A pattern recognition technique for structural identification using observed vibration signals: Nonlinear case studies

Abstract

A novel pattern recognition technique, support vector regression (SVR), has been introduced for linear structural parameter identification in a companion article. It is recognized that structural systems in general are designed to behave nonlinearly when subjected to extreme loading. Therefore SVR-based methods for nonlinear structural identification (SI) have been studied and they are summarized in this paper. The first method uses the SVR technique to identify nonlinear structural parameters, whereby the power parameter controlling the shape of the Bouc–Wen model is known. The second and third methods conduct nonlinear SI in the power parameter unknown condition, with the difference that the third method adopts a model selection strategy to enhance the nonlinear parameter estimation practicability. Five-story nonlinear structural systems whose restoring forces are expressed by the Bouc–Wen model are investigated to demonstrate the effectiveness of the SVR-based SI methods. Verification results show that the third method using the model selection strategy is the most efficient one for nonlinear structural parameter identification.