Soyez le premier à aimer ceci
Industries with variable inputs, such as the forest product industry (FPI), the mining industry or the recycling industry, must cope with material uncertainty, which affects their ability to predict output yields. To deal with this, one can either reduce uncertainty at the source, or plan operations taking uncertainty into account. In the FPI, the first approach is generally used. For instance, the softwood lumber industry has adopted sophisticated transformation technologies that adapt sawing patterns to the log characteristic using scanners technology to acquire accurate information about work-in-process status. Another approach to reduce uncertainty is input material classification. Specific characteristics can be measured to classify input material and therefore reduce uncertainty within each class. However, whether the process involves logs, mining ores or recycled papers, material classification has a value and a cost according to how detailed it is performed. This research first proposes a method based on classification tree analysis to classify hardwood logs. Next, using agent-based simulation, it analyses the value of different classification strategies, from detailed, to no classification at all. Results show in the context of the hardwood lumber industry that the benefit of detailed classification is offset by its cost, while a relatively simple classification dramatically improves output yield.