目的探讨基于Markov Chain Monte Carlo(MCMC)模型的多重估算法在处理医院调查资料缺失数据中的应用。方法运用SAS9.2编写程序,在分析数据的分布类型和缺失机制的基础上,采用MCMC法对缺失数据进行多次填补和联合统计推断,分析多重估算...目的探讨基于Markov Chain Monte Carlo(MCMC)模型的多重估算法在处理医院调查资料缺失数据中的应用。方法运用SAS9.2编写程序,在分析数据的分布类型和缺失机制的基础上,采用MCMC法对缺失数据进行多次填补和联合统计推断,分析多重估算法的优势。结果数据服从多元正态分布与随机缺失,采用MCMC法填补10次所得的结果最佳。结论多重估算既可反映缺失数据的不确定性,又可充分利用现有资料的信息、提高统计效率、对模型的估计结果更加可信,是处理缺失数据的有效方法。展开更多
现有安全稳定控制系统(简称稳控系统)的可靠性评估方法本质上属于静态建模,由于未能体现系统内各装置老化和检修等动态过程,在一定程度上影响了评估结果的准确性。为此,文中提出一种基于马尔可夫链蒙特卡洛(Markov chain Monte Carlo,MC...现有安全稳定控制系统(简称稳控系统)的可靠性评估方法本质上属于静态建模,由于未能体现系统内各装置老化和检修等动态过程,在一定程度上影响了评估结果的准确性。为此,文中提出一种基于马尔可夫链蒙特卡洛(Markov chain Monte Carlo,MCMC)的稳控系统动态可靠性评估方法。首先针对失效过程,构建四状态非齐次马尔可夫模型来模拟装置老化过程,并给出各状态评判方法;其次针对修复过程,分析不同检修策略对装置状态转移的影响以体现状态检修的差异性;最后考虑稳控装置状态转移过程的时序或条件相关性,对稳控系统可靠性进行动态建模。以实际稳控系统为例,仿真对比不同检修策略下的可靠性,并对模型参数进行灵敏度分析。评估结果表明,该方法可以求解稳控系统的时变可用度,用于指导稳控装置现场合理检修。展开更多
Most of the maintenance optimization models in condition-based maintenance(CBM) consider the cost-optimal criterion, but few papers have dealt with availability maximization for maintenance applications. A novel optim...Most of the maintenance optimization models in condition-based maintenance(CBM) consider the cost-optimal criterion, but few papers have dealt with availability maximization for maintenance applications. A novel optimal Bayesian control approach is presented for maintenance decision making. The system deterioration evolves as a three-state continuous time hidden semi-Markov process. Considering the optimal maintenance policy, the multivariate Bayesian control scheme based on the hidden semi-Markov model(HSMM) is developed, the objective is to maximize the long-run expected average availability per unit time. The proposed approach can optimize the sampling interval and control limit jointly. A case study using Markov chain Monte Carlo(MCMC)simulation is provided and a comparison with the Bayesian control scheme based on hidden Markov model(HMM), the age-based replacement policy, Hotelling’s T2, multivariate exponentially weihted moving average(MEWMA) and multivariate cumulative sum(MCUSUM) control charts is given, which illustrates the effectiveness of the proposed method.展开更多
Although it is known that exact sampling algorithm is easy to construct and less sensitive to noise, the samples distri- bution of the algorithm deviates from the target states distribution due to the local dependent ...Although it is known that exact sampling algorithm is easy to construct and less sensitive to noise, the samples distri- bution of the algorithm deviates from the target states distribution due to the local dependent coupling problem. A new algorithm, named exact sampling with directional threshold （ES-DT） is intro- duced. The main advantage of the new algorithm, in comparison with the traditional exact sampling algorithm, is that it can control the sampling with a rejection strategy in Markov chain during the path growth, and closely approach the ideal distribution based on maintaining the target density. Simulation experiments show the effectiveness of the proposed algorithm.展开更多
