in the volume of raw fibre required to produce the same volume of lumber. To remain competitive however, the industry must make a significant shift from higher volume to increased value recovery.
The purpose of this “technology intelligence” tour of European research organizations, technology suppliers and sawmills in September 2003 was to identify and view opportunities for the transfer of advanced scanning and optimization technologies and programs that could ultimately assist the B.C. wood processing industries to extract higher value from their indigenous resource.
Numerous opportunities have been identified for the transfer of commercially available technology, beta-testing of advanced technologies, and increased scientific collaboration.
The report includes summaries of the individual visits as well as several specific recommendations based upon conclusions drawn as a result of the tour.
VGrader, Veneer Grading Optimizer, was developed at Forintek to assist mills to optimize on-line veneer stress grading operations. So far, more than 10 copies of VGrader 1.0 software have been delivered to Forintek member mills. The software can recommend the optimum grading thresholds through analyzing the properties of veneer to help mills deal with “what-if” scenarios when veneer species, log source and diameter as well as final veneer products change. By tailoring veneer grades to the market requirements of LVL/plywood products, the software serves as a useful tool to characterize specific veneer for end use and help optimize veneer on-line stress grading and products lay-up options.
During the past year, the VGrader software has been upgraded to deal with either UPT-based (ultrasonic signal propagation time) veneer stress grading or E-based (modulus of elasticity) veneer stress grading or veneer visual grading. The software has also been upgraded to accommodate UPT data either from mills or laboratory testing of veneer samples. A direct linkage between laboratorial measurement span and desired wheel-span of the on-line grading system was also setup. The current version of the software is VGrader 3.0. To help mills optimize current on-line stress grading operations, the proper procedures to find the optimum UPT thresholds were established.
The proper procedures are as follows:
1) Sample veneer sheets representative of veneer population in the mill and perform stress wave testing for sampled sheets using a portable stress wave timer. Alternatively, full-size veneer sheets can be sampled right after the on-line grading system with UPT data being recorded for each veneer sheet;
2) Measure other relevant veneer properties such as thickness, density, moisture and knots;
3) Calibrate the stress wave time (or UPT) to find its zero offset value;
4) Store all measurement data into a VGrader compatible database;
5) Use the upgraded VGrader software to examine the distribution of veneer attributes/properties such as thickness, UPT, density and MOE;
6) Derive required veneer MOE based on the performance requirements of target veneer products;
7) Establish stress grading constraints and using VGrader 3.0 to perform computerized veneer stress grading through adjusting the UPT or E thresholds and examining the change of statistical veneer MOE, densities and volume breakdown per grade until all the grading constraints are satisfied;
8) Convert the optimum set of UPT or E thresholds from the VGrader software into those used for on-line veneer grading system to perform stress grading;
9) Make veneer products and test them to validate the grading results.
An example of establishing the above procedures was also demonstrated.
In this project, 5 species of veneer from 4 mills comprising aspen, hemlock, incised Douglas-fir and spruce/lodgepole pine veneer were sampled and evaluated. Also, non-incised Douglas-fir veneer was assessed. A portable Metriguard laboratory unit was employed to measure the stress wave time for each piece of veneer sheet. Other relevant veneer characteristics such as density, moisture content and knot area were also measured. All veneer samples were visually graded according to CSA Standard O151-M1978. A computer database was developed to record all measured data.
A practical user-friendly computer software package VGrader 1.0 was developed to assess veneer sorting strategies. This software provides users with panel lay-up options in connection with veneer grading results. Users can assemble their desired veneer products using either visual grades or stress grades by mixing species, grades and thickness. Further built into this software is an end product strength prediction model which was calibrated with experimental results obtained throughout this research. An electronic user help manual is built into the software, which guides users through the operation of this software. The intent of the software is to provide users with a tool to assist users understand the relationship between veneer visual grades, stress grades and performance of their final veneer products. The tool can assist those seeking to develop new veneer based composites with predictable strength properties for engineered applications. The software can give quick answers to questions such as what percentage of specific veneer can be used for making a target product, and what the optimum stress-grading thresholds are. It can be used to adjust and calibrate mill stress grading operations to meet the market requirements of final products. It can also serve as a management tool for mill managers to optimize products mix and keep track of mill production. Further, it can recommend appropriate adjustments of on-line production when veneer species, log source, log diameter and final veneer products change.
