From scan to decision: bucking, rotation and the size of the gains

A scanner earns money only when its data change a decision. This lesson looks at four decisions and the gains reported for them. Read every figure with the course rule from Lesson 1: region, sample, basis. The basis for all four studies below is simulation or modelling, not a before-and-after mill trial.

Simulation versus real-time optimisation

Simulation studies take scanned logs and test sawing strategies on a computer after the fact. They answer "how much could this be worth?" Real-time optimisation runs inside the mill's cycle time: the scan arrives, the optimiser searches candidates, and the result goes to the positioning equipment before the log reaches the saw. Simulations give a ceiling. The achievable fraction depends on positioning accuracy, on scanner errors and on whether the downstream process can execute the plan. The open studies reviewed here do not report cycle-time budgets for production-speed optimisation.

Four studies

What each result teaches

• Diminishing returns (Finland). If a later step already uses internal data, collecting it earlier may add cost without value.
• Price spread matters (Douglas-fir). The 20% scenario needs wide price differences between grades; a mill whose products sell at similar prices would see something closer to the 4% scenario. Positioning error was modelled on purpose.
• Algorithm matters (Appalachia). The better search method found about four more percentage points than the heuristic.
• Worst logs gain most (hardwood CT). Gains by grade were 27% (Grade 1), 47% (Grade 2) and 97% (Grade 3), because information about defects changes the best cut most where logs are worst.

Worked example: the algorithm gap

Lin et al. report 10.01% with the heuristic and 14.21% with dynamic programming.

gap = 14.21 - 10.01 = 4.20 percentage points

If a hypothetical mill sawed lumber worth 2,000,000 EUR a year at its current baseline, 4.20 points would be worth 2,000,000 x 0.042 = 84,000 EUR a year only if the simulation result transferred, which it need not. The 2,000,000 EUR is an invented example; the study gives no such figure.

Worked example: scenario thinking

Rais et al. report mean gains of 4%, 8% and 20% under three price scenarios. For a mill, the right question is not "which number is true?" but "which scenario looks like my price list?" Check the price difference between your top and bottom grades before quoting 20%.

Bucking inputs

A Swedish study (Barth et al., 2015) compared airborne laser scanning with traditional field inventory as input for bucking simulations on 16 stands (139.3 ha). Average stand-level error index was 0.15 for airborne scanning against 0.18 for traditional inventory; at total wood-flow level the traditional data gave a lower error index because the scanning tree lists were biased. Better input data gave small, not dramatic, improvements.

Checklist before believing a gain

• Is it simulated or measured in a mill?
• What is the baseline (horns up, diameter-based bucking, heuristic)?
• How many logs and which species?
• Which price spread or grade mix?
• Was positioning error included?

Key points

• The four reported gains (up to 5%, 4-20%, 10-14%, 46% mean) are simulations; none is a measured before-and-after mill result.
• The spread, from 4% to 97% for Grade 3 hardwood, shows the gain is situational: species, log quality, price spread and existing process all matter.
• Gains grow with price spread between grades and with poorer logs.
• Better search algorithms can add percentage points without new sensors (Lin et al.: 14.21 vs 10.01).
• Use published gains as an argument to measure your own baseline, not as a forecast.