Application of ensemble machine learning models for predicting ultimate tensile strength of trembling aspen (Populus tremuloides) dimension lumber

Summary

No reviewed summary yet. Read the abstract below or open the original publication.

Abstract

Trembling aspen (Populus tremuloides) dimension lumber is a promising product for light wood-frame construction, but its ultimate tensile strength (UTS) remains difficult to predict due to high variability from natural defects. This study developed four ensemble machine learning models, random forest, extreme gradient boosting, gradient boosting regression tree (GBRT), and adaptive boosting, to predict UTS of aspen lumber. The dataset covered 321 specimens with a 38 mm × 89 mm cross section and 12 input features, including grade, geometry, specific gravity, moisture content, knot characteristics, and modulus of elasticity (MOE) from machine stress rating (MSR) and longitudinal stress wave tests, all measurable in sawmills. GBRT achieved the best performance (R2 = 0.718), with the minimum MSR-MOE as the most influential predictor. Against empirical models (R2 = 0.239-0.487) and multivariable linear regression (R2 = 0.585), GBRT improved R2 by 0.133-0.479, supporting industry-oriented AI for individual-specimen UTS prediction and economical material use.

Abstract as supplied by the source. Full text may be restricted by the publisher.