Deep learning-based brace damage detection for concentrically braced frame structures under seismic loadings

Automated and robust damage detection tool is needed to enhance the resilience of civil infrastructures. In this article, a deep learning-based damage detection procedure using acceleration data is proposed as an automated post-hazard inspection tool for rapid structural condition assessment. The procedure is investigated with a focus on application in concentrically braced frame structure, a commonly used seismic force-resisting structural system with bracing as fuse members. A case study of six-story concentrically braced frame building was selected to numerically validate and demonstrate the proposed method. The deep learning model, a convolutional neural network, was trained and tested using numerically generated dataset from over 2000 sets of nonlinear seismic simulation, and an accuracy of over 90% was observed for bracing buckling damage detection in this case study. The results of the deep learning model were also discussed and extended to define other damage feature indices. This study shows that the proposed procedure is promising for rapid bracing condition inspection in concentrically braced frame structures after earthquakes.