Multivariate modeling improves quality grading of sawn timber

The quality grades are what determines the value of sawn timber. Therefore the grading process is essential for the profitability of a sawmill. At a modern sawmill in northern Sweden, a CT Log Computed Tomography is used in the saw line to optimize the cutting solutions by virtual 3D reconstruction of the log features. By adjusting the position of the log according to the optimal solution before cutting, the aim is to increase the quality and final resale value of the sawn timber. However, measurement errors in the virtual and final grading systems cause inconsistencies that decrease the agreement in grading. The grading process uses a rule-based system based on the Nordic Timber Grading Rules, which depends strongly on the size and shape of knots. If knots are measured incorrectly they could falsely exceed the allowed value for a certain quality, resulting in an inaccurate quality grade. The results from this initial project, show that using multivariate modeling instead of the traditional rule-based grading system improves the agreement between the virtual and final grading. The accuracy in grading increases with up to 19%, resulting in an agreement of 73%. A better agreement between the two systems would allow the process to take advantage of the full potential of the CT, increasing the profitability of the sawmill. The results are promising, but before implementing the method in the sawmill further testing and development have to be done to ensure optimal improvement