Investigating the influence of lumber sample subsets on simulated rough mill part yields.

摘要

Simulation techniques are widely used to simulate the cut-up of lumber in the secondary wood products industries to make operational, tactical, and strategic decisions. One challenge when conducting this kind of simulation is the determination and assembly of a sufficiently large lumber set (i.e., how many boards to include in the database, also referred to as lumber sample). The minimum required size of the lumber set is dependent on cutting bill part sizes and quantity requirements. If large cutting bill quantity requirements are to be simulated, there may not be enough sample boards available in the digital lumber database. One way to cope with this problem is to use the same lumber subset several times. However, this may influence the results of the simulation. To date, except for some empirical observations, detailed research on this problem has not been conducted. Two approaches exist for manipulating lumber set sizes: (a) creating a lumber set that is large enough to satisfy a cutting bill, or (b) generating a smaller lumber subset and using it repeatedly until the cutting bill is satisfied. T-test results indicated that mean yields are not statistically different when using either one large lumber set generated from different boards or a smaller lumber subset that is used repeatedly. From a practical viewpoint, however, it is recommended to include at least 40 BF in a lumber sample subset so that the lumber yield variation within replicates is small. For lower grade lumber (e.g., 2 A Common), a larger lumber subset (e.g., 313 BF) is recommended to minimize variance within replicates. This study examines these approaches to the lumber subset problem for four National Hardwood Lumber Association lumber grades.