摘要
基于反演问题的不确定性和目标函数的多峰性,本文提出引入小生境遗传算法,采用自适应控制交配概率和变异概率,求解出目标函数的若干个局部峰(或全局峰),然后利用先验知识,判定得到满意解。本文对实测资料进行了层速度与厚度的反演,通过实际值与反演值的比较,验证了进行多峰优化的有效性。
Due to the indeterminacy of inversion and multimodal of object function, a niching genetic algorithm is introduced in this paper and the adaptive control of probability of crossover and mutation is used. Several local peaks or several global peaks of the object function are found first, the satisfied solution is obtained based on prior information accordingly. The genetic algorithm has been used in interval velocity and thickness inversion of seismic data in detecting the structure of a dykes. Comparing the practical velocity and thickness with the inversion velocity and thickness, the efficiency of multimodal optimization is validated.
出处
《工程力学》
EI
CSCD
北大核心
2002年第6期16-21,共6页
Engineering Mechanics
基金
清华大学基础研究基金(199914004)
