The Effects of Discretization Errors on the High Frequency Content of In-Structure Response Spectra

Abstract

The next generation ground motion prediction equations predict significant high frequency seismic input for rock sites in the Central Eastern United States (CEUS). This high frequency motion is transmitted to basemat supported components and may be transmitted to components supported on elevated slabs. The existing ASCE 4 analysis requirements were initially developed based on seismic motions having lower frequencies, typical of ground motions in the Western United States (WUS). The adequacy of the existing ASCE 4 analysis requirements are examined using high frequency CEUS spectral shapes and the potential error inherent in using the existing approach to computing in structure response spectra is quantified. Modifications to reduce potential error in the existing ASCE 4 criteria are proposed. In structure response spectra are typically generated for a subsystem given the time history response of a building region. The building time history response is based on analyses that use either modal time history superposition, direct integration or complex frequency response analysis of the building and supporting soil. Input to the building analyses consist of either real or synthetic discretized ground motion records. The discretized ground motion records are often based on recorded ground motion seeds and are often limited to a 0.005 second time step. Thus the time step of the seed record often limits the frequency content of the problem. Both the building analyses and in structure response spectra subsystem analysis may interpolate the discretized ground motion records to obtain stable results. This interpolation generates errors that are propagated through the analyses used to calculate in structure response spectra. These errors may result in extraneous high frequency content in the in structure response spectra. Errors are quantified by comparison of time history parameters, Fourier components and in structure response spectra.