叠前弹性波反演非线性优化方法

Nonlinear optimization method on prestack elastic-wave inversion

关键词:  非线性多参数优化  叠前弹性波反演  并行遗传算法

Key words:  nonlinear multiparameter optimization  prestack elastic-wave inversion  parallel genetic algorithm

摘要:

针对传统遗传算法自身存在的早熟收敛、搜索空间小以及计算效率低的问题，在保证算法收敛和最大限度地 搜索模型空间的基础上，对遗传算子采取相应策略进行了改进，并通过界约束以增加解的稳定性。为了提高计算效 率,采用粗粒度并行遗传算法，将并行计算机的高速并行性和遗传算法固有的并行性相结合，进行多种群并行搜索。 选择合适的迁移拓扑结构和迁移策略，构建了并行模型，并给出了改进后并行遗传算法的设计流程图及详细算法描 述。采用该算法进行了叠前弹性波反演的实际计算，取得了良好的效果。

Abstract:

Due to the problems of premature convergence, searching space and computational efficiency in routine genetic algorithm, some relevant improved strategies were adopted for genetic operators in genetic algorithm ensuring convergence and the effective search of the model space. Additionally，search boundaries were set up to stabilize the solutions. In order to enhance computational efficiency, coarse-grained parallel genetic algorithm (PGA) which combines high-speed concurrency of parallel computer with inherent one of genetic algorithm was adopted to perform parallel search，select appropriate migrating topological architecture and migrating strategy，and establish parallel model. The designed flow chart and the detailed algorithm description of modified PGA were given. The calculated results show that the proposed algorithm is effective.