计算物理 ›› 2000, Vol. 17 ›› Issue (5): 511-517.

• 论文 • 上一篇    下一篇

抛物型方程的演化参数识别方法

熊盛武, 李元香, 康立山, 陈毓屏   

  1. 武汉大学软件工程国家重点实验室, 湖北 武汉 430072
  • 收稿日期:1999-02-23 修回日期:1999-08-05 出版日期:2000-09-25 发布日期:2000-09-25
  • 作者简介:熊盛武(1967~),男,湖北咸宁,副教授,主要从事并行计算、演化计算方面的研究,武汉大学软件工程研究所430072.
  • 基金资助:
    国家自然科学基金;武汉市晨光计划资助项目

PARAMETER IDENTIFICATION PROBLEM FOR PARABOLIC EQUATIONS USING EVOLUTIONARY ALGORITHMS

XIONG Sheng-wu, LI Yuan-xiang, KANG Li-shan, CHEN Yu-ping   

  1. State Key Lab of Software Engineering, Wuhan University 430072, P R China
  • Received:1999-02-23 Revised:1999-08-05 Online:2000-09-25 Published:2000-09-25

摘要: 给出了一种利用演化计算方法求解微分方程中的参数识别类型反问题的方法。该方法把参数识别问题转化为泛函的优化问题用演化算法来求解,指定待定参数的函数类形式,用遗传算法(Genetic Algorithms)来演化待求参数的最优估计值,并将该方法运用于线性扩散方程和拟线性对流扩散反方程反问题的数值模拟中。

关键词: 反问题, 参数识别, 演化计算, 遗传算法

Abstract: A general approach based on evolutionary algorithms to inverse parameter identification problems of PDEs is introduced. The evolutionary algorithms is used for solving the inverse problem as an optimization problem. This approach can evolve the optimal coefficient estimation by genetic algorithms (GA) in the giving function space of unknown parameter from observation data objectively and automatically. The approach has also been applied to parameter identification problem of linear and quasi linear parabolic equations. The numerical results demonstrated the effectiveness of this method.

Key words: inverse problem, parameter identification, evolutionary algorithms, genetic algorithms

中图分类号: