基于逐减随机震源采样法的频率域二维黏滞声波方程全波形反演

发布时间：2018-02-27 23:01

  本文关键词： 逐减随机震源采样法 频率域 二维黏滞声波方程 全波形反演 无记忆拟牛顿算法 L-BFGS　出处：《吉林大学学报(地球科学版)》2016年06期 　论文类型：期刊论文

【摘要】：全波形反演方法利用叠前地震波场的运动学和动力学信息重建地下速度结构,具有揭示复杂地质背景下构造与岩性细节信息的潜力。然而,巨大的计算量是阻碍其发展的一个瓶颈问题。为此,研究者们提出了震源编码技术来减少计算量,但是此方法在模型更新过程中会引进随机串扰噪声,降低反演结果准确性。所以,在保证计算精度的情况下,本文提出了采用逐减随机震源采样的方法来高效计算全波形反演问题。笔者将此方法应用于频率域二维黏滞声波波动方程全波形反演,开始了在频率域进行随机震源采样类方法的研究,计算过程中共使用了依次增大的8个频率段;并应用Overthrust模型来验证此类随机震源采样法的正确性。实验结果表明:基于逐减随机震源采样法的反演结果与实际Overthrust模型的拟合误差为0.065 65,而应用基于全部震源的全波形反演方法得到的反演结果与实际Overthrust模型的拟合误差为0.064 64,两者差别不大;但计算用时由740min减少到291.2min,即计算效率提高了2.54倍。为了更好地确定方法的有效性,将其应用于Marmousi模型进行试算。模型试算结果表明:基于逐减随机震源和基于全部震源得到的反演结果与实际Marmousi模型的拟合误差分别为0.080 12和0.078 97,相差不大;但计算用时由1 218.9min减少到274.4min,计算效率提高了4.44倍。综上,在保证反演精度的情况下,基于逐减随机震源采样法的频率域全波形反演方法大大减少了计算量,具有不可替代的计算优势,并且没有引进随机串扰噪声。
[Abstract]:The full waveform inversion method uses the kinematics and dynamics information of pre-stack seismic wave field to reconstruct the underground velocity structure, which has the potential to reveal the details of structure and lithology in complex geological background. However, The huge amount of computation is a bottleneck problem that hinders its development. For this reason, the researchers put forward the source coding technique to reduce the computation load, but this method will introduce random crosstalk noise in the process of model updating. Reduce the accuracy of the inversion results. So, while ensuring the accuracy of the calculation, In this paper, a method of reducing random source sampling is proposed to efficiently calculate the full waveform inversion problem, which is applied to the full waveform inversion of two-dimensional viscous acoustic wave equation in frequency domain. The method of random source sampling is studied in frequency domain. Eight frequency bands are used in the calculation. The Overthrust model is used to verify the correctness of this random source sampling method. The experimental results show that the fitting error between the inversion result based on the subtractive random source sampling method and the actual Overthrust model is 0.065 65, while the application of the method based on the full source. The fitting error between the full waveform inversion method and the actual Overthrust model is 0.064 64, the difference between them is not significant. However, the computational time was reduced from 740min to 291.2 mins, that is, the computational efficiency was increased 2.54 times. The results show that the fitting errors of the inversion results based on the decreasing random source and all the sources are 0.08012 and 0.078 97, respectively, compared with the actual Marmousi model. However, the calculation time is reduced from 1 218.9 min to 274.4 min, and the computational efficiency is increased by 4.44 times. In summary, under the condition of guaranteeing the inversion accuracy, the frequency domain full waveform inversion method based on the subtractive random source sampling method greatly reduces the calculation amount. It has irreplaceable computational advantages and no random crosstalk noise is introduced.
【作者单位】： 吉林大学地球探测科学与技术学院;
【基金】：国家自然科学基金重点项目(41430322) 国家重点基础研究发展计划(“973”计划)项目(2013CB429805)~~
【分类号】：P631.4
，

本文编号：1544691