黑龙江科技大学学报2025，Vol.35Issue(1)：95-102,8.DOI:10.3969/j.issn.2095-7262.2025.01.015
基于试验数据库的神经网络RC柱恢复力预测
Prediction of RC column restoring force based on neural network with test database
摘要
Abstract
This paper seeks to address the long time consuming,difficulty in solving,and problems hard to deal with by resilience model in the process of conducting the elastic-plastic time-course analysis of building structures relying on traditional finite element analysis methods.The study involves adopting 253 groups of the structural parameters and hysteresis curve characteristics of RC columns as the Input vectors collected by Pacific Earthquake Engineering Research Center(PEER)under the action of the re-ciprocating anthropomorphic static forces,training the BP neural network RC column restoring force mod-el,and analyzing the prediction accuracy of the neural network structural restoring force under the differ-ent numbers of input variable parameters and different numbers of training samples.The results show that when the input vector only considers the RC column structural parameters to train the network resilience model,the network is not able to predict the resilience well.Compared with the input variable consider-ing three variables,the prediction accuracy of the neural network with five and eight variables is improved by 76.3%and 44.4%,respectively.With the increase of the number of training samples,the prediction accuracy and training efficiency of the neural network are also improved,and the accuracy can be im-proved by a maximum of 88.23%.关键词
神经网络预测/RC柱恢复力/输入变量参数/训练样本数量Key words
neural network prediction/RC column restoring force/input variable parameters/num-ber of training samples分类
信息技术与安全科学引用本文复制引用
王涛,周雨晨,孟丽岩,谢婧怡,孙立飞..基于试验数据库的神经网络RC柱恢复力预测[J].黑龙江科技大学学报,2025,35(1):95-102,8.基金项目
国家自然科学基金项目(52278173 （）
52078398) （）
