基于粘弹性AVO理论的海底物性参数正则化反演方法研究

Seafloor properties play an important role in both civilian and military aspects. AVO (Amplitude Variation with Offset) technology is an effective method to extract physical parameters of the underground media from pre-stack seismic data. However, conventional researches on AVO technology are mainly focused on the strata under the seafloor sediments. The AVO response of the interface between seawater and seafloor sediments is always ignored. The goal of this study is to estimate seafloor sediments properties with the AVO response of the interface between seawater and seafloor sediments. Most of the seafloor sediments are loose and unconsolidated, which have a strong attenuation affect on seismic waves. Previous elastic based researches are not applicable for that seafloor situation. This project will conduct research based on viscoelastic model which takes seismic wave attenuation into account. Additionally, for the previous research, the inversion is performed from one CRP to another successively. The previous inversion strategy makes the inversion result easy to be affected by noise and lack of transverse continuity. In order to overcome the defects of the previous inversion strategy, we will conduct research on simultaneous multi-CRP inversion method based on regularization theory. Finally, this project will form a set of regularization inversion method based on viscoelastic AVO theory for the physical parameters of the seafloor sediments.

海底沉积物物性参数不论在民用还是军用领域都有非常重要的作用。AVO技术作为一种从叠前地震数据中提取地下介质物性参数的有效手段，主要关注了海底以下地层物性参数的反演，对于与海水交界的海底沉积物物性参数的反演往往被忽略。本项目研究目的在于利用地震波海底反射的AVO效应实现海底沉积物物性参数的反演。绝大部分海底沉积物都是疏松的、未固结的，会对地震波产生较强的吸收作用。前期弹性假设下的反演方法不能满足这样的海底情况，本项目将考虑海底沉积物对地震波的吸收作用，在粘弹性模型下开展反演方法研究。同时，前期依次对单个共反射点（CRP）道集进行反演的策略使得反演结果易于受噪声影响、缺乏横向连续性。为克服这一缺陷，同时降低反演多解性，本项目将通过正则化方法实现多CRP道集同时相互约束反演，在反演过程中考虑相邻CRP对当前位置反演结果的影响。最终将形成一套基于粘弹性AVO理论的海底沉积物物性参数正则化反演方法。

海底沉积物物性参数的获取不论在民用还是军用方面都有着重要的意义。本项目利用地震波在海底反射的AVO效应实现海底沉积物物性参数的反演，研究内容围绕海底反射中的正演问题和反演问题两方面展开，各项研究内容均已按照研究计划顺利完成。推导了粘弹性海底沉积物反射系数随入射角变化关系的理论表达式，分析了纵横波品质因子对不同入射角反射系数的影响。针对前期依次对单个CRP道集进行反演的策略存在稳定性差的问题，发展了基于正则化理论的多CRP同时相互约束反演方法。针对上述正则化反演中正则化参数设置过大导致的过正则化问题，发展了一种混合驱动的正则化反演方法。为提高计算效率，针对海底叠前地震反演的特点，设计了一个6层的神经网络构架，并发展了相应的网络训练方法，实现了基于机器学习的海底物性参数反演。对相关算法进行优化整合形成海底沉积物物性参数反演软件一套。项目研究期间，发表SCI论文3篇，申请发明专利4项（目前3项已授权、1项实审中），登记软件著作权1项。