Modeling of network traffic is a fundamental building block of computer science. Measurements of network traffic demonstrate that self-similarity is one of the basic properties of the network traffic possess at large ...Modeling of network traffic is a fundamental building block of computer science. Measurements of network traffic demonstrate that self-similarity is one of the basic properties of the network traffic possess at large time-scale. This paper investigates the change of non-stationary self-similarity of network traffic over time,and proposes a method of combining the discrete wavelet transform (DWT) and Schwarz information criterion (SIC) to detect change points of self-similarity in network traffic. The traffic is segmented into pieces around changing points with homogenous characteristics for the Hurst parameter,named local Hurst parameter,and then each piece of network traffic is modeled using fractional Gaussian noise (FGN) model with the local Hurst parameter. The presented experimental performance on data set from the Internet Traffic Archive (ITA) demonstrates that the method is more accurate in describing the non-stationary self-similarity of network traffic.展开更多
针对多阶段退化产品,基于性能退化响应数据的回归统计拟合,建立变速率复杂退化可靠性评估模型。首先利用分段回归拟合对性能退化数据进行变点回归建模,推导变点分布的极大似然函数;其次鉴于模型的复杂性,变点估计的最大似然估计无法得...针对多阶段退化产品,基于性能退化响应数据的回归统计拟合,建立变速率复杂退化可靠性评估模型。首先利用分段回归拟合对性能退化数据进行变点回归建模,推导变点分布的极大似然函数;其次鉴于模型的复杂性,变点估计的最大似然估计无法得到显式解,采用分层Bayes进行建模;然后结合MCMC(Markov chain Monte Carlo)算法中的Gibbs采样对模型进行参数诊断,依据Schwarz信息准则(Schwarz information criterion,SIC)构建经验似然比对模型的变点进行估计与检验;最后根据产品失效退化的定义,推导该失效模式下产品的可靠度函数。以手机运行的游戏性能数据建模分析,演绎说明了多阶段退化模型适应性强、可行性高的特点。与传统两阶段建模相比,多阶段退化建模考虑了各个阶段的退化信息,提高了数据利用率和产品的可靠性评估可信度。展开更多
基金the National High Technology Research and Development Program (863) of China(Nos. 2005AA145110 and 2006AA01Z436)the Natural Science Foundation of Shanghai of China(No. 05ZR14083)the Pudong New Area Technology Innovation Public Service Platform of China(No. PDPT2005-04)
文摘Modeling of network traffic is a fundamental building block of computer science. Measurements of network traffic demonstrate that self-similarity is one of the basic properties of the network traffic possess at large time-scale. This paper investigates the change of non-stationary self-similarity of network traffic over time,and proposes a method of combining the discrete wavelet transform (DWT) and Schwarz information criterion (SIC) to detect change points of self-similarity in network traffic. The traffic is segmented into pieces around changing points with homogenous characteristics for the Hurst parameter,named local Hurst parameter,and then each piece of network traffic is modeled using fractional Gaussian noise (FGN) model with the local Hurst parameter. The presented experimental performance on data set from the Internet Traffic Archive (ITA) demonstrates that the method is more accurate in describing the non-stationary self-similarity of network traffic.
文摘对液力耦合器(liquid coupling device,LCD)的两阶段退化过程进行可靠性建模及评估。液力耦合器退化过程中呈现出明显的两阶段特性,即LCD在退化过程中存在某一时刻,在该时刻前后,LCD分别服从不同的退化过程。传统做法只基于第二阶段进行可靠性分析,丢失了第一阶段的信息。鉴于此,建立两阶段维纳退化过程模型,并推导出该模型下的可靠度函数;同时,基于Schwarz信息准则(Schwarz information criterion,SIC),获得模型变点的估计。最后,运用所建模型对LCD实例进行可靠性评估,并与传统做法进行对比分析。结果表明该模型对LCD退化过程拟合效果良好,且给出更为可信的评估结果。
文摘针对多阶段退化产品,基于性能退化响应数据的回归统计拟合,建立变速率复杂退化可靠性评估模型。首先利用分段回归拟合对性能退化数据进行变点回归建模,推导变点分布的极大似然函数;其次鉴于模型的复杂性,变点估计的最大似然估计无法得到显式解,采用分层Bayes进行建模;然后结合MCMC(Markov chain Monte Carlo)算法中的Gibbs采样对模型进行参数诊断,依据Schwarz信息准则(Schwarz information criterion,SIC)构建经验似然比对模型的变点进行估计与检验;最后根据产品失效退化的定义,推导该失效模式下产品的可靠度函数。以手机运行的游戏性能数据建模分析,演绎说明了多阶段退化模型适应性强、可行性高的特点。与传统两阶段建模相比,多阶段退化建模考虑了各个阶段的退化信息,提高了数据利用率和产品的可靠性评估可信度。
