采用MEA-AdaBoost-BP模型的工程结构可靠性分析方法  

Reliability Analysis Method of Engineering Structure Based on MEA-AdaBoost-BP Model

作　　者：胡启国[1] 李致明 HU Qiguo;LI Zhiming(School of Mechanical and Vehicle Engineering,Chongqing Jiaotong University,Chongqing 400074,China)

机构地区：[1]重庆交通大学机电与车辆工程学院,重庆400074

出　　处：《华侨大学学报（自然科学版）》2022年第3期291-296,共6页Journal of Huaqiao University(Natural Science)

基　　金：国家自然科学基金资助项目(51375519);重庆市基础科学与前沿技术研究专项(cstc2015jcyjBX0133)。

摘　　要：针对工程结构可靠性设计中算法和计算存在的问题,提出基于MEA-AdaBoost-BP神经网络算法模型的可靠性求解方法.运用思维进化算法(MEA)求解训练集权值和阈值优化的BP神经网络,并构造为弱预测器函数.然后,运用AdaBoost算法将多个优化后的BP神经网络弱预测器函数迭代训练,形成MEA-AdaBoost-BP神经网络算法模型强预测器函数.最后,利用逼近隐性功能函数求解可靠性指标,并将其与AdaBoost-BP算法和Monte-Carlo算法进行比较.研究结果表明:所提算法在计算中与Monte-Carlo算法相比,其迭代次数分别仅为16次和46次,效率高,计算精度与Monte-Carlo法接近;而和AdaBoost-BP法相比,其可靠性指标误差分别仅为1.59%和1.88%,计算结果更精确.Aiming at the algorithm and calculation in engineering structure reliability design,a reliability solving method was proposed based on MEA-AdaBoost-BP neural network algorithm model.The optimized BP neural network was constructed as a weak predictor function by solving the training set weights and threshold values through the Mind Evolutionary Algorithm(MEA).Then the AdaBoost algorithm was used to iteratively train several optimized BP neural network weak predictor functions to form the strong predictor function of the MEA-AdaBoost-BP neural network algorithm model.Finally,the approximate the implicit function was used to solve the reliability index,which was compared with AdaBoost-BP algorithm and the Monte-Carlo method.The results show that the proposed method is efficient because its number of iterations is only 16 and 46 respectively compared with the Monte-Carlo method and its calculation accuracy is close to that of the Monte-Carlo method;the error of reliability index is only 1.59% and 1.88% respectively compared with that of AdaBoost-BP method,and the calculation result is more accurate.

关 键 词：可靠性指标 思维进化算法(MEA) AdaBoost-BP神经网络 MEA-AdaBoost-BP算法 强预测器函数 

分 类 号：TB114.3[理学—概率论与数理统计] TP183[理学—数学]