采用局部、非线性的广义最小残量方法，进行三维频率-空间域迭代法声波叠前全波形速度正演和反演方法研究。正演使用三个预条件和计算迭代更新残差的方法，使得迭代法的收敛速度快，从而获得稳定高效率的迭代算法。在计算机机群上并行地快速迭代求解三维频率域声波正演数值模拟。使用拉普拉斯域波形反演得到大量的低频成分，生成精确的初始速度模型。使用多尺度的策略（从低频到高频进行全波形反演）和合适的预条件技术，从而减少全波形反演解的非唯一性（即减轻全波形反演强烈的非线性）。使用多重网格循环的方法和合成炮的技术，很大程度提高了三维频率域迭代法声波叠前全波形速度正演、反演的计算速度。把每个迭代步化为线性问题，由最速下降法求解。深入研究、阐明控制三维频率-空间域迭代法声波叠前全波形速度反演的机制，提出减少全波形反演解非唯一性的策略和方法，反演得到高分辨率的速度，为现实大小的地质体的油气勘探提供有效的反演方法和技术。

Using a local nonlinear generalized minimal residual method, research will concentrate on 3D iterative acoustic prestack full-waveform velocity forward-modeling and inversion development in frequency-space domain. Forward-modeling uses three preconditioners and the method used to calculate the iterative update, to get a rapid convergence of iteration, so as to obtain a stable and high efficiency iterative algorithm. Parallel iteration for solving the 3D acoustic forward numerical simulation with a computer cluster in frequency-space domain. We will use Laplace domain waveform inversion to get a large amount of low frequencies, in order to obtain an accurate starting velocity model for full-waveform inversion. We use multiscale strategy (full-waveform inversion from low frequency to high frequency) and suitable preconditioning techniques, thus reducing the non-uniqueness of the full-waveform inversion solution (ie, reducing the strong non-linearity of the full-waveform inversion).We use multigrid cycle method and composite shot technology, improving the 3D iterative full-waveform velocity inversion's calculation speed in frequency-space domain significantly. Each iteration step into a linearized problem, solved by the steepest descent method. Thoroughly study and evaluate the mechanism of 3D iterative acoustic prestack full-waveform velocity inversion in frequency-space domain, proposing new strategies and innovative methods to reduce the non-uniqueness of the full-waveform inversion solution, to get a high-resolution velocity model, providing an effective inversion method and tool for oil and gas exploration on a realistic geological scale and in acceptable computational time.