Empirical-based support vector machine method for seismic assessment and simulation of reinforced concrete columns using historical cyclic tests

Abstract

Support vector machine models are generally considered as “black box” procedures, which sometimes render them unexplainable and prone to over-fitting or under-fitting in absence of sufficient data. In this study, an empirical-based support vector machine (EM-based SVM) method is proposed to model the complicated relationships among the basic characteristics of reinforced concrete columns and their corresponding critical performance metrics on hysteretic curves. Moreover, a database of historical pseudo-static cyclic tests is established for model training and test after rigorous screening of specimens. The developed model is then validated by comparing with the results of common SVM method and semi-empirical formulas, as well as pseudo-static and shake table tests. Finally, its applicability and effectiveness for seismic damage assessment and simulation are explored, which demonstrates its potential value in engineering applications.