Damage location of Runyang cable-stayed bridge based on BP neural network

The damage location of long span bridge remains a challenge. This paper aims to develop a damage location method based on BP neural network to diagnose the cable damage of a long span cable-stayed bridge (Runyang North Bridge). First the damage patterns are defined based on plentiful dynamical calculation. The careful analysis of damage pattern reveals that the damage patterns caused by different damage location appear inherent distinctness, while the damage extent only linearly amplifies the damage pattern curves. And the fourth, sixth and seventh frequencies are canceled form the patterns because of the insensitiveness to cable damage. Then a Back-Propagation neural network is designed by trail and error to describe the 7 dimensions mapping space of damage pattern. Identification results prove that the properly organized Back-Propagation network could effectively grasp the damage pattern and identify the damage location correctly.