Full-field identification of multi-cable tension using segmented super-sensitivity optical flow with spatially averaged spectral analysis: Wind tunnel tests and field study

Cables are primary load-bearing components but typically most fragile in cable-supported structures; measuring the tension of in-service cables, which is a critical indicator of their health state, is essential to ensure structural safety. As a potential alternative to conventional contact sensors, vision-based techniques, combining with vibration theory, provide low-cost, non-contact, and intrinsically full-field measurement capability for cable tension measurement. However, existing methods suffer insufficient accuracy under long measurement distances and weak surface textures. In addition, they are constrained by point-wise spectral analysis that leads to strong dependence on the selection of measurement location and limited robustness against local disturbances. To address these challenges, this study develops a new vision-based framework for accurate simultaneous estimation of multi-cable tension at full field by integrating super-sensitivity displacement measurement with spatially-averaged spectral analysis. The key is to effectively exploit full-field vibration measurement information in both the time and spectral domains. Specifically, the segmented super-sensitivity optical flow enables accurate pixel-level displacement estimation of each cable at full field beyond the sensitivity limit of photogrammetry; whereas spatial averaging of response power spectral densities yields modal-rich and local-fluctuation-suppressed spectra with reduced dependence on measurement locations. Wind tunnel experiments on a stay cable show that the developed method is able to identify the multi-modal cable displacements across the full field and provides enhanced identifiability of vibration frequencies for tension estimation, achieving an accuracy comparable to coherent laser displacement sensors. Field tests on a cable-stayed bridge of a 580 m main span are conducted for further validations. It is observed that despite low-resolution conditions at long measurement distance, displacements below 0.071 pixels are resolved and tensions of full-field cables are simultaneously estimated by the developed method, in excellent agreement with the installed accelerometer measurements. Finally, the applicability and challenges of the developed method are discussed.