基于均匀试验人工神经网络的实用可靠性方法

根据均匀试验的设计提出了一种全新的可靠性计算方法，它有机地综合了人工智能网络.人工神经网络弃用有限单元法（FEM）而改用该方法，一个根本原因在于它能极大地减少计算量.首先，根据随机变量分布情况，通过均匀试验提取出有限样本，将它们看作是有限元法分析的输入参数.其次，基于有限元分析结果利用这些有限的训练样本构建最优的人工神经网络.利用最优人工神经网络的泛化能力，随机得到一个有效的响应值，然后计算出结构系统的可靠性指数.最后，这一计算方法还为进行可靠性分析提供新尝试，在对复杂系统进行实际试验时该方法表现得切实可行且有效.

Artificial Neural Reliability Calculating Method Based on Uniform Test Network

This paper presents a design of a new reliability calculation method based on uniform test,which organically combines artificial intelligence network.Artificial neural networks abandons the finite element method (FEM) and applies this method,because  the new law can greatly reduce the amount of computation.First,according to the distribution of the random variable,limited samples are extracted by uniform test,which are taken as for the input parameters finite element analysis.Secondly,based on the finite element analysis results,the limited training samples are used to  construct the optimal artificial neural network.Optimal artificial neural network generalization ability is applied to obtain a valid response,and then the reliability index of the structural system is calculated.Finally,this calculation method provides a new attempt  for reliability analysis in the actual testing of complex systems,which is proved to be  practical and effective.