Influence of Dominant Response Modes on Structural Seismic Demand Modeling

A seismic demand model attempts to describe the behavior of a structure in terms of a set of predictor variables that represents the loading. For buildings, the most frequently used demand parameter and predictor variables are the maximum interstory drift ratio (MIDR) and the spectral accelerations of the ground motion at various modal periods, respectively. An adequate and optimal demand model should be independent of the suite of records that is used to calibrate it. It is shown that this is not the case with currently used demand models and that the dominant dynamic modes imposed by the ground motion suite have a significant effect on the model predictions. In this study, this influence is quantified in terms of the coefficient of partial determination. It is shown that the marginal contribution of the included variables in the demand model is dependent on the response mode that yields the MIDR. An alternative method of estimating the regression coefficients via Ridge estimation is discussed as an approach that minimizes the influence of the dominant mode on the demand model. The performance of the Ridge estimation is compared with the least squares (unbiased) counterpart using the cross-validation method. These findings have a major impact on the selection of ground motions for seismic assessment of structures. (C) 2015 American Society of Civil Engineers.