Automated wireless monitoring system for cable tension forces using deep learning

As demand for long-span bridges is increasing worldwide, effective maintenance has become a critical issue to maintain their structural integrity and prolong their lifetime. Given that a stay-cable is the principal load-carrying component in cable-stayed bridges, monitoring tension forces in stay-cables provides critical data regarding the structural condition of bridges. Indeed, various methodologies have been proposed to measure cable tension forces, including the magneto-elastic effect-based sensor technique, direct measurement using load cells, and indirect tension estimation based on cable vibration. In particular, vibration-based tension estimation has been widely applied to systems for tension monitoring and is known as a cost-effective approach. However, full automation under different cable tension forces has not been reported in the literature thus far. This study proposes an automated cable tension monitoring system using deep learning and wireless smart sensors that enables tension forces to be estimated. A fully automated peak-picking algorithm tailored to cable vibration is developed using a region-based convolution neural network to apply the vibration-based tension estimation method to automated cable tension monitoring. The developed system features embedded processing on wireless smart sensors, which includes data acquisition, power spectral density calculation, peak-picking, post-processing for peak-selection, and tension estimation. A series of laboratory and field tests are conducted on a cable to validate the performance of the proposed automated monitoring system.