Two variants of systematic resampling （S-RS） are proposed to increase the diversity of particles and thereby improve the performance of particle filtering when it is utilized for detection in Bell Laboratories Layer...Two variants of systematic resampling （S-RS） are proposed to increase the diversity of particles and thereby improve the performance of particle filtering when it is utilized for detection in Bell Laboratories Layered Space-Time （BLAST） systems. In the first variant, Markov chain Monte Carlo transition is integrated in the S-RS procedure to increase the diversity of particles with large importance weights. In the second one, all particles are first partitioned into two sets according to their importance weights, and then a double S-RS is introduced to increase the diversity of particles with small importance weights. Simulation results show that both variants can improve the bit error performance efficiently compared with the standard S-P^S with little increased complexity.展开更多
针对现有 Va R计算中主流方法的缺陷 ,创新性地提出了一种基于马尔科夫链蒙特卡洛(Markov Chain Monte Carlo,MCMC)模拟的 Va R计算方法 ,以克服传统 Monte Carlo模拟的高维、静态性缺陷 ,提高估算精度 .通过对美元国债的实证分析和计...针对现有 Va R计算中主流方法的缺陷 ,创新性地提出了一种基于马尔科夫链蒙特卡洛(Markov Chain Monte Carlo,MCMC)模拟的 Va R计算方法 ,以克服传统 Monte Carlo模拟的高维、静态性缺陷 ,提高估算精度 .通过对美元国债的实证分析和计算 ,验证了 MCMC方法的优越性 .展开更多
基于弹着点空间分布对目标毁伤效能的差异化影响,构建导弹命中目标不同重要区域的概率分布模型,实现对传统命中精度概念的扩展。针对导弹实打试验过程复杂、费用高、次数少的实际,采用贝叶斯方法融合多源信息,基于区域划分-分布确定-先...基于弹着点空间分布对目标毁伤效能的差异化影响,构建导弹命中目标不同重要区域的概率分布模型,实现对传统命中精度概念的扩展。针对导弹实打试验过程复杂、费用高、次数少的实际,采用贝叶斯方法融合多源信息,基于区域划分-分布确定-先验融合-后验求解的思路进行导弹命中精度估计。选取Dirichlet分布作为命中精度参数的先验分布,运用D-S(Dempster-Shafer)证据理论对先验信息进行融合处理,基于马尔可夫链蒙特卡罗(Markov chain Monte Carlo, MCMC)方法对精度参数的后验分布进行求解。示例表明,该方法能够细致描述导弹命中目标不同重要区域的概率,并科学融合多源命中精度先验信息,为导弹命中精度估计方法及测试方案优化提供理论借鉴。展开更多
文摘目的探讨基于Markov Chain Monte Carlo(MCMC)模型的多重估算法在处理医院调查资料缺失数据中的应用。方法运用SAS9.2编写程序,在分析数据的分布类型和缺失机制的基础上,采用MCMC法对缺失数据进行多次填补和联合统计推断,分析多重估算法的优势。结果数据服从多元正态分布与随机缺失,采用MCMC法填补10次所得的结果最佳。结论多重估算既可反映缺失数据的不确定性,又可充分利用现有资料的信息、提高统计效率、对模型的估计结果更加可信,是处理缺失数据的有效方法。
文摘现有安全稳定控制系统(简称稳控系统)的可靠性评估方法本质上属于静态建模,由于未能体现系统内各装置老化和检修等动态过程,在一定程度上影响了评估结果的准确性。为此,文中提出一种基于马尔可夫链蒙特卡洛(Markov chain Monte Carlo,MCMC)的稳控系统动态可靠性评估方法。首先针对失效过程,构建四状态非齐次马尔可夫模型来模拟装置老化过程,并给出各状态评判方法;其次针对修复过程,分析不同检修策略对装置状态转移的影响以体现状态检修的差异性;最后考虑稳控装置状态转移过程的时序或条件相关性,对稳控系统可靠性进行动态建模。以实际稳控系统为例,仿真对比不同检修策略下的可靠性,并对模型参数进行灵敏度分析。评估结果表明,该方法可以求解稳控系统的时变可用度,用于指导稳控装置现场合理检修。
基金supported by the National Natural Science Foundation of China(51705221)the China Scholarship Council(201606830028)+1 种基金the Fundamental Research Funds for the Central Universities(NS2015072)the Funding of Jiangsu Innovation Program for Graduate Education(KYLX15 0313)
