PSI - Issue 78

Available online at www.sciencedirect.com

ScienceDirect

Procedia Structural Integrity 78 (2026) 25–32

© 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of XX ANIDIS Conference organizers Keywords: Reclaimed timber ; Visual grading ; Bending strength ; Finite-element modelling ; Stochastic simulation ; Machine-learning classifier ; Circular economy ; Seismic reuse Abstract Reclaiming load-bearing timber from demolition sites can significantly reduce the environmental footprint of new construction; however, the absence of grading rules tailored to reclaimed members often forces high-value material down the waste hierarchy. This work introduces a visual-based classification framework that quantifies the bending-strength loss caused by small through holes, typical defects of previous service, and assigns each beam to either a reusable or non-reusable category. A stochastic, plane-stress finite-element model, experimentally validated on four-point bending tests, generated more than 1 . 5 × 10 4 random hole configurations in beams of three common cross-sections. From the elastic stress fields, a dimensionless reduction factor, k red , was derived to relate hole geometry to residual capacity. The simulated dataset fed two classifier families: (i) machine-learning algorithms (logistic regression, linear SVM, random forest and XGBoost) and (ii) an intentionally simple conditional rule based only on the sum of hole diameters in two visually identifiable depth zones. The conditional rule, with thresholds t 1 = 12mm for the outer quarter-depth strips and t 2 = 28 mm for the central strip (holes with d < 3 mm ignored), achieved a balanced accuracy of 0.80 and an F 2 , score of 0.82, on par with the best machine-learning models while remaining transparent and easy to apply. All misclassifications were conservative: no under-strength beam was ever labelled reusable. The resulting procedure provides a fast and camera-friendly grading aid that can unlock the safe and high-value reuse of reclaimed timber, advancing circular economy goals in the construction sector. XX ANIDIS Conference A visual-based classification approach for regrading reclaimed timber Angelo Aloisio a, ∗ , Dag Pasquale Pasca b , Audun Øvrum b , Massimo Fragiacomo a a Department of Civil, Construction-Architectural and Environmental Engineering, Universita` degli Studi dell’Aquila, 67100 L’Aquila, Italy b Norsk Treteknisk Institutt (Norwegian Institute of Wood Technology), Børrestuveien 3, 0373 Oslo, Norway

1. Introduction

Earthquakes leave behind large piles of timber from damaged buildings. Reusing this wood keeps it out of landfill, meets the EU circular-economy goals, and helps when new timber prices jump after a quake [2]. The main obstacle is

∗ Corresponding author. E-mail address: angelo.aloisio1@univaq.it.

2452-3216 © 2025 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of XX ANIDIS Conference organizers 10.1016/j.prostr.2025.12.004