【摘要】 利用径向基函教(Radial Basis Function,RBF)神经网络来预测结构初期损伤对整体的影响,可以有效地判断结构的稳定性.由于神经网络可以通过对样本的反复学习来反映整体结构复杂的非线性演化关系,其预测精度可以满足要求.RBF神经网络作为一种性能良好的前馈网络,具有更好的逼近能力和全局最优特性.本文通过有限元计算得出样本作为基础,采用RBF神经网络建立初期损伤的预测系统,通过最近邻聚类学习算法实行整体结构预测,这种研究思路具有结构简单、学习速度快、预测精度高的特点,网络的外推能力也较强,计算效率明显优于传统方法.本系统采用Fortran语言编写,最后通过一个实例说明本系统的有效性及实用性.更多还原

【Abstract】 Using the RBF neural network to forecast the impact of the initial injury on the overall structure,it is effective in judging the stability of the structure.Because that the neural network can learn the samples repeatedly to reflect complex nonlinear evolution of the overall structure,the prediction accuracy can meet the requirements. As a good feed-forward network,the RBF neural network has a better approximation of optimal capacity and the overall optimal characteristics.Based on the samples calculated by the finite element method,this paper used RBF neural network to establish the initial injury forecast system and then used the nearest neighbor clustering algorithm to predict the overall structure.Such research idea has the characteristics of simplicity in structure,learning fast and high prediction accuracy.At the same time,this idea also has a stronger ability for extrapolation of the net- work,so its calculation efficiency is obviously better than that of traditional methods.The program is compiled by Fortran.At last,the effectiveness and practicality of the system have been proved by an example.更多还原