【作者】 潘纪顺；

【导师】 张先康；

【作者基本信息】 中国地震局地球物理研究所 ， 固体地球物理学， 2007， 博士

【摘要】 本文首先回顾了地震层析成像和地球物理反演的发展历史，分析了全局优化算法和局部线性化反演方法各自的优缺点，指出将两类方法相结合可能做到取长补短、优势互补。然后着重介绍了混合反演算法的国内外研究现状，提出利用遗传算法和单纯形法相结合，得到一种高效、健全的非线性走时反演方法，进而形成一种新的地震层析成像方法。并对青藏高原、青藏高原东北缘的研究现状作了简要介绍。本文第二章介绍了遗传算法和单纯形方法的背景知识，详细阐述了基于变尺度混合反演算法的速度成像的基本原理：正问题采用有限差分波前走时计算，反问题采用遗传算法和单纯形相混合的反演方法，其成像策略是变尺度逐步逼近。具体成像过程是，把速度场划分为不同的空间尺度，定义网格节点上的速度作为待反演参数，采用双三次样条函数速度模型参数化，首先由遗传算法在较大的尺度范围内全局寻优，经过充分的演化后，将其符合终止条件的最佳个体提供给单纯形方法作为初始值，然后由单纯形方法进行快速局部寻优，这样结合的目的在于既降低计算成本，又避免陷入样本函数的局部极小值。然后，逐步减小空间尺度范围，重复上述过程，直至满足终止判别标准。本文第三章进行了一系列测试函数试验，将遗传算法、单纯形方法以及两者相混合方法的寻优能力做了对比，进一步阐述了全局寻优方法、局部线性化方法以及混合反演算法各自的优缺点。进行了低速异常体、高速异常体、梯度变化体、向斜、背斜、直立断层等速度模型的数值模拟试验。试验结果表明，基于变尺度混合反演算法的速度成像方法是有效的。利用对低速异常体数值模拟的观测走时加上5％和10％水平的随机噪声，然后进行速度成像的抗噪声试验，其结果表明5％的随机噪声对成像结果几乎没有影响，10％的随机噪声对成像结果影响不大，从而验证了本文提出的速度成像方法的健全性。本文第四章应用基于变尺度混合反演算法的速度成像方法对阿尼玛卿缝合带及其两侧的上地壳速度结构进行了成像，并将成像结果与他人的结果做了相应的对比。成像结果表明，阿尼玛卿缝合带东段的上地壳速度结构呈现一个横向宽度大于20km的低速度带的特征；研究区浅地表(2km以上)速度横向变化不大，2km以下速度横向变化剧烈，260～280km桩号之间呈现一个低速带，280～290km桩号之间，速度值突然抬升，290～300km桩号之间又出现一个规模较小的低速条带，300～310km桩号之间速度急剧升高，310km桩号已北的西秦岭褶皱带内，按照速度特征大体分为两段，340km桩号为速度分界线，以南为高速，以北为低速。库赛湖—玛沁断裂穿过283km桩号附近，速度从低速剧变为高速，基底深度由深突然变浅；在320～330km桩号之间，速度横向变化亦较大，基底深度从2.2km突然加深至4.5km，此处是武都—迭部断裂的体现；340km桩号是低速与高速的分界线，是舟曲—两当断裂的反映。本文第五章利用Zelt的Rayinvr软件包对阿尼玛卿缝合带及其两侧的二维地壳速度结构进行了研究，同时对Zelt的Rayinvr软件包的使用技巧进行了探讨。能够处理首波是Zelt的Rayinvr软件包的一个特点，可以将Pg波震相视作来自基底以上的回折波和来自基底界面的首波，进行有关基底的速度和界面深度的同时成像。结果表明，在阿尼玛卿缝合带内基底界面剧烈下凹，最深达5.47km；阿尼玛卿缝合带两侧相对而言，西秦岭褶皱带的基底埋深较松潘—甘孜微块体浅，在缝合带南侧的松潘—甘孜微块体内，基底埋深在3.5kin左右，基底界面在松潘—甘孜微块体也呈一定的下凹形态，在桩号170～250kin之间，基底下凹深度达4.0km；从阿尼玛卿缝合带过渡到西秦岭褶皱带，基底界面急剧变浅至1.8km，继而又急剧变深至4.7kin的深度，之后变得平坦。在使用Zelt的Rayinvr反演程序时，可以试验不同的反演方案，探寻不同的反演参数的效果，最好的方案是找到合适的参数，进行速度和深度的同时反演，其次是先速度后深度的反演方案。获得了关于阿尼玛卿缝合带及其两侧的二维地壳速度结构的一些重要认识：阿尼玛卿缝合带东段在深度20-45公里范围，存在贯穿整个中下地壳的低速构造，相同深度速度低于两侧约0.2～0.3km／s，这种低速构造的分布自上而下逐步减小的趋势。阿尼玛卿缝合带两侧的南北地壳结构存在明显区别，南侧的复杂程度明显高于北侧。整个地壳厚度沿测线横向变化不大，大约48～51km，阿尼玛卿缝合带略有增厚。松潘—甘孜地块有向西秦岭褶皱带下地壳俯冲的迹象。研究揭示的地球动力学含义是，研究区的构造背景以走滑、水平错断为主，下地壳物质有侧向流动的可能性。最后在第六章中对本文的研究成果进行了总结。更多还原

【Abstract】 In this thesis, the development history of seismic tomography and geophysicalinversion methods is reviewed first. The advantages and defects of the globaloptimization algorithms and local linearized methods are discussed. It is pointed outthat the combination of these two types of methods will make it possible to overcomethe weakness caused by using any single method. Then, the current status of researchon the hybrid inverse methods is introduced. It is put forward that the combination ofgenetic algorithms (GA) and simplex methods will be an efficient and robust strategyof non-linear travel time inverse methods, and then a new seismic tomographymethod is developed. Besides, a brief introduction is given to the actual status ofresearch on crustal structures of Qinghai-Tibet Plateau and its northeastern margin.In Chapter 2, the general description about genetic algorithms and simplexmethods is presented. The basic principle of seismic velocity imaging based on themulti-scale hybrid inverse algorithms is expounded in detail. The forward problem issolved by use of the finite-difference method, and the hybrid method combininggenetic algorithms and simplex algorithms is applied to the inverse