عنوان مقاله [English]
Due to the lack of earthquake records and limitations on one hand, and the increasing demand on time-history dynamic analysis of structures on the other hand, generation of artificial earthquake accelerograms becomes a more and more important research topic. In such a situation, using artificial accelerograms seems to be the most logical and practical approach. These accelerograms should be constructed in such a manner that includes the seismic characteristics of the region in which a structural design is performed. Because of the random nature of earthquakes, one should generate many records, on which an acceptable analysis can be based. The best accelerogram is one that has compatible characteristics with the desired area. Therefore, it is difficult or may be impossible, in some cases, to choose a proper record for a design area, because the recorded and processed accelerograms of the design location are few. Besides, other location records do not satisfy the geo-seismic characteristics of the desired location. Wavelets are capable of decomposing a time series to different levels, such that each level covers a definite frequency domain. The purpose of the classic wavelet is to provide an alternate way of breaking a signal down into its constituent parts. Low and high frequencies have narrow and wide bands, respectively. Therefore, a signal could be separated into two sub-signals of approximation and detail by using low and high pass filters. In this paper, a combination of artificial neural networks and wavelets are employed for the formation of artificial earthquake accelerograms with a response spectrum similar to that of the target spectrum. Since wavelets permit the decomposition of each signal, it enables use of the parameters to generate many records, where all these follow a single characteristic of the target spectrum. This paper presents a simple and applicable method to generate many artificial records having the given spectrum. Besides, the method applies real records, which are more reliable than other random methods. The feasibility and reliability of the method have been verified with different examples.