A summary evaluation of output-only damage-sensitive features for structural health monitoring of offshore platforms subjected to ambient loads

Abstract

The purpose of structural health monitoring (SHM) is to detect damage as early as possible. To this end, damage-sensitive features (DSFs) must be evaluated to achieve the objectives. Researchers have developed numerous features. However, some gaps remain in the literature for feature selection; thus, conducting research that collects, evaluates, and suggests suitable DSFs are deemed essential. Consequently, this article seeks to identify effective DSFs and damage diagnosis recommendations. Seven categories are examined for output-only DSFs: 1) statistical features, 2) spectral function features, 3) modal property-based features, 4) entropy-based features, 5) transmissibility function (TF), 6) time series models, and 7) vibration parameter features. The sensitivity of DSFs to noise is also investigated. Three types of offshore platforms subjected to ambient loads are modeled in SAP2000. For each model, nine damage scenarios with three types of damage (cross-section area, moment of inertia, and pinned connection) are considered. The results of this study include a comparison of the performance of categories and features, efficient features for identifying the damage occurrence, and efficient features for locating the damage in noise-free and noisy conditions. Finally, suggestions are made for effective damage diagnosis.