problem. Themulti-scale successive approximation strategy is adopted in the inverse process. Theimaging process is that the velocity field is firstly divided into different spatial scales,the velocities on the grids are taken as the inverted parameters, and then the model isparameterized by a bicubic spline function. Genetic algorithms (GA) are used firstlyto search the global optimization in a larger scale. The best individual obtained bygenetic algorithms, which experienced fully evolutions, acts as the initial model forthe simplex methods. And then the simplex methods are used to search rapidly thelocal optimization. The spatial scale will be reduced when the terminal conditions aresatisfied. The process mentioned above is repeated until the terminal conditions aresatisfied. The reason for combining genetic algorithms and simplex methods is that itcan not only reduce the calculation costs, but also avoid falling into the localoptimization area.In Chapter 3, a series of function tests and numerical simulated tests aredescribed. In the function tests, the searching abilities among genetic algorithms,simplex methods and their hybrid algorithms are compared. The advantages and defects of global, local and their hybrid algorithms are discussed in detail. Thegradient velocity, anomalous low velocity, anomalous high velocity, syncline,anticline and steep fault models are involved in the numerical simulated tests. All theresults of numerical simulated tests show that the velocity imaging method based onmulti-scale hybrid inverse algorithms proposed here is efficient. Anti-noise tests arealso done by adding 5% and 10% level random noise to the theoretical traveltimes ofthe anomalous low velocity model. It shows that the 5% level random noise makesalmost no differences to the inverse result and the 10% level random noise makes afew changes to the inverse result. The result of anti-noise tests verify that the velocityimaging method mentioned above is robust.In Chapter 4, the velocity imaging method based on the multi-scale hybridinverse algorithms is applied to research the upper crustal velocity structure inA’nyemaqen suture zone and its adjacent areas. The results obtained here arecompared with the others. The velocity imaging shows that the upper crustal velocitystructure of the eastern part of A’nyemaqen suture zone is characterized by a lowvelocity distribution, with a width of more than 20 km. The velocity structure abovethe depth of 2km is almost homogeneous laterally, but strong heterogeneities ofvelocity structures appear below the depth of 2km. A low velocity zone appearsbetween the stake numbers 260kin and 280kin. The P wave velocities increase rapidlybetween the stake numbers 280km and 290km. A smaller scale low velocity zoneappears between the stake numbers 290km and 300km, and the velocities increaserapidly again between the stake numbers 300km and 310km. The P wave velocitydistribution in the West Qinling fold zone to the north of the stake number 310km canbe divided into two parts depending on their velocity features. The P wave velocitiesare higher in the area to the south of the stake number 340km as compared with thevelocities in the area to the