项目名称: 基于偏最小二乘理论的结构可靠度分析代理模型方法
项目编号: No.51308158
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 建筑科学
项目作者: 赵威
作者单位: 哈尔滨工业大学
项目金额: 25万元
中文摘要: 针对现有结构可靠度分析代理模型方法在处理高维非线性问题时存在计算精度不高的缺陷。通过引入偏最小二乘理论,探索结构可靠度分析高效高精度计算方法。首先,从样本角度出发,研究可靠度分析由样本间多重相关性引起计算精度不足的问题,建立最基本的线性偏最小二乘代理模型方法。以此为基础,针对高维变量问题,在偏最小二乘代理模型方法中引入递阶思想,将变量分块、建模分层,建立可靠度分析线性递阶偏最小二乘代理模型方法。同时,针对高度非线性问题,将偏最小二乘理论结合高维特征空间的径向基映射技术,研究可靠度分析非线性偏最小二乘代理模型方法。最后将递阶思想与非线性映射结合,建立一整套在偏最小二乘框架下的高精度高效率可靠度分析代理模型方法,有效解决传统方法计算精度和效率之间的矛盾,为大型复杂体系的结构可靠度分析开辟一条新思路。本项研究目的不仅在于丰富和完善结构可靠度理论并指导工程实际应用,而且对于其他相关学科具有借鉴价值
中文关键词: 结构可靠度;代理模型;偏最小二乘;高维非线性问题;实验设计
英文摘要: In order to improve the accuracy of the present reliability analysis methods when they are confronted with high-dimensional non-linear problems, this project investigates the structural reliability analysis methods by introducing the partial least squares. Firstly, from the point of view of samples, the lack of accuracy induced by the correlation between the samples will be studied, and the linear partial least squares surrogate model method will be established. Based on this, According to high-dimensional problems, by introducing the hierarchical idea of the variable block and modeling stratification, the linear hierarchical partial least squares method will be investigated. While according to the non-linear problems, the radial basis function mapping technology in high-dimensional feature space will be combined with the partial least squares theory, and the non-linear partial least squares surrogate model method will be also investigated. Finally, the hierarchical idea and the non-linear mapping technology will be combined to establish a set of high efficiency and accuracy surrogate model methods under the partial least squares framework for structural reliability analysis, and the methods provide a new concept for solving reliability problems of large-scale complex structures. This project is important not on
英文关键词: structural reliability;surrogate model;partial least square;high dimensional nonlinear problems;experimental design