基于自适应步长OBFGS算法的快速时间域全波形反演  

Fast time domain full waveform inversion based on self-adaptive step OBFGS algorithm

作　　者：张天泽[1] 韩立国 张盼[1] 胡勇[1] 毛博 ZHANG Tainze;HAN Liguo;ZHANG Pan;HU Yong;Mao Bo(Jilin University,Changchun 130012,China)

机构地区：[1]吉林大学地球探测科学与技术学院,长春130012

出　　处：《物探化探计算技术》2018年第1期1-7,共7页Computing Techniques For Geophysical and Geochemical Exploration

基　　金：国家自然科学基金(41674124);国家重点研发计划(2016YFC0600301)

摘　　要：将OBFGS(Online-BFGS)优化算法应用到全波形反演中,OBFGS算法是一种不依赖Wolfe准则求取步长的优化算法。基于OBFGS算法的特点,提出了自适应步长策略,可以根据当前目标函数的变化情况自适应调整步长。基于自适应步长的OBFGS算法不要求在调整步长时计算正演和梯度,因此可以加快计算速率。通过对Marmousi型进行反演的精度分析表明,基于自适应步长OBFGS算法全波形反演提高了反演精度。反演效率分析结果表明,基于自适应步长的OBFGS算法的反演耗时较短,具有更快的收敛速度,提高了反演的效率。FGS algorithm is one of the most popular optimization method,but it requires the Wolfe principle to achieve convergence and that requires a lot of computing time.In this paper we first introduce the OBFGS(Online-BFGS),which is a method does not require line search,into FWI and we also introduce a self-adaptive step computing method into the OBFGS algorithm,which can make the inversion faster because it does not need to calculate gradient and forward modeling while the step is changing.We use Marmousi model to test the inversion result based on self-adaptive step OBFGS FWI.We prove that the full waveform inversion based on self-adaptive step OBFGS can improve the accuracy and speed of full waveform inversion.

关 键 词：时间域全波形反演 步长搜索 OBFGS算法 自适应步长 

分 类 号：P631.4[天文地球—地质矿产勘探]