عنوان مقاله [English]
Time-history earthquake records are critical means for many earthquake engineering applications, including step-by-step numerical solution of dynamic, nonlinear or highly damped systems, seismic design of new buildings and
infrastructures, seismic control, risk assessment and vulnerability evaluation of existing structures, and so on. They are provided using statistical simulated records, geological model-based artificial earthquakes or recorded
accelerograms of real experienced earthquakes. The latter is preferred in many cases, due to expensive computational effort and lack of reliability in simulating events of strong ground motion using artificial methods. However, the resulted structural responses are highly sensitive to the selected earthquake record, so that their variation from one record to another is a major source of uncertainty in further design and decision making. In order to overcome such a challenge, well-known seismic design regulations have been accepted using a set of earthquake records instead of one, provided they are scaled to match a target design spectrum. It is especially beneficial in the absence of sufficient site-specific earthquake records or the impossibility of their extraction via seismic hazard analyses. As no further limitation than over-riding the design spectra is implied, the resulted mean spectrum may over-estimate the target, leading to an economically undesired design. In addition, the corresponding scale factors will not be unique for a set of records unless taken as similar or optimized by proper methods. Hence, it is reasonable to select a set of accelerograms from an available catalogue that is optimally compatible with the design spectrum; formulated as the first optimization problem in the present work. While the set of optimal records is identified, their scaling factors are also optimized as the second problem formulation. The present work employs genetic algorithm as a reliable meta-heuristic, and utilizes its operators to solve both optimization problems. In addition to conventional binary coding, an integer index coding is also utilized for the special case of optimization problems in this research. The superiority of the integer coding over the binary type is not only discussed in reducing search space cardinality, but is also shown via the results of a numerical example, using a list of 136 earthquake records, to match the target spectra, according to the Iranian Standard of Practice; No.2800. The results also show the proposed method is highly beneficial in matching the target
design spectrum, with respect to common practice, with a trial-selected, uniform-coefficient set of records.