Vision-based vibration analysis and causal discovery for structural damage localisation
NONDESTRUCTIVE TESTING AND EVALUATION, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1080/10589759.2026.2633573
- Dergi Adı: NONDESTRUCTIVE TESTING AND EVALUATION
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Applied Science & Technology Source, Compendex, INSPEC, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO), Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
- Ankara Üniversitesi Adresli: Evet
Özet
This study presents a vision-based non-destructive testing (NDT) framework for structural damage localisation by integrating optical flow - based vibration measurement with the Temporal Causal Discovery Framework (TCDF). A laboratory-scale steel frame with adjustable bolted connections is used as the experimental model, enabling flexible damage scenarios without physical intervention. As a non-contact and non-destructive sensing technique, multiple high-resolution cameras capture full-field displacement responses from QR-code targets attached to the structure. Optical flow is employed to extract precise displacement time series, which are subsequently analysed using TCDF to reveal causal relationships among structural components. The proposed approach enables accurate damage localisation by identifying damage-sensitive causal nodes and provides insight into damage propagation mechanisms within the structure. By coupling vision-based NDT measurements with data-driven causal discovery, the methodology strengthens the reliability and interpretability of Structural Health Monitoring (SHM) systems and offers proof-of-concept for automated, sensor-free damage diagnosis. The novelty of this work lies not in modifying the TCDF algorithm itself, but in its first systematic integration into an NDT-based SHM framework, where TCDF is combined with optical flow - based vibration measurements and optimised through hyperparameter tuning using Optuna.