【作者】 崔树果；

【导师】 刘怀山；

【作者基本信息】 中国海洋大学 ， 海洋地球物理， 2008， 博士

【摘要】 多次波是地震记录中一种常见的相干噪音,尤其是在海洋地震记录中。如何有效地消除或压制各种类型的多次波,并最大限度地保留一次波信号是地震勘探中的一个重要课题。由于多次波的存在,使得地震资料的信噪比降低,干扰人们对有效波的识别,从而导致速度分析、叠前及叠后偏移的极大困难,影响地震成像的真实性和可靠性,并将导致假的成像,严重影响地震解释工作。因此,研究和压制多次波,具有重要的现实意义和发展前景。本学位论文依托导师刘怀山的国家高新技术研究发展计划(863计划)课题“浅海地震资料特殊干扰波形成机理与剔除方法(2006AA09Z339)”。自20世纪50年代以来,已研究实现了多种压制多次波的方法技术,它们分别基于不同标准来区分多次波和一次波。压制多次波的方法主要分为两大类:一类是基于有效波和多次波之间的可分离性和其他差异性(如多次波周期性、时差差异等)来预测与压制多次波。比较典型的方法有预测反褶积、正常时差变换叠加、f - k变换、τ-p变换、抛物线Radon变换和聚束滤波等,它们可以简称为基于几何地震学方法或“几何滤波类方法”;第二类主要是基于弹性波动理论的方法,通过模拟或反演方法来预测原始数据中多次波,继而从原始数据中匹配减去所预测的多次波。这类方法也称为波动预测减去法,主要有:波场外推法、反馈环法、反散射级数法。基于波动方程理论,压制和利用多次波是20世纪90年代发展起来的一种新思路,也是国外多次波研究的发展趋势。这种思路方法,综合考虑自由界面多次波传播的运动学特性和动力学特性,针对多次波产生的机理,有效避免了各种非波动方程方法解决问题所带来的局限性,使多次波研究发展步入了新的里程,标志着多次波理论与应研究进入了一个新的层次和水平。本人在上述背景下,开展博士学位论文工作。与其他方法相比,自由表面多次波压制(Surface-related Multiple Elimination,简称SRME)预测过程无需知道宏观速度场,从而增强了预测方法的适应性。这是SRME的最大优点。由于SRME推导过程中,做了四个假设:①震源子波不发生改变;②自由界面的反射系数为-1;③在数据采集过程中,检波器保持稳定;④地层是层状介质。如果假设条件不满足,多次波压制效果就会受到影响。另外,SRME是一种Kirchhoff频率域积分法,所以易受野外采集观测系统影响。为了克服原始数据驱动预测多次波过程受数据中子波和多次波的影响,本文采用反射率法实现了脉冲响应炮集记录。与其他正演方法(射线追踪、有限差分)相比,反射率法是f-k域实现的,计算速度较快。用反射率法可以在任何检波器位置模拟脉冲响应单炮,为了后续的SRME Kirchhoff积分提供了数据保障,同时在一定程度上克服了子波的影响。与原始数据相比,模拟出来的记录不含子波的影响,同时弥补了原始数据空间采样不足的问题。利用模拟出来的叠前脉冲响应炮集与原始数据相结合预测多次波模型,可以在一定程度上克服空间采样率不足和对多次波预测过头的问题。本文模拟出的叠前脉冲响应炮集,与实际地层介质难免存在差异,因此预测出的多次波也会受到残留子波的影响。在迭代去除多次波的过程中,需要进一步消除残留子波的影响。但是在实际操作过程中如何求准子波始终是数据匹配过程的一个难题。为了匹配准确,作者提出了Huber混合模构造目标函数,然后利用伪牛顿梯度算法求解匹配算子。与其他模相比,Huber混合模是一种有效的误差估计准则,它综合了L2模的优点(对小误差的平滑作用)和L1模的优点(对大的野值不敏感)。另外Huber模目标函数是可微的,所以可以利用梯度类优化方法最小化。多次波模型与实际多次波的匹配一个关键参数就是时窗位置的选取,如果选取不当的话,就会导致多次波向有效波靠拢,损害有效。根据水平地层多次波的周期性特点,本文利用多次波时距曲线约束时窗位置(在多数情况下,水平地层假设是可以适用的),提高了匹配滤波的准确度。利用时窗约束Huber模估算匹配算子克服了多次波与有效信号匹配不准的问题。模拟数据和东海实际资料处理均表明,本文所提出的方法能够有效地压制多次波。更多还原

