摘要
本文基于多尺度反演思想提出波动方程多尺度反演方法。该方法的基本思路是:对反演的目标函数进行多尺度分解,在每一尺度下利用正则化方法反演求解,克服了反演问题的不适定性,获得全局最优解,使得反演结果的分辨率和保真性得到提高。本文利用小波变换实现多尺度反演,并且是先通过在大尺度上反演得到一个比较好的参数估计,再将此估计作为较小尺度的初值进行反演,直至获得全局最优解。文中的应用实例说明了该方法具有对初始模型依赖性低、收敛速度快、反演结果稳定的优点。
Based on the multi-scale idea, the paper presented a method of multi-scale inversion of wave e-quation. The basic ideas of the method are as follows : multi-scale decomposition of inverted objective function and inversion is conducted by using normalization method under each scale,which overcome the ill-posedness of inversion problem, acquired global optimum solution and improved resolution and hi-fi of inverted results. The paper uses wavelet transform to conduct multi-scale inversion. The better estimation of parameters is acquired first by inversion in large scale,then the estimated values are taken as initial values in smaller scale to invert until the global optimum solution is obtained. The application cases showed that the method has advantages of low dependence on initial model, rapid convergent speed and stable inverted results.
出处
《石油地球物理勘探》
EI
CSCD
北大核心
2005年第3期273-276,共4页
Oil Geophysical Prospecting
