Browsing by Subject "laaturaja"

The right wood raw material should be delivered to the right wood processing mill at the right time, while minimizing costs. In other words, the log demand distributions and quality requirements of mills should be fulfilled. Bucking control tries to solve two problems; what kind of wood assortments to cut from each stand, and what dimensions to cut within each wood assortment. The success of bucking is important, as it is difficult to fix a poor bucking outcome. Modern harvesters collect stem data, from which virtual trees can be created. The bucking of these virtual trees can be simulated with different product instructions and settings. A bucking simulator can be utilized in predicting and optimizing bucking outcome. However, without quality-information of stems a bucking simulator is overly optimistic. Stem quality-information utilized in previous studies has been laborious to collect, and there has been little validation of log distributions simulated with stem quality-information. The object of this study was to determine whether a method, where stem quality-information is derived from harvester-collected stem and log data, could be utilized to improve the accuracy of simulated wood assortment recoveries and distributions. The study focused on Metsä Fibre’s sawmill in Vilppula and clear-cut stands. The usability of the method in bucking control and wood procurement in Finland was of central interest. Results were examined based on the theoretical framework of the study, which outlined the significance of the results to business operations. The bucking of real stems registered by harvesters was simulated with real quality-information derived from harvester-collected data, predicted quality-information derived from the previously mentioned quality data, and without any quality-information. The bucking outcomes were compared to real bucking outcomes registered by harvesters. The focus was on validating the real quality-information derived from harvester-collected data, while the predicted quality-information worked as an example. Without quality-information the simulated total log distribution and the real total log distribution had a match of 84.2 % (butt log: 86.4 %; top and middle log: 80.8 %). With the real quality-information this percentage increased by 2.9 percentile points (butt log: 2.4 pp; top and middle log: 2.4 pp). With the predicted quality-information the percentage increased by 1.6 percentile points (butt log: 1.4 pp; top and middle log: 2.7 pp). The relative bias and the RMSE of the simulated log recovery were 19.0 % (42.6 m³) and 24.6 % (55.1 m³) with no quality-information. With the real quality-information the relative bias and the RMSE of the simulated log recovery were -0.4 % (-1.0 m³) and 4.1 % (9.1 m³). With the predicted quality-information the relative bias and the RMSE of the simulated log recovery were 2.0 % (4.4 m³) and 12.3 % (27.6 m³). The real quality-information was derived from data that was cost-effectively collected by harvesters. Quality-information derived from harvester-collected data enables the consideration of complex dimensional and qualitative requirements, and the derivation of quality-information predictions, in a situation where clear quality-sections of stems can’t be defined.