文摘Most of the maintenance optimization models in condition-based maintenance(CBM) consider the cost-optimal criterion, but few papers have dealt with availability maximization for maintenance applications. A novel optimal Bayesian control approach is presented for maintenance decision making. The system deterioration evolves as a three-state continuous time hidden semi-Markov process. Considering the optimal maintenance policy, the multivariate Bayesian control scheme based on the hidden semi-Markov model(HSMM) is developed, the objective is to maximize the long-run expected average availability per unit time. The proposed approach can optimize the sampling interval and control limit jointly. A case study using Markov chain Monte Carlo(MCMC)simulation is provided and a comparison with the Bayesian control scheme based on hidden Markov model(HMM), the age-based replacement policy, Hotelling’s T2, multivariate exponentially weihted moving average(MEWMA) and multivariate cumulative sum(MCUSUM) control charts is given, which illustrates the effectiveness of the proposed method.
文摘Although it is known that exact sampling algorithm is easy to construct and less sensitive to noise, the samples distri- bution of the algorithm deviates from the target states distribution due to the local dependent coupling problem. A new algorithm, named exact sampling with directional threshold （ES-DT） is intro- duced. The main advantage of the new algorithm, in comparison with the traditional exact sampling algorithm, is that it can control the sampling with a rejection strategy in Markov chain during the path growth, and closely approach the ideal distribution based on maintaining the target density. Simulation experiments show the effectiveness of the proposed algorithm.
基金supported by the National Natural Science Foundation of China(6047209860502046U0635003).
文摘Two variants of systematic resampling （S-RS） are proposed to increase the diversity of particles and thereby improve the performance of particle filtering when it is utilized for detection in Bell Laboratories Layered Space-Time （BLAST） systems. In the first variant, Markov chain Monte Carlo transition is integrated in the S-RS procedure to increase the diversity of particles with large importance weights. In the second one, all particles are first partitioned into two sets according to their importance weights, and then a double S-RS is introduced to increase the diversity of particles with small importance weights. Simulation results show that both variants can improve the bit error performance efficiently compared with the standard S-P^S with little increased complexity.
文摘针对现有 Va R计算中主流方法的缺陷 ,创新性地提出了一种基于马尔科夫链蒙特卡洛(Markov Chain Monte Carlo,MCMC)模拟的 Va R计算方法 ,以克服传统 Monte Carlo模拟的高维、静态性缺陷 ,提高估算精度 .通过对美元国债的实证分析和计算 ,验证了 MCMC方法的优越性 .
文摘基于弹着点空间分布对目标毁伤效能的差异化影响,构建导弹命中目标不同重要区域的概率分布模型,实现对传统命中精度概念的扩展。针对导弹实打试验过程复杂、费用高、次数少的实际,采用贝叶斯方法融合多源信息,基于区域划分-分布确定-先验融合-后验求解的思路进行导弹命中精度估计。选取Dirichlet分布作为命中精度参数的先验分布,运用D-S(Dempster-Shafer)证据理论对先验信息进行融合处理,基于马尔可夫链蒙特卡罗(Markov chain Monte Carlo, MCMC)方法对精度参数的后验分布进行求解。示例表明,该方法能够细致描述导弹命中目标不同重要区域的概率,并科学融合多源命中精度先验信息,为导弹命中精度估计方法及测试方案优化提供理论借鉴。
