基于采样的非线性滤波方法

在系统阐述贝叶斯估计理论的基础上,分析和总结了扩展卡尔曼滤波、Sigma点卡尔曼滤波以及粒子滤波等方法的特点、使用条件以及局限性.介绍了扩展卡尔曼滤波方法并指出其缺陷;介绍了无轨迹卡尔曼滤波、中心差分卡尔曼滤波等确定性采样方法,并在加权统计线性回归意义下将其归结为Sigma点卡尔曼滤波方法;介绍了随机采样方法-粒子滤波方法,并指出其主要的研究方向.最后,对非线性滤波的发展趋势进行了展望.     

组合评价方法的装备维修性定性指标综合评价

针对在新型装甲装备设计阶段对其维修性进行综合评价的需求,结合装备设计阶段的其维修性指标验证与评价的特点,构建了基于设计阶段的装备维修性综合评价定性指标体系,通过分析线性加权与非线性加权两种综合评价方法的特点,提出了利用两种方法组合对装甲装备维修性进行综合评价的方法,并将该方法应用于实例验证,所得结果与物理样机的实验结果一致,证明了本方法的可行性和科学性,对基于设计阶段的新型装甲装备维修性综合评价有重要参考价值。     

基于贝叶斯网络的船舶动力装置战场损伤分析

针对船舶动力装置战场损伤特点,对其损伤分析进行了研究,提出了基于贝叶斯网络的动力装置战场损伤模型.以主机振动为例,分析了模型的建立、更新和完善等.分析结果表明,该模型和方法提高了船舶动力装置的战场损伤分析的速度和准确性.     

MIMO-OFDM最优训练序列设计

推导得到了宽带 MIMO-OFDM 信道模型。在空间相关情况下推导了非盲信道估计的 Bayesian CRB及 Bayesian MMSE。利用 Bayesian CRB 准则对发送的训练序列进行了功率分配,而后利用一种线性规划的方法对带有保护带宽 MIMO-OFDM 训练符号进行子载波功率分配,并利用发端酉矩阵进行最优相位设计。     

MIMO-OFDM系统Turbo接收机中的信道估计器

设计了一种新的MIMO-OFDM系统Turbo接收机中的信道估计器.考虑了信道估计误差对编码比特外信息计算的影响,利用软球形译码器的搜索列表和解码器反馈的先验信息对传统EM信道估计中的软信息近似处理进行了修正,获得了更为准确的软符号统计信息用于信道估计,提出了考虑软符号互相关性的Bayesian EM(BEM)信道估计算法.仿真结果表明,新算法较传统EM算法具有更低的误码率和更小的估计均方误差值.     

Bayesian enhanced decision making for deteriorating repairable systems with preventive maintenance

Since a system and its components usually deteriorate with age, preventive maintenance (PM) is often performed to restore or keep the function of a system in a good state. Furthermore, PM is capable of improving the health condition of the system and thus prolongs its effective age. There has been a vast amount of research to find optimal PM policies for deteriorating repairable systems. However, such decisions involve numerous uncertainties and the analyses are typically difficult to perform because of the scarcity of data. It is therefore important to make use of all information in an efficient way. In this article, a Bayesian decision model is developed to determine the optimal number of PM actions for systems which are maintained according to a periodic PM policy. A non‐homogeneous Poisson process with a power law failure intensity is used to describe the deteriorating behavior of the repairable system. It is assumed that the status of the system after a PM is somewhere between as good as new for a perfect repair and as good as old for a minimal repair, and for failures between two preventive maintenances, the system undergoes minimal repairs. Finally, a numerical example is given and the results of the proposed approach are discussed after performing sensitivity analysis. © 2007 Wiley Periodicals, Inc. Naval Research Logistics, 2008     

Network reliability assessment in a random environment

In this paper we consider networks that consist of components operating under a randomly changing common environment. Our work is motivated by power system networks that are subject to fluctuating weather conditions over time that affect the performance of the network. We develop a general setup for any network that is subject to such environment and present results for network reliability assessment under two repair scenarios. We also present Bayesian analysis of network failure data and illustrate how reliability predictions can be obtained for the network. © 2003 Wiley Periodicals, Inc. Naval Research Logistics 50: 574–591, 2003     

基于稀疏贝叶斯学习的扩展目标雷达关联成像*

雷达关联成像不依赖于雷达与目标的相对运动,是一种高分辨凝视成像方式。传统的关联成像方法未考虑复杂扩展目标的结构信息,在高分辨成像时的应用受到限制,为此提出一种自适应结构配对稀疏贝叶斯学习方法。该算法在稀疏贝叶斯学习的框架内针对扩展目标建立一种结构配对层次化高斯先验模型,然后采用变分贝叶斯期望-最大化算法交替进行目标重构和参数优化。该算法将某一信号分量的重构与周围信号分量联系起来,并能在迭代过程中自适应地调整表征各信号分量相关性的参数。实验结果表明,该方法针对扩展目标可以有效地进行高分辨成像。     

滑模控制的新型双幂次组合函数趋近律

提出一种基于双幂次组合函数趋近律的新型滑模控制方案。与现有的快速幂次或双幂次趋近律相比,具有更快的收敛速度,同时还保持了全局固定时间收敛特性,收敛时间上界与滑模初值无关。当系统存在有界扰动时,能够使滑模变量在有限时间内收敛到稳态误差界内,同时其稳态误差要小于现有方法的。仿真实验验证了该方法的有效性及理论分析的正确性。     

多核贝叶斯优化的模型决策树算法

构造模型决策树时超参数较多,参数组合复杂,利用网格搜索等调参方法将会消耗大量的时间,影响模型性能的提升。提出了一种多核贝叶斯优化的模型决策树算法,该算法为应对不同分类数据特性,采用三种高斯过程建模寻优,利用贝叶斯优化技术,选出最优的参数组合。实验结果表明,所提算法在参数寻优上要优于传统的模型决策树寻优方法,并且能够在迭代次数不多的情况下找到全局最优参数值,在一定程度上提升了算法的分类性能,节省了大量的调参时间。