Structural performance‐based anomaly detection for velocity pulse

Abstract Pulse‐like ground motion can cause extreme damage to long‐period structures. An automatic algorithm is proposed to identify pulse‐like ground motions, in which improved anomaly detection is applied and the structural performance is considered. To characterize the intrinsic pulse‐like features, the distance‐based anomaly detection algorithm is improved, and the relative cumulative energy is added to the Manhattan distance. The first application of the anomaly detection algorithm performs well in pulse‐like motion identification. The pulse‐like ground motions are identified according to the performance of typical long‐period engineering structures in Hazus by the improved capacity spectrum method. The method is applied to 214 strong ground motions in the Next Generation Attenuation West 2 database. Comprehensive comparisons of the proposed approach with the wavelet analysis and the energy method are presented to illustrate the significance of the structural performance on the pulse‐like identification method.