Path planning and topology-aided acoustic emission damage localization in high-strength bolt connections of bridges

As the high-strength bolt loosening, corrosion, fracture, falling off, and plate cracking problems become increasingly severe, real-time monitoring of critical bolt connections in bridges is in great demand. Considering the complex geometry of large-scale bolt connections with multiple holes, a path planning and topology-aided method was presented for damage localization using acoustic emission (AE). The path planning algorithm, artificial potential field-guided rapidly-exploring random tree* (APF-RRT*), was introduced to adaptively estimate the shortest distances from the source to sensors while avoiding bolt holes, which were characterized as the actual propagation paths of AE waves. The topology was constructed to represent the geometric relationship between potential source locations and sensors. The shortest distances corresponded to the edges in the topology that connected the vertices of the structure and sensors. The vertex that minimized the error function of arrival times and distances was the predicted damage location. With the help of path planning and topology, it eliminates the straight-line wave propagation assumption, the time-consuming iterative optimization process, and the need for extensive training data. Through the experiment on a full-scale high-strength bolt connection plate, the proposed method was demonstrated to achieve real-time damage localization with higher accuracy and robustness than the time-of-arrival (TOA) method, delta-T mapping method, Gaussian process (GP)-based method, artificial neural network (ANN)-based method, and the counterpart method without topology. • An AE damage localization method is proposed for large-scale bolt connections. • APF-RRT* estimates actual wave propagation paths by avoiding geometric obstacles. • Topology abstracts the geometric relationship between damage sources and sensors. • It eliminates the straight-line wave path assumption and the need for training data. • It provides more accurate real-time damage localization results than existing methods.