摘要
本文将优化领域应用较广的全局随机非线性粒子群算法与局部迭代梯度法相结合,构造了一种粒子群-梯度算法,并将其应用于频率域波形速度结构反演.数值实验结果表明,粒子群-梯度算法能继承梯度法快速收敛和粒子群法全局寻优的特点,适用于频率域波形反演问题,算法具有一定的抗噪能力,无论在计算精度还是在降低解的非唯一性方面,都有较明显的改善.
A PSO-gradient algorithm combined with the widely used global non-linear stochastic particle swarm optimization （PSO） algorithm and local iteration gradient method is proposed. It is applied to waveform velocity structure inversion in frequency domairu The results show that the PSO-gradient algorithm can inherit the rapid convergence characteristics of gradient method; it also has the same ability of searching global optimization as PSO method. As a suitable tool tackling waveform inversion problems, the PSO-gradient algorithm has the ability of anti- noise in a certain extent and significant improvements both in calculation accuracy and reducing the nonuniqueness of solution.
出处
《地球物理学进展》
CSCD
北大核心
2013年第1期180-189,共10页
Progress in Geophysics
基金
国家自然科学基金项目(41204041、40874024)
国家重点基础研究发展计划“973”计划(2007CB209603)联合资助
关键词
粒子群-梯度算法
频率域
波形反演
全局寻优
PSO-gradient method, frequency domain, waveform inversion, global optimization
