Author(s)

Hao Wang, Kun Ma

Corresponding Author

Kun Ma

Abstract

As most structural health problems of buildings are cumulative damages, they are difficult to be de-tected in real time. The complexity of actual structure and environmental noise makes structural health monitoring more difficult. The existing methods require a large amount of data when training models, but the marking of data is very complicated in practice. In order to overcome this problem, a wireless sensor network is equipped, and curvature mode and strain mode are adopted to realize bridge structure health monitoring. Then curvature mode and strain mode algorithm training are car-ried out on the basis of feature extraction through a large number of unlabeled instances, so as to re-alize data dimension compression and unlabeled data preprocessing. Secondly, the grid environment Morphin search tree algorithm is used to realize the prediction of bridge structure health monitoring categories. Meanwhile, Hessian optimization is improved based on linear conjugate gradient. The semi-positive definite Gaussian-Newton curvature matrix is used to replace the uncertain Hessian matrix, and secondary target combination is carried out, so as to improve the efficiency of grid envi-ronment Morphin search tree algorithm. The experimental results show that the proposed grid envi-ronment Morphin search tree bridge structure safety detection algorithm realizes high-precision structural health monitoring at the curvature mode and strain mode levels of environmental noise.

Keywords

Curvature mode, Strain mode, Beam bridge structure, Damage identification