基于响应面重构的一种可靠度计算方法

大型复杂结构的可靠度分析中,极限状态方程往往无法显式表达。针对这种情况,文章采用二次多项式函数重构响应面来寻求设计验算点;然后在设计验算点附近,采用BP神经网络重构响应面,拟合极限状态函数;在此基础上,采用基于最优化原理的蒙特卡罗方法求解结构的可靠性指标。与其他单一的响应面重构方法相比,文中方法具有计算简便及精度高的特点;工程计算语言Matlab采用矩阵操作,并提供了包括优化、统计、神经网络在内的大量工具箱,而且语言简单,使得编程效率大大提高,能快速方便地实现可靠度计算程序的编制。

A kind of reliability computational method based on reconstruction of the response surface

In this paper,the quadratic polynomial function is used to reconstruct the response surface in search of the possible breakdown point,and the BP neural network is adopted to reconstruct the response surface around the possible breakdown point,fitting the limit state function.Then according to the method of Monte Carlo in the optimization theory,the reliability index of the structure can be obtained.Compared with other methods of reconstructing response surfaces,the method in this paper works easily and has high precision.The calculation program of reliability can be compiled conveniently and quickly by using the engineering calculation language,Matlab,for it adopts the matrix method and provides a great many toolboxes including such functions as optimization,statistics and neural network and the language is simple,making the program compiling efficient.