Urban safety network for long-term structural health monitoring of buildings using convolutional neural network

Structural response recovery and prediction technologies are required for long-term structural health monitoring (SHM). In this study, a structural response prediction method was proposed to evaluate the long-term safety of structures in which SHM system failure occurred by constructing a safety network of buildings in a given region. In this method, the correlation is defined by machine learning between the dynamic structural response of nearby buildings that are under the same load effects. Furthermore, a convolutional neural network was introduced in which the structural response correlation between two nearby buildings in a region is defined as a safety network. The safety network can be individually applied to multiple buildings, resulting in multiple networks that can be expanded to construct an urban safety network. To examine the applicability of the proposed technique, a safety network was constructed between two nearby buildings to predict the structural response of the target building.

Automated summary

This study proposed a structural response prediction method to evaluate long-term safety of structures, particularly when SHM system failure occurs. The method defined correlation between dynamic structural responses of nearby buildings under the same load effects using machine learning, and introduced a convolutional neural network to define the structural response correlation as a safety network. The method has the potential to be applied to multiple buildings and construct an urban safety network.