结构可靠性灵敏度分析的低偏差抽样方法.pdfVIP

第 27 卷第 1 期 Vol.27 No.1 工 程 力 学 2010 年 1 月 Jan. 2010 ENGINEERING MECHANICS 104 文章编号：1000-4750(2010)01-0104-05 结构可靠性灵敏度分析的低偏差抽样方法 * 戴鸿哲， 王 伟 (哈尔滨工业大学土木工程学院，黑龙江，哈尔滨 150090) 摘 要：由于蒙特卡罗(MC)方法具有程序结构简单，收敛速度与问题维数无关等优点，故其在结构可靠性及可靠 性灵敏度分析中得到了广泛应用。但是计算效率低这一主要缺点限制了该方法的应用范围。通过引进单位超立方 体中不同的低偏差点集代替伪随机数序列，并结合重要抽样技术建立了结构可靠性灵敏度分析的低偏差抽样方 法。该方法不但可以大幅度减少抽样点数目，还能够得到确定性的估计值避免传统 MC 方法只能得到概率意义下 误差的缺陷。通过数值算例可以看出该方法具有较高的计算精度和效率。 关键词：结构可靠性；灵敏度；低偏差抽样；重要抽样；蒙特卡罗法 中图分类号：TU311.2 文献标识码：A LOW-DISCREPANCY SMAPLING METHOD FOR STRUCTURAL RELIABILITY SENSITIVITY ANALYSIS * DAI Hong-zhe , WANG Wei (Harbin Institute of Technology, School of Civil Engineering, Harbin, Heilongjiang 150090, China) Abstract: Monte Carlo (MC) method has found a wide application in structural reliability and reliability sensitivity problems not only for its simple program structure but also for its rate of convergence, completely independent from the dimension. The major disadvantage of MC method is its inefficiency due to the excessive number of samples and this property limits its application in structural reliability sensitivity analysis. This paper introduces and investigates various low-discrepancy sequences instead of pseudo random numbers to develop low-discrepancy sampling method for structural reliability sensitivity analysis by combing these sequences with importance sampling technique. The proposed method holds advantages over MC in particular by being more accurate and having a deterministic estimate of e