误差反传播神经网络法地震反演

SEISMIC INVERSION USING ERROR-BACK-PROPAGATION NEURAL NETWORK

针对测井资料约束下的地震反演具体问题，在假定反演目标和地震资料之间存在某种非线性映射的情况下，用神经网络逼近反演问题中的正演和反演过程，综合构成一个大网络系统，并根据地震反演的具体问题，给出该系统的能量函数．系统采用误差反传播法进行学习，从而实现用神经网络自适应地外推测井资料，有机地将神经网络与地震反演结合起来．对实际资料的测井速度外推结果表明，此法具有好的应用前景．

In connection with inversion problem restrained by well log information,a large network system is built in this paper to approach the direct modeling and inversing process of the inverse problem on condition that a nonlinear map exists between the inverse goal and the seismic data.The energy function of the system is given in the light of inverse problem, and the error back propagation learning algorithm is used to realize adaptively extrapolatting well log information,so the neural network is Combined with inverse problem better here. The result of the extrapolation of well log velocity using actual information shows: this method has good applied prospects.