Selection and scaling of accelerograms for time history analysis using inelastic intensity measures

With advances in computational facilities and progress in structural dynamics, nonlinear time history analysis is accepted as an efficient method for seismic design. Because of time-consuming analyses, and limited number of available records, a compromise in the number of records is essential for a practical analysis. The common approach in building codes and in FEMA 356 is to select and scale few records to estimate the response that would be obtained under the ideal scenario where no compromise was required.;Using spectral shape indices, 40 accelerograms recorded on Site Class C of the International Building Code 2006 are ranked with respect to the closeness of their elastic pseudo spectral acceleration spectrum to the design spectrum of FEMA 356 for a coastal Californian region. Genetic Algorithms are used to find an optimal set of scale factors to minimize the area between the average scaled spectrum for seven selected records and the target spectrum. The method did not provide dependable results when the records do not possess spectral shapes close to the target spectrum.;Seismic codes recommend modifying records such that the average elastic response spectrum matches an elastic design spectrum scaled by a response modification factor. However, because inelastic response is not proportional to the excitation amplitude, the procedure does not guarantee close agreement between the computed and the expected inelastic response. Knowing that most structures respond inelastically to major earthquakes, inelastic spectral matching is pursued in this study using an iterative scheme based on wavelet transformation and the damage index proposed by Park and Ang. Four records with different spectral shapes are modified to match a smooth and a jagged damage spectrum for a ductility factor of 4. The modified spectra show close agreement with the target spectra. The method maintains the characteristics of the frequency content of the original records. It supersedes the traditional methods developed for elastic spectral matching and can also be utilized in damage-based design methodologies. Results also indicate that a wavelet transformation has a better performance than a Genetic Algorithm for elastic spectral matching. Since both wavelets and Genetic Algorithms estimate the design spectrum over the entire range of periods, they can capture the response of higher modes in a multi-degree-of-freedom system and the response at increased period of an inelastic system.;By adding a third axis that represents ductility to a response spectrum, the concept of "spectral surface", referred to as a "damage surface" if a damage index is used as the response parameter, is introduced. A response surface gives a better insight on the variations of the response with both period and ductility. The feasibility of modifying a record to produce the desired response for any pair of period and ductility is investigated by extending the method used for inelastic spectral matching. The efficiency of the method depends on the record and the target spectrum. Using surface spectral matching, the structure can be designed for more than one performance level using one modified record.