基于优化算法的能量等值延性指标评价

Evaluation on Energy Equivalent Ductility Index Based on Optimized Algorithm

摘要:

设计多层前馈BP、径向基RBF和遗传优化GABP算法训练并预测能量等值法极限承载延性指标μΔe，采用5个控制参数及PEER库与拟静力试验作为训练和预测样本.编写BP及其隐含层节点数和学习率的优化算法，编写RBF及其节点数和中心宽度的优化算法，采用GA优化BP算法权值wij和阈值θj并抑制局部最优及逼近数据规律，并实现算法预测足尺框架柱延性指标μΔe值.研究表明GABP具有最佳适用性和计算优越性，提供结构损伤评价新方法.

Abstract:

back propagation(BP), radial basis function(RBF) and genetic algorithmback propagation (GABP) are used to train and predict the ultimate ductility index μΔeof dualenergy equivalent method. Five control parameters and PEER database and quasistatic test are used as training and prediction samples. BP and optimization algorithm of its node numbers in hidden layer and learning rate are compiled, RBF and optimization algorithm of its node numbers and central width are compiled. Genetic algorithm (GA) is used for optimizing weight wij and threshold θj in backpropagation (BP) to inhibit the local optimum, making relationship between testing data approximate to the real mapping principle. Algorithms is realized to predict μΔe of fullscale columns. Research indicates GABP has optimal adaptability and computing superiority, provide new method for structural damage evaluation.