论文部分内容阅读
复杂地球物理资料的反演问题往往是一个求解多参数非线性多极值的最优解问题。而鸟和蚂蚁等群体觅食的过程,正好与寻找地球物理反演最优解的过程相似。基于自然界群体协调寻优的思想,本文提出了交叉学科的群体智能地球物理资料反演方法,并给出了其对应的数学模型。用一个有无限多个局部最优解的已知模型对该类方法进行了试验。然后,将它们应用到了不同的复杂地球物理反演问题中:(1)对噪声敏感的线性问题;(2)非线性和线性同步反演问题;(3)非线性问题。反演结果表明,群体智能反演是可行的。与常规遗传算法和模拟退火法相比,该类方法有收敛速度相对快、收敛精度相对高等优点;与拟牛顿法和列文伯格一马夸特法相比,该类方法有能跳出局部最优解等优点。
The inversion of complex geophysical data is often an optimal solution to the multi-parameter nonlinear multi-extremum. The foraging process by groups such as birds and ants happens to be similar to finding the optimal solutions for geophysical inversion. Based on the idea of the coordinated optimization of nature groups, this paper proposes a population intelligent geophysical data inversion method of interdisciplinary, and gives its corresponding mathematical model. This type of method has been tested with a known model with an infinite number of local optimal solutions. Then, they are applied to different complex geophysical inversion problems: (1) noise-sensitive linear problems; (2) nonlinear and linear synchronous inversion problems; and (3) nonlinear problems. The inversion results show that the population intelligent inversion is feasible. Compared with the conventional genetic algorithm and the simulated annealing method, this kind of method has the advantages of relatively fast convergence speed and relatively high convergence precision. Compared with Quasi-Newton method and Levenberg-Marquardt method, this kind of method can jump out of local optimum Solutions and other advantages.