Structural Health Monitoring of A Cable-Stayed Bridge

A small set of dynamic features of a cable-stayed bridge in a short-term monitoring process is provided to utilize in research activities regarding structural health monitoring (SHM). This dataset contains 216 modal/natural frequencies of nine days of measurements. The modal frequencies (dynamic features) have been identified by an automated frequency domain decomposition (FDD) technique in four stable modes. The first 192 samples of the modal frequencies are related to the undamaged condition of the bridge (i.e., the first eight days of measurements) and the remaining 24 samples of the modal frequencies belong to the damaged state.

This dataset is suitable for applying to SHM problems based on data-driven methods under the concept of machine learning/unsupervised learning/anomaly detection. The major challenges in this dataset, which can be incorporated into further research, can be categorized as technical and engineering issues. The former is concerned with the problem of generalization of machine learners and their overall performances under small data. The latter pertains to the problem of environmental and/or operational variability and their effects on the modal frequencies.

If you use this dataset, please cite the following article:

Daneshvar, M. H., and Sarmadi, H. (2022). "Unsupervised learning-based damage assessment of full-scale civil structures under long-term and short-term monitoring." Engineering Structures, 256, 114059. https://doi.org/10.1016/j.engstruct.2022.114059