Nonlinear multiclass support vector machine–based health monitoring system for buildings employing magnetorheological dampers

Abstract

In this article, a nonlinear multiclass support vector machine–based structural health monitoring system for smart structures is proposed. It is developed through the integration of a nonlinear multiclass support vector machine, discrete wavelet transforms, autoregressive models, and damage-sensitive features. The discrete wavelet transform is first applied to signals obtained from both healthy and damaged smart structures under random excitations, and it generates wavelet-filtered signal. It not only compresses lengthy data but also filters noise from the original data. Based on the wavelet-filtered signals, several wavelet-based autoregressive models are then constructed. Next, damage-sensitive features are extracted from the wavelet-based autoregressive coefficients and then the nonlinear multiclass support vector machine is trained by a variety of damage levels of wavelet-based autoregressive coefficient sets in an optimal method. The trained nonlinear multiclass support vector machine takes new test wavelet-based autoregressive coefficients that are not used in the training process and quantitatively estimates the damage levels. To demonstrate the effectiveness of the proposed nonlinear multiclass support vector machine, a three-story smart building equipped with a magnetorheological damper is studied. As a baseline, naive Bayes classifier–based structural health monitoring system is presented. It is shown from the simulation that the proposed nonlinear multiclass support vector machine–based approach is efficient and precise in quantitatively estimating damage statuses of the smart structures.