Unsupervised machine learning has recently gained attention in structural health monitoring as engineers seek methods that can interpret large and complex data sets without prior labeling. Traditional diagnostic approaches often rely on predefined models or manual analysis, which limits their adaptability and efficiency when dealing with evolving structural behaviors or unforeseen conditions. Despite the growing interest in this domain, the literature remains fragmented, with limited systematic and bibliometric reviews that consolidate progress, identify prevailing trends, and clarify methodological limitations. This study addresses this gap through a comprehensive systematic and bibliometric review of research on unsupervised machine learning applied to structural health monitoring. The review examines the evolution of techniques, their applications in various structural systems, and the frequency and quality of their adoption in scientific literature. The aim is to present a consolidated understanding of how unsupervised learning contributes to data-driven monitoring, fault detection, and predictive maintenance. This research is important as it provides an evidence-based overview that supports future methodological improvements and guides the selection of suitable algorithms for practical engineering applications, ensuring a stronger link between data science and structural reliability.
Alariyan et al. (2026) studied this question.