【Abstract】 In seismic prospecting especially on the sea, it is the multiple wave that causes the SNR of seismic data to reduce greatly, disturbs the identifiability of primary reflection, brings enormous difficulty on velocity analysis and excursion before and after stacking, and then affects the seismic imaging authenticity and reliability, sometimes may be false. Multiple may seriously influence seismic interpretation work also. So it is very important and necessary for multiple suppression. The research of this dissertation is funded by the Hi-tech research and development program of China (Grant No. 2006AA09Z339).Since the 20th century 50’s, a lot of multiple suppressing methods have come out, which make use of different characteristics of multiple. methods that suppress multiples can be classified into two broad categories: One is based on the moveout between primary and multiples. Typical methods of the first class consist of predictive deconvolution, normal moveout stacking, f-k transformation, Tau-p transformation, parabola Radon transformation and beamforming. They can be called as seismography method based on the geometry. The second one is mainly based on wave equation, which predict, match it with the original data and then subtract multiples from the input data. The wave equation methods include Wavefield Extrapolated method, Iterative Feedback method and Inverse Scattering Series method. In this dissertation, a comprehensive study has been done on the wave equation-based methods.From the perspective of the acoustic wave equation theory, using the Kirchhoff Integral solution, we obtained the wave field extrapolation formula in the frequency-space domain. Regardless of surface reflection, the primary propagates downward, reflects at the subsurface reflector, and then propagates back. Taking the interface of seawater-to-air into account, the so-called surface-related multiple has been modeled. The surface-related multiple elimination can be treated as an inverse procedure of surface-related multiple modeling. Algorithms, formulas, program flowcharts for the surface-related multiple have been given in detail in this dissertation.A surface-related multiple-elimination method can be formulated as an iterative procedure: the output of one iteration step is used as input for the next iteration step. In this paper it is shown that the procedure can be made very efficient if a good initial estimate of the multiple-free data set can be provided in the first iteration, and in many situations, the synthetic method may provide such an estimate. It is also shown that for each iteration, the inverse source wavelet can be accurately estimated by a nonlinear (Huber norm) inversion process. The iterative multiple elimination process, together with the source wavelet estimation, are illustrated with numerical experiments as well as with field data examples. The results show that the surface related multiple-elimination process is very effective in time gates where the moveout properties of primaries and multiples are very similar.From the physical insight, surface multiple attenuation is to combining primary events to predict a multiple is similar to the diffraction-aperture problem of classical optics. Although the algorithm requires no assumptions or modeling regarding the positions and reflection coefficients of the multiple-causing reflectors, it does require complete internal physical consistency from the optical veiwpoint. SRMA also has several weaknesses. The wavelet in the data contaminates the multiple prediction operator and has to be removed in order to predict multiples accurately. Since the prediction operator is the data, it is only sampled for locations where shots were initiated, not in general the same as the locations were traces were recorded. The SRMA operator tends to overpredict higherorder multiples because it should be convolving with an operator containing only subsurface primaries and interbeds. As a result it must resort to some type of iterative scheme to drive this contamination down. To sidestep acquisition limitations, source wavelet contamination and overprediction of higher order multiples, we have developed a method which essentially models surface-related multiples by adding a synthetic shot to the data. More specifically, each input trace is convolved with a synthetically-generated shot record consisting of one or more key subsurface primary reflections. These records are impulse response of the beneath, which are generated by the so-called reflectivity method. The resulting operators have several appealing features. They are clean operators, having no source wavelet contamination and little or no aliasing. By construction they are available for any source-receiver combination desired. By modeling only primaries into the operator, we avoid the geometric overprediction of higherorder multiples that plagues SRMA. Because l 1norm adaptive filter is incorporated, we retain much of the ability of SRMA to honor lateral amplitude variations induced by these reflectors in the predicted multiples.The synthetic and real examples clearly justify the validity of the technology presented in this dissertation.更多还原