Application of a parameter-shifted grey wolf optimizer for earthquake dynamic rupture inversion

• 作者: Zhenguo Zhang Yuchen Zhang
• 来源期刊: 地震学报（英文版）
• 年，卷（期）: 2021,（6）
• 页码范围: 507-521
• 页数: 15页
• 语种: 英文
• DOI: 10.29382/eqs-2021-0049

摘要

Global optimization is an essential approach to any inversion problem. Recently, the grey wolf optimizer (GWO) has been proposed to optimize the global minimum, which has been quickly used in a variety of inv-ersion problems. In this study, we proposed a parameter-shifted grey wolf optimizer (psGWO) based on the conven-tional GWO algorithm to obtain the global minimum. Com-pared with GWO, the novel psGWO can effectively search targets toward objects without being trapped within the local minimum of the zero value. We confirmed the effectiveness of the new method in searching for uniform and random objectives by using mathematical functions released by the Congress on Evolutionary Computation. The psGWO alg-orithm was validated using up to 10,000 parameters to dem-onstrate its robustness in a large-scale optimization problem. We successfully applied psGWO in two-dimensional (2D) synthetic earthquake dynamic rupture inversion to obtain the frictional coefficients of the fault and critical slip-weakening distance using a homogeneous model. Furthermore, this alg-orithm was applied in inversions with heterogeneous dist-ributions of dynamic rupture parameters. This implementation can be efficiently applied in 3D cases and even in actual earthquake inversion and would deepen the understanding of the physics of natural earthquakes in the future.