【作者】 魏巍；

【导师】 刘学伟；

【作者基本信息】 中国地质大学（北京） ， 地球探测与信息技术， 2009， 博士

【摘要】 地震资料去噪和压制多次波是地震数据处理的重要内容。在矿区、厂区等特殊环境进行地震数据采集时,经常会受到周围机械设备、高压电干扰等强噪声影响,致使地震记录噪声干扰严重,资料品质低下。由于机器噪声本身固有的特点,常规去噪方法在此已无法适用,需要寻求新的处理技术。另外,在压制多次波方面,基于波动方程的方法现已成为压制多次波的主流技术,其中多次波自适应相减是关键步骤。而盲信号处理技术作为目前信号处理中最热门的学科之一,具有可靠的理论基础和许多方面的应用潜力。其中自适应噪声抵消和独立分量分析技术作为盲信号处理的主要技术,目前在语音处理、生物医学等领域得到广泛的应用。本文重点研究了盲信号处理技术在消除机器噪声和多次波自适应相减两个方面的应用,其主要研究内容:(1)对地震资料的常规去噪方法进行分析总结,并对地震信号与机器噪声特征进行全面的分析研究。(2)研究学习盲信号处理技术的理论基础,掌握其基本原理;对各种算法的具体步骤进行详细研究并且重新推导公式;编制用于地震数据去噪的算法程序。(3)对常规多次波自适应相减方法进行研究总结,得出要求多次波与一次波正交是导致算法低效的主要原因。(4)重点研究基于独立分量分析的自适应相减算法,通过程序实现对数据进行处理,并与常规方法的结果进行比较分析。取得的主要创新点及成果:(1)提出利用自适应噪声抵消技术消除机器噪声的方法,并针对常规技术存在的不足,提出了一种改进的算法。(2)首次引入基于频域和时域两种独立分量分析方法用于地震数据去噪,针对频域算法分离结果存在振幅、排序模糊性问题,提出切实可行的解决方案,综合验证表明方法的有效性和优势。(3)提出了一种多次波自适应相减的优化方案,即通过拟多道匹配滤波技术改善波形匹配问题,再由独立分量分析方法来分离一次波和多次波以避免正交性问题。理论分析及资料处理表明该技术可以大大改善消除多次波的效果,并能很好保持一次波的有效能量。更多还原

【Abstract】 Seismic data noise removal and multiple attenuation are the important part of seismic data process. The noise generated by the working machines and the power interference around the detectors collecting the seismic signal in some special fields such as factories usually contaminated the seismic data. The conventional methods proved inadequate in removing the machine noise with special characteristic from the seismic signal. Some new method should be studied to remove it. In addition, the method based on wave-equation has been the important technique on multiple attenuation. The adaptive multiple subtraction algorithm plays an important role in wave-equation method. Moreover Blind Signal Processing as the most popular subject has reliable theory and abroad application potential. Adaptive noise cancellation technique and Independent Component Analysis (ICA) which are the main methods of Blind Signal Processing, have been widely applied in some fields such as voice processing, biomedicine. This paper investigates the application of Blind Signal Processing in machine noise removal and multiple attenuation. The main contents are shown as follows:(1)Summarized the conventional noise removal methods and analyzed the characteristic of the signal and machine noise in the round.(2)Studied the theory of Blind Signal processing; Masterd the fundamentals;Studied some algorithms and newly deduced the formulas;Achieved the program based on Blind Signal Processing for reducing the machine noise.(3)Investigated and summarized the conventional adaptive multiple subtraction methods; Research showed that the primary vector is not vertical with the multiple vector make conventional methods can not get better result.(4)Studied the adaptive multiple subtraction method based on the Independent Component Analysis; Achieved the program of this method and processed the data; Compared the results with conventional methods’outcome.The main innovative points and achievements:(1) Presented the adaptive noise cancellation technique to reduce the machine noise; Developed a new algorithm aiming at the shortcoming of the conventional technique. (2) Primarily presented two ICA methods operated separately in frequency domain and in time domain for reducing seismic data noise; Solved the illegibility problem in amplitude and order of the frequency domain method; The results of tests and real data processing showed the effectiveness and the superiority.(3)Presented a optimized project for adaptive multiple subtraction; Used pseudomultichannel matching filter to revise the wavelet difference between the predicted multiple and the real multiple in the record before using the ICA adaptive multiple subtraction method which can get better result when primary and multiple vector have overlap; Theory and data processing proved that this method can eliminate the multiples effectively without hurting the primary.更多还原