A new approach to estimate a mixed model-based ground motion prediction equation
Abstract
A derivative-free approach based on a hybrid genetic algorithm (HGA) is proposed to estimate a mixed model-based ground motion prediction equation (attenuation relationship) with several variance components. First, a simplex search algorithm (SSA) is used to reduce the search domain to improve the convergence speed. Then, a genetic algorithm (GA) is employed to obtain the regression coefficients and the uncertainties of a predictive equation in a unified framework using one-stage maximum-likelihood estimation. The proposed HGA results in a predictive equation that best fits a given ground motion data set. The proposed HGA is able to handle changes in the functional form of the equation. To demonstrate the solution quality of the proposed HGA, the regression coefficients and the uncertainties of a test function based on a simulated ground motion data set are obtained. Then, the proposed HGA is applied to fit two functional attenuation forms to an actual data set of ground motion. For illustration, the results of the HGA are compared with those used by previous conventional methods. The results indicate that the HGA is an appropriate algorithm to overcome the shortcomings of the previous methods and to provide reliable and stable solutions. © 2007, Earthquake Engineering Research Institute.
Publication Title
Earthquake Spectra
Recommended Citation
Tavakoli, B., & Pezeshk, S. (2007). A new approach to estimate a mixed model-based ground motion prediction equation. Earthquake Spectra, 23 (3), 665-684. https://doi.org/10.1193/1.2755934
