EFEFCTS OF GROUND MOTION UNCERTAINTY ON PREDICTING THE RESPONSE OF AN EXISTING RC FRAME STRUCTURE

摘要

Estimation of structural response may be significantly affected by the representation of seismic ground motion uncertainty. A complete probabilistic presentation of ground motion can be constructed by specifying a stochastic model that depends on seismic source parameters. Alternatively, the ground motion uncertainty can be represented by adopting parameters known as the intensity measures (IM), and using attenuation relationships to relate the IM to seismic source parameters. The uncertainty in the prediction of structural response can be expressed in terms of the probability of exceeding a given value of the structural response. In this study, the uncertainty in the ground motion is represented in these two alternative ways: (a) a full probabilistic representation using an advanced simulation technique known as subset simulation (Au [4]) based on a stochastic ground motion model conditional on magnitude and distance proposed by Atkinson and Silva (Atkinson [3]), and, (b) by adopting spectral acceleration at the small amplitude fundamental period as the intensity measure. In alternative (b), a suite of ground motion recordings are used to represent ground motion characteristics not already captured by the IM. The attenuation relation relating the IM to the seismic source parameters is obtained by two alternative approaches: (I) by simulating stochastic ground motions and applying them to an elastic SDOF system and (ii) by using the empirical regression equation of Abrahamson and Silva (1997). The alternative approaches are compared based on their prediction of the uncertainty in structural response. Another comparison is done between the following two cases: (1) by predicting the structural response following alternative (b) and using a suite of real ground motion recordings and (2) by predicting the structural response following alternative (b) and using a suite of synthetic records. The suite of synthetic records are generated according to the same stochastic model used in alternative (a) for a given magnitude and distance; this provides a common basis for comparisons. In order to emulate selection of a suite of real records from a bin, alternative (a) and case (2) of alternative (b) described above are repeated using a suite of synthetic records generated for magnitude and distance as uncertain variables belonging to a designated bin. An existing 7-story reinforced concrete structure is used as a case-study. The structural model (Jalayer [9]) includes stiffness and strength degradation.