Thesis (MScEng (Industrial Engineering))--University of Stellenbosch, 2011. / ENGLISH ABSTRACT: During the 2009 financial year the sawlog production from plantations in South Africa amounted to 4.4 million m 3 and sawn timber of R4.2 billion was produced from these logs. At the current average price for structural timber, a 1% increase in volume recovery at a medium-sized South African sawmill with an annual log intake of 100 000m 3 will result in additional profit of about R2.2 million annually. The purpose of this project was to evaluate the potential of increasing in value recovery at sawmills through optimization of the positioning of a log at the primary workstation by considering the internal knot properties. Although not yet commercially available, a high speed industrial log CT scanner is currently in development and will enable the evaluation of the internal characteristics of a log before processing. The external profiles and the internal knot properties of ten pine logs were measured and the whole log shape was digitally reconstructed. By using the sawmill simulation program Simsaw, explicit enumeration was performed to gather data. This data include the monetary value that can be earned from sawing the log in a specific log position. For every log a total of 808 020 sawing positions were evaluated. In the sawmill production environment only a few seconds are available to make a decision on the positioning of each log. Meta-heuristic optimization algorithms were developed in order to come to a near optimal solution in a much shorter time than that required when simulating all possible log positions. The algorithms used in this study include the Genetic algorithm, Simulated Annealing, Population Based Incremental Learning and the CrossEntropy method. An Alternative algorithm was also developed to incorporate the trends identified through analysis of the sawmill simulation results. The effectiveness of these meta-heuristic algorithms were evaluated using the sawmill simulation data created. Analysis of the simulation data showed that a maximum increase in product value of 8.23% was possible when internal knot data was considered compared to using conventional log positioning rules. When only external shape was considered a maximum increase in product value of 5% was possible compared to using conventional log positioning rules. The efficiency of the meta-heuristic algorithms differed depending on the processing time available. As an example the Genetic algorithm increased the mean product value by 6.43% after 200 iterations. Finally, a method to evaluate the investment decision to purchase an internal scanning and log positioning system is illustrated. / AFRIKAANSE OPSOMMING: Gedurende die 2009 finansiële jaar is daar 4.4 miljoen m 3 rondehout op plantasies in Suid Afrika geproduseer en saaghout ter waarde van R4.2 biljoen is hieruit vervaardig. Met die huidige gemiddelde prys vir strukturele hout, kan ‘n 1% verhoging in volumeherwinning by ‘n gemiddelde grootte saagmeul in Suid Afrika met ‘n jaarlikse rondehout inname van 100 000 m 3 ‘n bykomende wins van R2.2 miljoen lewer. Die doel van hierdie projek was om die potensiële verhoging in waardeherwinning by ‘n saagmeul te evalueer, indien die posisionering van ‘n stomp by die primêre werkstasie geoptimeer word deur interne kwas eienskappe in ag te neem. Kommersiële CTskandeerders word tans nog nie hiervoor aangewend nie, maar ontwikkelinge in tegnologie sal dit moontlik binnekort prakties moontlik maak om die interne karakteristieke van ‘n stomp te evalueer voor prosessering. Die eksterne profiel en interne kwas eienskappe van tien Pinus rondehout stompe is gemeet en die al tien stompe is digitaal geherkonstrueer. Met behulp van die saagmeulsimulasieprogram, Simsaw, is 808 020 verskillende saagsimulasielopies uitgevoer. Elk van hierdie simulasielopies het ‘n ander beginposisie gehad in terme van rotasie, skeefheid en horisontale verskuiwing. Die finansiële waarde wat verdien kan word deur ‘n stomp in ‘n sekere posisie te saag is telkens bepaal. In die saagmeulomgewing is daar slegs ‘n paar sekondes beskikbaar om ‘n besluit te maak oor hoe ‘n stomp geposisioneer moet word. Meta-heuristiese optimisering algoritmes is ontwikkel om ‘n naby optimale oplossing te bepaal in ‘n baie korter tyd as wanneer alle saagposisies geëvalueer word. Vyf verskillende meta-heuristiese algoritmes is teen mekaar opgeweeg. Vier van hierdie algoritmes is bestaande heuristieke wat vir verskeie ander optimeringsprobleme ingespan word. Die vyfde algoritme is spesifiek vir doeleindes van hierdie projek ontwikkel om die neigings wat tydens die data-analise van die saagmeulsimulasie geïdentifiseer is, te inkorporeer. Die effektiwiteit van hierdie meta-heuristiese algoritmes is bepaal deur van die saagmeul simulasiedata wat gegenereer is gebruik te maak. Analise van die simulasiedata toon dat ‘n maksimum toename in produk waarde van 8% moontlik is wanneer interne kwaseienskappe ook geïnkorporeer word tydens besluitneming teenoor die konvensionele stompposisioneringreëls. Wanneer slegs die eksterne stompprofiel in ag geneem word, is ‘n maksimum produkwaardeverhoging van tot 5% moontlik teenoor resultate wat verkry word met konvensionele stompposisioneringsreëls.