模拟退火独立分量分析方法及其应用

Simulated annealing independent component analysis and its application

（1.西南石油大学，四川成都610500； 2.中国石油集团工程设计有限责任公司西南分公司，四川成都610017）

Southwest Petroleum University, Chengdu 610500, China

独立分量分析通过追求多源混合信号中各个独立源的非高斯性，使各个分量获得统计学上的独立，但现有的诸多独立分量分析方法普遍追求的是运算速度，忽略了对算法本身精确性的要求。为提高分离信号的质量，在原有独立分量分析算法的基础上，结合非线性优化算法——模拟退火算法，提出了一种新的独立分量分析算法——模拟退火独立分量分析算法。给出了独立分量分析方法的基本原理和目标函数；讨论了模拟退火独立分量分析算法的实现过程。数值模拟和实际资料处理表明，该方法在叠前地震记录中能很好地将有效信号和干扰信号分离出来，从本质上起到了压制噪声、突出有效信号的叠前去噪处理作用，改善了叠加剖面的品质。信噪比分析表明，经模拟退火独立分量分析方法去噪处理后，叠加剖面的信噪比由1.3936提高到2.1056。

The independent component analysis is to make each component be independent statistically through pursuing the nongaussianity of each independent source in multiple source composite signals. Most of the existing independent component analysis parsing algorithms pursue the computing speed but neglect the accuracy. In order to enhance the quality of the separated signal, the simulation annealing independent component parsing algorithm is proposed based on the original independent component parsing algorithm and the simulation annealing algorithm. The basic principles and the objective function of the independent component analysis method have been given and the realizing procedures of the method discussed. The numerical modeling and the actual data processing indicate that the method can well separate the effective signals from the interfering signals in prestack seismic record and improve the quality of the stacking profile. By using this de-noise processing method, the signal to noise ratio of the stacking profile can be increased from 1.3936 to 2.1056.