Cable force monitoring and prediction for cable group of long-span cable-supported bridges

Cable force monitoring is an essential and critical part of structural health monitoring for long-span cable-supported bridges. Considering economical and efficient issues, the accuracy and quality of safety assessment depend considerably on a reasonable cable-monitoring scheme, especially the number and locations of limited sensors. This paper presents the optimal sensor placement for cable force monitoring and cable force prediction in non-sensor positions with optimal sensor arrangement. Bond-energy algorithm was used to obtain optimal sensor placement strategy, where mutual information was employed to describe the inherent spatial correlation of cable group. To maximize useful cable force information of cable group utilizing from optimal sensor arrangement, particle-swarm optimization-based Kernel Extreme Learning Machine model was employed for forecasting cable forces in non-sensor positions. The analysis results illustrated that the proposed kernel extreme learning machine can achieve better predictive performance than multiple linear regression model and multiple adaptive spline regression model with higher accuracy and generalization. The cable force monitoring and prediction with notable decrease of sensor number lays a comprehensive basis for bridge safety assessment and verified the effectiveness of the proposed method.