结构可靠度逐步逼近径向基神经网络响应面法

提出了逐步逼近径向基神经网络响应面法计算结构可靠度,这种数值方法较好地解决了结构功能函数非线性及结构随机变量非正态分布时采用结构可靠度指标度量结构可靠度存在误差的问题.经过多个数值试验表明,该算法迭代收敛迅速、计算准确性较高、应用过程简单.利用这种方法对含有多个随机变量的结构进行可靠性分析和设计,特别是当结构功能函数具有较强的非线性或结构随机变量分布具有较大的偏斜度时,具有一定的应用价值.

An iterative RBF ANN response surface method for structural reliability analysis

Structural reliability index for response surface method of RBF ANN was developed,which could resolve the error of structural reliability when the structure performance function was nonlinear or the variables were nonnormally distributed.An improved approach named as iterative radial basis function(RBF) artificial neural network(ANN) response surface method(RSM) was characterized by rapid iterative convergence speed,higher calculation accuracy and simple process.This method allows for reliability analysis and design of structure comprising multiple random variables,especially when the structure performance function is highly nonlinear or the variables are nonnormally distributed.