借助Matlab遗传算法反演火山区压力源参数  

Inversion of pressure source parameters based on Matlab GA toolbox

作　　者：姜友谊[1] 巨小文 胡亚轩[3] 王雄[3] 

机构地区：[1]西安科技大学测量工程系,陕西西安710054 [2]陕西省测绘局产品质量监督检验站,陕西西安710054 [3]中国地震局第2监测中心,陕西西安710054

出　　处：《西安科技大学学报》2008年第1期91-95,共5页Journal of Xi’an University of Science and Technology

基　　金：国家自然科学基金资助项目(40574041);地震科学联合基金资助项目(605004)

摘　　要：首先分析了Mogi模型的经典最小二乘反演由于出现矩阵奇异或接近奇异而发散的局限性,决定采用具有全局高效搜索性和优化计算时不需要求导的遗传算法来反演火山区压力源参数。反演程序调用Matlab7.0 Release14工具箱中的选择、交叉和变异等相关程序。最后应用该程序和长白山天池火山区水准和GPS观测资料联合反演火山区压力源参数,经过比较遗传算法和阻尼最小二乘方法的反演结果,可看出2种方法结果近似,可见此方法方便快捷地实现了目标函数的最小化。First limitation of classic least squares method in Mogi model is analyzed, because the method can cause matrix singular or near singular. Genetic Algorithm （GA） can be used better to inverse the pressure source parameters because searching is entire and efficient and it need not calculate partial derivative. Then the Matlab programs based on Matlab7.0 Release 14 including the main GA process-selection,crossover and mutation etc designed in the toolbox is introduced. At last, through comparing the results based on both vertical and GPS deformation data in Chang-baishan-Tianchi volcano using different methods between the GA and damping Least squares method, the conclusion is GA can be used to realize the smallest targeted function.

关 键 词：遗传算法 最小二乘法 火山形变 

分 类 号：P317[天文地球—固体地球物理学]