基于贝叶斯网络的计算机网络端到端服务故障诊断

研究了利用贝叶斯网络不确定推理技术实现端到端服务故障诊断的方法,详细描述了贝叶斯网络故障诊断模型的建立方法,设计了基于Pearl信念传播机制的故障诊断算法,并对其进行了改进,以提高诊断效果.最后,通过仿真验证了该方法的有效性,并提出了下一步的研究方向.     

基于ANN的高炮火力单元射击有利度评估

采用ANN(人工神经网络)方法对高炮火力单元射击有利度进行评估是一种新的尝试。在对各影响因素进行分析与量化处理的基础上,建立了基于ANN的高炮火力单元射击有利度评估网络模型,并提供了网络的解算方法。通过某型高炮对不同目标射击的已知有利度评估值对网络进行的训练,确定了网络的各权值,并使学习误差达到了预先规定的标准,实现了对该型高炮对空射击有利度的评估。     

空间信息栅格体系与服务聚合技术

空间信息应用中存在的数据海量、分布异构、处理复杂、计算密集等问题制约了空间数据和信息的广泛共享和应用。空间信息栅格(SIG)以Grid和WebServices技术为核心,共享并汇集地理上分布的海量空间信息资源,实现动态整合与集成。在构建开放性SIG七层应用体系的基础上,重点研究SIG服务聚合技术,基于Petri网和图论提出了一种新的服务聚合流程模型(服务/资源网),同时构建了可扩展的服务动态选择模型(SISM),实现服务聚合过程中的动态服务选择。结合国家地质调查业务应用,通过地下水资源评价应用实例对SIG的实用性和有效性进行了验证。     

振动矩形机翼非线性绕流数值研究

发展了可用于模拟包含运动边界的非定常流动的数值方法。该方法采用非结构动网格技术并构造了一种MUSCL类型有限体积格式,求解考虑了动网格效应的三维非定常Euler方程。非结构网格的变形运动由弹簧近似技术实现。用上述方法计算了作正弦振动的矩形机翼非线性绕流,计算结果与实验结果吻合,得到了流体运动与固体运动耦合的非线性流动特性。计算表明,该数值方法具有较好的精度和可靠性,完全可用于气动弹性或其他流固耦合问题的数值模拟。     

基于粗集的神经网络在目标类型识别中的研究

为克服传统的目标类型识别方法的不足,提出将粗集和神经网络紧耦合建立新的识别模型,即经过识别信息预处理、样本数据粗集方法简化、神经网络学习训练及待识信息网络识别等步骤,充分融合了粗集强大的规则提取能力和神经网络优良的分类能力。实验表明,该模型减少了识别的主观因素,简化了神经网络结构,提高了运算速度。     

基于HLA的军事信息系统网仿真体系架构研究

分布式仿真系统的体系架构对于整个系统的功能实现和运行性能起着决定性的影响,为能充分利用仿真系统测试和评估军事信息系统网的性能,该文首先介绍了军事信息系统网(Military Information System Network:MISN)的组成和特点,并在分析基于数据库的仿真系统和基于HLA(高层体系架构)单联邦仿真系统局限性的基础上,得到了基于HLA的军事信息系统网分布式仿真多联邦体系框架,实现了联邦间功能转换所需的桥接联邦.该体系框架较好地完成了多通信网系的分布式交互仿真.     

Short‐term capacity allocation problem with tool and setup constraints

We consider a short‐term capacity allocation problem with tool and setup constraints that arises in the context of operational planning in a semiconductor wafer fabrication facility. The problem is that of allocating the available capacity of parallel nonidentical machines to available work‐in‐process (WIP) inventory of operations. Each machine can process a subset of the operations and a tool setup is required on a machine to change processing from one operation to another. Both the number of tools available for an operation and the number of setups that can be performed on a machine during a specified time horizon are limited. We formulate this problem as a degree‐constrained network flow problem on a bipartite graph, show that the problem is NP‐hard, and propose constant factor approximation algorithms. We also develop constructive heuristics and a greedy randomized adaptive search procedure for the problem. Our computational experiments demonstrate that our solution procedures solve the problem efficiently, rendering the use of our algorithms in real environment feasible. © 2005 Wiley Periodicals, Inc. Naval Research Logistics, 2005     

Scheduling a production–distribution system to optimize the tradeoff between delivery tardiness and distribution cost

We consider a make‐to‐order production–distribution system with one supplier and one or more customers. A set of orders with due dates needs to be processed by the supplier and delivered to the customers upon completion. The supplier can process one order at a time without preemption. Each customer is at a distinct location and only orders from the same customer can be batched together for delivery. Each delivery shipment has a capacity limit and incurs a distribution cost. The problem is to find a joint schedule of order processing at the supplier and order delivery from the supplier to the customers that optimizes an objective function involving the maximum delivery tardiness and the total distribution cost. We first study the solvability of various cases of the problem by either providing an efficient algorithm or proving the intractability of the problem. We then develop a fast heuristic for the general problem. We show that the heuristic is asymptotically optimal as the number of orders goes to infinity. We also evaluate the performance of the heuristic computationally by using lower bounds obtained by a column generation approach. Our results indicate that the heuristic is capable of generating near optimal solutions quickly. Finally, we study the value of production–distribution integration by comparing our integrated approach with two sequential approaches where scheduling decisions for order processing are made first, followed by order delivery decisions, with no or only partial integration of the two decisions. We show that in many cases, the integrated approach performs significantly better than the sequential approaches. © 2005 Wiley Periodicals, Inc. Naval Research Logistics, 2005     

火炮动态后坐装置有限元分析与优化

活塞缸是火炮动态后坐装置中的关键部件,其结构强度设计是整个实验装置强度设计的核心。利用I-DEAS软件,在对其进行实体建模和有限元网格划分的基础上,计算了活塞缸整体的应力应变值,得到了活塞缸工作时的应力应变图,并对其结构中应力状况恶劣的部位进行了结构优化。     

基于NNLI技术的齿轮箱故障诊断

将NNLI技术应用到齿轮箱的故障诊断中,提出了基于NNLI的特征提取方法,并将该方法与神经网络结合起来,进一步提出基于NNLI特征提取的神经网络故障诊断方法,给出了两种不同的网络分类器,通过齿轮箱故障诊断实例验证了该方法的有效性。