文章摘要
郑垚*,韩洋*,杨泳*,彭翔* **.一种考虑多峰分布的反向不确定性传播方法[J].高技术通讯(中文),2025,35(11):1250~1262
一种考虑多峰分布的反向不确定性传播方法
An inverse uncertainty propagation method considering multi-peak distributions
  
DOI:10. 3772 / j. issn. 1002-0470. 2025. 11. 009
中文关键词: 反向不确定性传播分析; 多峰分布; 灵敏度矩阵; 置信区间; 概率密度函数
英文关键词: inverse uncertainty propagation analysis, multi-peaked distribution, sensitivity matrix, confidence interval, probability density function
基金项目:
作者单位
郑垚* (*浙江工业大学机械工程学院杭州 310023) (**浙江工业大学台州研究院台州 318014) 
韩洋*  
杨泳*  
彭翔* **  
摘要点击次数: 77
全文下载次数: 61
中文摘要:
      传统的不确定性传播分析方法在解决多峰分布问题时仍按照单一随机分布处理,导致效果不佳。本文旨在提出一种实用的方法,以保证在多峰随机分布的稀缺数据环境下更准确地执行反向不确定性量化。首先,基于先验知识和高斯混合模型(Gaussian mixture model, GMM)构建输入随机变量的多峰概率密度函数(probability density function, PDF)。其次,在输出数据的有限样本条件下,基于核密度估计(kernel density estimation, KDE)与最大信息熵原理,估计输入变量的统计参数值。最后,通过灵敏度矩阵和置信区间(confidence interval, CI)进行参数的校准验证,确保了统计参数结果的可靠性。通过算例对该方法的可行性和有效性进行了验证。
英文摘要:
      Traditional methods of uncertainty propagation analysis still follow a single random distribution when solving multi-peak distribution problems, leading to poor results. The aim of this study is to propose a practical approach to ensure more accurate execution of inverse uncertainty quantification in a scarce data environment with multi-peak random distribution. First, the multi-peak probability density function(PDF) of the input random variable is constructed based on prior knowledge and Gaussian mixture model(GMM). Second, the values of the statistical parameters of the input variables are estimated based on the kernel density estimation(KDE) with the principle of maximum information entropy under the finite sample condition of the output data. Finally, the calibration validation of the parameters is carried out through the sensitivity matrix and confidence intervals(CI) to ensure the reliability of the statistical parameter results. The feasibility and validity of the method are verified by arithmetic examples.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