Principal Component Analysis for Predicting the Response of Nonlinear Base-Isolated Buildings

Abstract

In the design of base-isolated buildings, a critical parameter governing the selection of isolator parameters is the peak displacement of the isolation layer. In this study, a model is developed in order to predict the peak base displacements utilizing multiple ground motion parameters, called intensity measures (IMs), as the inputs. The issue of correlation between various IMs is addressed through principal component analysis (PCA). This method also lends itself to dimensionality reduction, as those components that do not contribute significantly to the variance are discarded. The prediction intervals from the model are compared with the results from nonlinear dynamic analysis. An important conclusion is that by using the PCA based model, the standard errors remain relatively small and constant for a wide range of isolation periods. It is therefore clear that by utilizing multiple IMs and accounting for their correlation effects, it is possible to estimate the responses of base-isolated buildings with good confidence.