north. The Hoh Sai Lake-Maqin fault goes through theprofile at the stake number 283km, where the P wave velocities change severely andthe depths of the basement vary suddenly as well. The P wave velocity obviouslyvaries laterally between the stake numbers 320km and 330km, where the depth of thebasement varies in the range between 2.2km and 4.5km, and Wudu-Tewo fault goesthrough here. The velocity transition boundary between low and high velocities is located near the stake number 340km., where Zhouqu-Liangdang fault goes through.In Chapter 5, the 2D crustal section of A’nyemaqen suture zone and its adjacentarea is constructed by use of Rayinvr software package written by Zelt. Meanwhile,the skills for using the package are discussed. Rayinvr software package can deal withhead wave data, and that is its major distinguishing characteristic. The Pg waves canbe regarded as turning waves or head waves penetrating the upper crust, and thus itcan be used to reconstruct both of velocity distributions of the upper crust andtopography of the basement interfaces. The results reveal that the basement inA’nyemaqen suture zone is concave dramatically. Its maximum depth is about 5.47kmin it. As regards the both sides of A’nyemaqen suture zone, the basement in WestQinling fold zone is shallower than in Songpan-Ganzi geological block. The depth ofthe basement interface in Songpan-Ganzi block is about 3.5km. The concave featureof the basement interface also appears in Songpan-Ganzi block, and the depth of thebasement is about 4.0km between the stake numbers 170km and 250km. In thetransition zone between A’nyemaqen suture zone and West Qinling fold zone, thebasement shallows rapidly up to a depth of 1.8km, then deepens northwards up to4.7km, and finally becomes flat afterwards. Different inverse schemes and parametersare tested in order to compare their effects, when Rayinvr software package is used. Itis found that the simultaneous inversion of velocities and depths with suitableparameters is better than their individual inversion.The following are some major results about the 2D crustal section ofA’nyemaqen suture zone and its adjacent area. There are low velocity distributions inthe middle and lower crust in the depth range from 20km to 45km in the easternsegment of A’nyemaqen suture zone. The decrease of velocity values is about0.2～0.3km/s in A’nyemaqen suture zone and the scale of the low velocity distributionsreduces with depths. Obvious discrepancies of velocity structures exist between thesouthem and northem parts of A’nyemaqen suture zone: a more complicated structureis found in the south as compared to the north. Thickness of the crust varies very littlealong the profile, with an approximate value of about 48～51km, though a slightincrease appears in A’nyemaqen suture zone. Imaging of the crustal section shows thatthe lower crust of Songpan-Ganzi block dives under West QinLing fold zone. Geodynamic implications revealed by the results can be summed up as follows: thatstrike-slip and horizontal dislocating tectonic is the major tectonic background in thisarea and substance of lower crest tends to flow laterally. Finally, the results aresummarized in Chapter 6.更多还原