Appraisal of the evolutionary-based methodologies in generation of artificial earthquake time histories



Through the last three decades different seismological and engineering approaches for the generation of artificial earthquakes have been proposed. Selection of an appropriate method for the generation of applicable artificial earthquake accelerograms (AEAs) has been a challenging subject in the time history analysis of the structures in the case of the absence of sufficient recorded accelerograms. In this paper we have spotlighted the application of the evolutionary algorithms in the AEAs generation approaches. In this regard, we have statistically apprised the two novel methods; the genetic algorithm-based and the hybrid evolutionary neural network-based methods. The main feature of this paper is to provide some statistical information of the two proposed methods to make some quantitative criteria for assessing the future models and algorithms. The assessment is performed based on three major functions of the spectrum-compatibility, the stochastic diversity of generated seismographs and the computational efforts. The results demonstrate the practical advantages of the evolutionary algorithms in this context.