Rapid seismic response prediction of rocking blocks using machine learning.

The paper proposes the use of supervised machine learning (ML) methods for quickly predicting the seismic response of rocking systems when subjected to seismic excitations. Different supervised ML algorithms are discussed, while a relatively simple and a more sophisticated algorithm are examined in detail. Specifically, the two algorithms compared are the k-Nearest Neighbor (k-NN) and the Support Vector Machine (SVM). The performance of the ML models is demonstrated considering both sine pulses and different sets of natural ground motion records. The results are practically perfect for sine pulses, while accurate results were also obtained for the case of natural ground motions. The proposed ML-based tool allows to quickly assess the risk of damage for rocking systems, while it is also very important when a large number of rocking blocks have to be studied, e.g. in the case of a building's inventory. [ABSTRACT FROM AUTHOR]