长沙理工大学学报（自然科学版）2025，Vol.22Issue(4)：161-171,11.DOI:10.19951/j.cnki.1672-9331.20231129002
基于贝叶斯神经网络的锈蚀RC梁抗弯承载力时变概率模型
Time-varying probability model of flexural capacity of corroded reinforced concrete beams based on Bayesian neural network
摘要
Abstract
[Purposes]This paper aims to analyze the uncertainty of the parameters affecting the corrosion of steel bars in reinforced concrete(RC)beams and the uncertainty of the resistance prediction model and then accurately evaluate the flexural capacity of corroded RC beams.[Methods]Firstly,based on the model prediction results and experimental data,the uncertainty coefficient of the model expressed by the longitudinal reinforcement corrosion rate was proposed.Then,by using the Monte Carlo and Latin hypercube sampling methods to establish a probability model for the influencing parameters of the flexural capacity of corroded RC beams,a time-varying prediction model for the flexural capacity of corroded RC beams based on a Bayesian neural network was established.The model was compared with the traditional back propagation(BP)neural network and the genetic algorithm(GA)-BP neural network prediction model.Finally,based on the case,the time-varying probability distribution of the flexural capacity of corroded RC beams within 100 years of service life was obtained.[Findings]Compared with the BP neural network and GA-BP neural network prediction model,the Bayesian neural network prediction model has the least number of iterations and the highest efficiency,and its average prediction accuracy is comparable to the current general GA-BP neural network prediction accuracy.The average value of the ratio of the predicted value to the experimental value is 0.99,and the variance is 0.007 0.The prediction accuracy of the Bayesian neural network is 0.074 4 higher than that of the traditional BP neural network.[Conclusions]The time-varying probability model of the resistance of corroded RC beams satisfies the lognormal distribution.The established time-varying prediction model for the resistance of corroded RC beams based on a Bayesian neural network can obtain more accurate results when assessing the flexural capacity of corroded RC beams.关键词
钢筋混凝土梁/锈蚀/抗弯承载力/贝叶斯神经网络/模型不确定性系数Key words
reinforced concrete beam/corrosion/flexural bearing capacity/Bayesian neural network/model uncertainty coefficient分类
交通工程引用本文复制引用
管海江,张军辉,张石平,唐利民..基于贝叶斯神经网络的锈蚀RC梁抗弯承载力时变概率模型[J].长沙理工大学学报（自然科学版）,2025,22(4):161-171,11.基金项目
国家重点研发计划项目(2021YFB2600900) （2021YFB2600900）
湘江实验室重大项目(22XJ01009) （22XJ01009）
国家自然科学基金资助项目(52278434) （52278434）
湖南省教育厅自然科学研究重点项目(21A0206) Project(2021YFB2600900)supported by the National Key Research and Development Program of China （21A0206）
Project(22XJ01009)supported by the Major Program Project of Xiangjiang Laboratory （22XJ01009）
Project(52278434)supported by the National Natural Science Foundation of China （52278434）
Project(21A0206)supported by Research Foundation of Education Bureau of Hunan Province （21A0206）
