机械设计特征值反问题求解的神经网络算法研究

Research on the Inverse Solution of Eigenvalue Problem in Machine Design Based on Artificial Neural Network

研究了特征值反问题求解的几种神经网络模型:直接逆模型,间接逆模型,优化方法模型,指出了各种方法的应用范围。研究表明,神经网络的直接逆模型使用范围十分狭窄,间接逆模型不一定收敛,而基于有限元分析的神经网络-优化方法模型,在结构参数到频率响应之间的映射关系为非双射时,也完全能获得正确结果。数值实例表明了该方法的有效性。

In this paper,several artificial neural network based models of solving inverse eigenvalue problem and their characters are studied. These models are direct,indirect and optimization inverse models. Researches show that the optimization inverse model based on artificial neural network and genetic algorithm is a universal means of solving inverse eigenvalue problem,while the application of direct model is limited and the convergence of indirect model is uncertain.