流体饱和多孔隙介质多参数反演的小波多尺度-正则化高斯牛顿法

Multiparameter Inversion in the Fluid-saturatedz Porous Medium with Wavelet Multiscale-regularized Gauss Newton Method

将小波多尺度方法和正则化高斯牛顿法相结合,充分利用两种方法的优点,以小波尺度分解作为引导算子确保反演算法的搜索路径,在每一个分解后的尺度上采用正则化高斯牛顿法作为求解算子以解决反问题的不适定性问题,构造了小波多尺度-正则化高斯牛顿法,有效地解决了流体饱和多孔隙介质多参数反演过程中的局部极值和不适定性的问题.通过与传统的正则化高斯牛顿法数值比较,显示了小波多尺度-正则化高斯牛顿法法是一个大范围收敛方法.数值模拟的结果验证了方法的有效性和可行性.

To overcome ill conditioning and multiple minima in inversion wavelet muhiscale-regularized Gauss Newton method is presented muhiparameter inversion of fluid-saturated combination of three operators：a restriction process efficiently , a and applied to the porous medium. The method is described as the operator, a relaxation operator, and a prolongation operator. The wavelet scale decomposition and reconstruction are regarded as the lead operator （including the restriction operator and the prolongation operator） to promote a better searching path, the regularized Gauss Newton method is chosen as the relaxation operator to solve the ill conditioning at each scale. The results of numerical simulations demonstrate that the method is a widely convergent optimization method and exhibits the advantages of conventional regularized Gauss-Newton method on computational efficiency and precision.