The key results from this research are as follows:
Veneer properties vary from species to species, stand to stand, and from mill to mill. They further vary with block positions and from sap to heart to core. According to this study, there exist two groups among the veneer species studied. One group is Douglas-fir, aspen and hemlock, which are suitable for making LVL and high strength plywood; the other is mixed spruce/lodgepole pine, which is suitable for making plywood or using it as inner layers for LVL manufacture.
There is little or no correlation between veneer visual grades and stress grades. Hence, it is not accurate to visually sort veneer on a strength basis. The stress grading operation is threshold-dependent, which differs from visual grading in both strength properties and percentages of grade volume. Compared to visual grading, stress grading can sort veneer into distinct strength groups with much smaller variation for quality assurance, and can extract more high-grade veneer for high value LVL manufacture. To maximize the value of veneer products, the best strategy is to extract the strongest veneer via stress grading to make market-demanding LVL and use the rest to make either low-grade LVL or plywood. It is also strongly recommended that veneer/plywood operations first perform stress grading to sort veneer, followed by veneer visual grading. By combining stress grading with visual grading, high-grade or high-value plywood can be produced, and veneer panels requiring high visual grade face veneer combined with strength can be manufactured.
A significant correlation exists between veneer MOE and LVL edgewise MOE and MOR for all the species tested. However, the correlation between LVL flatwise MOE and MOR, shear strength and veneer MOE is less or much less significant and differs from species to species, and from mill to mill. A calibration with experimental data is needed when trying to predict panel MOR and shear strength with veneer MOE. Good correlations between plywood MOE and MOR and average MOE of veneer layers parallel to the testing span were identified for all the species tested, which can set up a benchmark for predicting the strength properties of structural plywood panels for engineered applications using stress graded veneer.
Using VGrader 1.0 software, an optimum set of veneer stress grading thresholds can be established, which makes it possible for adjustment and calibration of mill on-line stress grading systems based on requirements of market-oriented veneer products. By periodically sampling veneer, mill operations can be diagnosed and optimized, and mill profits can be maximized.
The objective of this project was to develop a new feed speed control system for band saws. This system regulates the feed speed based on the cutting depth measured ahead of the saw. In comparison to other feed speed control systems presently used in sawmill operations, the new system allows maximum feed speeds without overfeeding the saw and also avoids underfeeding. The equipment developed basically involves two laser light generators which place laser lines along the saw cut on the saw entry and saw exit side, two cameras which locate the position of the laser or saw lines, and a microprocessor which processes the information from the cameras, determines the feed speed based on the known relationship between sawing variables and gives the signal to the carriage drive. Considerable experimenting was required to accurately measure the distance between the laser lines on the saw log due to colour variations of the long surface and more so due to the irregular geometry of many saw logs. Also a high working speed of the system had to be achieved to correspond to the high feed speeds used in sawmill operations. The new feed speed control system was tested with Forintek's 5-ft band mill. It was found that the actual feed speed set by the feed speed control, and the calculated feed speed were in close agreement showing that the system is able to effectively control the feed speed based on the cutting depth. Additional work is needed and will be carried out to further refine the system before commercialization can be undertaken.
This report presents the results of the first stage of an investigation into the feasibility of developing a machine capable of automatically tensioning a bandsaw blade. The present work involves an experimental and analytical investigation of the effects of roll tensioning upon the cutting performance of the bandsaw. In order to understand the role of roll tensioning its effects on internal stress distribution, torsional and lateral natural frequencies and stiffness of the blade have been investigated. The results of strain measurements induced during different rolling patterns and with different thickness of plate and differing rolling pressures are presented and an analytical explanation of the results is given. Experimental results showing how the stiffness of the blade and its natural frequencies are affected by the roll tensioning are also presented. An accurate analytical model that relates the rolling pattern to the lateral stiffness has not been found. Cutting tests have been conducted in which the performance of a blade with no tension is compared with a blade with different levels of tensioning. The results of these tests are presented and indicate that the relationship between cutting accuracy and tensioning is very subtle.