中文摘要：

中国大陆中西部乃至全球造山带普遍具有复杂地壳结构．随着矿产资源勘探和深部探测研究的深入，探测造山带及盆山耦合区下方地壳精细结构正逐渐成为当前面临的巨大挑战．人工源深地震测深方法正越来越清晰地揭示出不同构造域地壳速度结构的基本特征，然而传统的层状结构模型参数化方法难以准确描述复杂地质模型，通常情况下多忽略速度结构的精细间断面且采用层边界平滑处理，难以满足地壳精细结构成像的发展要求．针对上述困难，本文采用最近发展的块状结构建模方案构建三维复杂地壳模型，基于逐段迭代射线追踪正演走时计算方法，推导了走时对三角形界面深度以及网格速度的偏导数，开展了非线性共轭梯度走时反演方法研究．发展了利用直达波和反射波等多震相走时数据对界面深度和网格速度的多参数联合反演方法，并引入不同种类震相数据的权系数和不同类型参数偏导数归一化的方法．数值算例表明，基于块状结构的非线性共轭梯度走时反演方法适用于复杂地壳结构模型，在利用人工源走时数据反演复杂地壳精细结构领域具有良好的应用前景．

英文摘要：

Complex crustal structures generally characterize global orogenic belts and Midwest China. With the deepening of mineral resources exploration and increasing detection of the deep earth, it is becoming a great challenge to use new methods to probe fine crustal structure beneath orogenic belts and basin-mountain coupling regions. Basic characteristics of crustal structure of different tectonic domains are becoming clearer and clearer in deep seismic soundings. However, it is difficult for the traditional layered structure modeling method, in which layer boundaries are smoothing and fine velocity discontinuities are often ignored, to describe complex geological models, making it difficult to meet the development requirements of fine structure imaging of crust. In view of the above difficulties, based on a newly developed block modeling scheme to describe three-dimensional complex crustal models and a corresponding segmentally iterative ray- tracing （SIRT） method, we derive the travel-time partial derivatives of triangular interface depth and grid velocity, and develop a 3-D nonlinear conjugate gradient travel-time inversion method. Block modeling scheme is able to construct any complex geological models in theory and can be used to efficiently build initial models by combining various priori velocity and interface information. In the inversion process, PRP （Polak-Ribiere-Polyak） type of the conjugate gradient method is used to solve the constrained damping least squares problem. We develop the joint inversion of grid velocity and interface depth based on multi seismic phases like direct waves and reflected waves, and make a great improvement of the inversion resolution compared with the traditional method which is based on single phase. To improve the convergence accuracy of inversion results, strategies like different weighting factors for different seismic phases and normalization of travel-time partial derivatives of different parameters are introduced in the joint inversion process. Numerical examples a