Machine learning for structural response assessment in timber engineering: a review on mechanical performance, durability, and fire resistance

Summary

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

Abstract

Timber, a renewable and low-carbon building material, is pivotal in sustainable construction. Effective structural response assessment (SRA) is crucial for ensuring the safety and serviceability of timber structures. Traditional SRA methods, including experimental, theoretical, and numerical methods, face limitations due to the inherent variability and anisotropy of timber. Machine learning (ML), as a data-driven paradigm, offers a potent alternative for addressing these challenges. This paper presents a comprehensive review of ML-based SRA in timber engineering. It critically examines state-of-the-art ML applications in four aspects: static mechanical properties, seismic performance, durability, and fire resistance. It is found that current research reveals a clear trend: the field is shifting from applying traditional ML algorithms to predict mechanical properties of wood towards employing advanced deep learning (DL) and ensemble algorithm (EA) models to solve complex problems related to engineered wood products. Furthermore, SRA in timber engineering faces small-data challenges due to high variability of wood, and the lack of publicly available benchmark datasets, which significantly hinders progress in the field. Finally, this paper explored potential future research directions for addressing existing challenges.

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