Value recovery and production control in the forestry-wood chain using simulation technique

Sammanfattning: This thesis deals with value recovery and production control in the forestry–wood chain for improved competitiveness of sawmills through higher profit and better adaptation to product requirements of the customers. The subject was approached using simulation technique with a database of virtual logs and a sawing simulator capable of processing the logs in the database. CT images of young Scots pine (Pinus sylvestris) sawlogs were processed with artificial neural networks (ANN) for identifying knots in sapwood where the contrast in the images is low. ANN classifications were deemed a feasible method where traditional image analysis methods failed. Further processing of the classified image allowed for parametric descriptions of the logs in a format compatible with the previously established Swedish Pine Stem Bank (SPSB). Static models of stem shape and interior knot structure were used to create stems that were also compatible with the SPSB. Processing the stems with the sawing simulator demonstrated the possibility of predicting timber grade recovery and volume yield from stands based on site, stand and tree characteristics. It was also shown that timber values in logs can be predicted using variables derived from 3-dimensional (3D) scanning of stems’ external geometry as well as from 3D scanning in combination with X- ray log scanning. The improvement achieved with the combined scanning was rather low compared to using 3D scanning alone. Results of a study of bucking methods, log sorting methods and production control showed that the more detailed information the bucking and log sorting decisions are based on, the higher the value recovery. Furthermore, the more processing stations involved in production control, the better are the demand targets met. In a study aiming at increased share of target board lengths, different bucking alternatives were evaluated. It was concluded that optimizing forest operations, value recovery and production as separate entities will not produce optimal results. A case study of a sawmill where the length of the timber was of high interest showed that increasing the share of target lengths of small dimensions can only be done at a relatively high cost in terms of volume yield loss. It was also shown that log classes should be defined with varying diameter limits for different log lengths at the conventional diameter-based log sorting. In order to meet the desired length distribution of the timber, it is necessary to alter the log length distribution, and this can be done with adaptive control heuristics that dynamically updates control log prices during bucking. It is concluded that there is an unexploited value potential in the forestry–wood chain which can be reached using modern measurement techniques and that a better characterization of the wood raw material will facilitate an improved customer-order orientation.