Mahalanobis Nearest-Neighbour Index for Vibration-Based Damage Detection under Environmental and Operational Variability: Validation on a Full-Scale Concrete Bridge

Articles in Press

Document Type : Research Paper

Authors

Department of Civil Engineering, Se.C., Islamic Azad University, Semnan, Iran

10.22124/jcr.2026.34412.1754

Abstract

Damage detection in operating civil structures, particularly concrete bridges, using vibration responses under environmental and operational variability (EOV) remains a major challenge, because these effects can mask or distort damage-induced dynamic changes. This study presents an unsupervised framework based exclusively on healthy reference data for damage detection using nearest-neighbour indices. The reference space is constructed from healthy observations, and inspection samples are evaluated using Mahalanobis and Euclidean distance metrics. In this approach, each inspection sample is compared with its nearest neighbour in the healthy reference space, and the resulting distance score is used as a damage-sensitive index; therefore, labelled damage data or explicit modelling of environmental and operational effects is not required. To examine the influence of feature type, autoregressive coefficients, root-mean-square values of autoregressive residuals, principal-component-analysis-based features, and modal frequencies are compared. The proposed framework is first evaluated using the Los Alamos three-storey laboratory frame and is then validated on the full-scale post-tensioned concrete Z24 bridge under long-term monitoring and progressive damage. The results show that the Mahalanobis-distance-based index provides clearer and more stable separation between healthy and damaged states than the Euclidean-distance-based index, owing to its ability to account for the covariance structure of the feature space. The findings also indicate that autoregressive residual features and modal frequencies exhibit the highest sensitivity to changes in structural condition. Overall, the proposed framework provides a simple, interpretable, and computationally efficient strategy for baseline-only vibration-based structural health monitoring under EOV conditions, with particular relevance to full-scale concrete bridges.

Keywords

Main Subjects

Concrete Research

Articles in Press, Accepted Manuscript
Available Online from 16 September 2026

Files

Share

How to cite

Statistics