Real-Time Value Optimization of Edging and Trimming Operations for Rough, Green Hardwood Lumber

For edging/trimming operations in hardwood sawmills, an operator examines both sides—or just the waney-edged side—of each board, and makes a quick assessment of grade potential. The operator then decides where to edge and/or trim the board to achieve the intended grade and, presumably, maximum value. However, human operators can only achieve lumber values that are 62-78% of optimal. To achieve higher performance rates, computer-aided processing is needed to search through the millions of possible edging/trimming options. Using a computer to solve the edging/trimming problem means that board data must be provided from a scanning system. We have developed a prototype scanning system for rough, green hardwood lumber that can automatically describe a board and its defects. Those data are then used in a branch-andbound (B&B) search for an optimal edging/trimming solution. In comparing this B&B algorithm with known maximal solutions for a suite of boards, we found that the B&B method attains the correct maximal solution for 92% of the boards and achieves 99% of the total maximal lumber value. Ninety-four percent of the boards attained the maximal value in less than 10 seconds (233 MHz processor). For all boards for which the algorithm obtained a maximal value, it did so in less than 8 seconds. Currently available computers can easily reduce that search time to 1-2 seconds per board. In comparing the B&B algorithm to the edging/trimming option in UGRS (the only other software that performs edging/trimming), our method has better accuracy (1-2% higher value recovery, 5-11% higher correctness) and is 40-100 times faster. The high value recovery of the B&B algorithm essentially eliminates that source of error for an automated edging/trimming system. Performance is therefore only limited by scanning accuracy.