螺线管磁场性能的优化

采用正反“单匝”线圈优化螺线管轴向磁场的均匀性和陡化端口磁场的下降沿。计算和实验研究表明,该方法明显改善磁场的均匀性和端口磁场的下降沿。     

地面雷达故障诊断策略研究

本文对现役地面雷达的故障诊断方式进行了综合比较分析,提出了诊断策略优化的基本标准和要求,并推荐一种新的诊断方式——两级结合诊断方式。     

时序电路故障可测性分析

对不可测故障进行测试产生是影响时序电路测试产生效率的一个重要因素。文中提出了一种基于简化可控性计算的识别时序电路中不可测故障的算法,运用该算法无须搜索便可识别出时序电路中相当一部分不可测故障。针对ISCAS89电路的实验结果也验证了其有效性。     

像管电子光学系统多目标优化的研究

将多重网格法用于静电像管电子光学场分布计算,采用改进的约束变尺度法进行静电聚焦像管的优化设计,较大地提高了电场分布的计算精度和像管设计效率,减少了计算时间和设计周期。研究的加权最小二乘法构造优化计算的多目标函数,普遍适用于调焦、定焦、变倍、选通和校畸变等各种类型静电像管的设计与计算。     

姿控发动机布局方式优化分析

在液体推进剂动力系统质量模型的基础上,针对采用双组元推进剂和挤压式输送系统的小推力空间飞行器姿控发动机,在控制总冲量和总冲量矩相同的情况下,对动力系统总质量最轻的姿控发动机最优布局方式进行了优化分析     

VRML三维模型文件的优化技术

目前在Web3D领域,VRML是主流的开发平台。但在建模过程中,通常使用第三方三维建模软件。这样极大地减轻了制作模型的难度,节约了时间;但由于建模平台本身的原因,转换后的VRML文件的体积比较大,不利于网络传输与浏览。因此,对模型的空间位置进行调整,尽可能地优化模型的视觉效果,缩减其文件体积,是建立虚拟环境非常重要的环节。主要针对这个问题提出了一些具体的优化技术与算法以及新的数据结构。     

Quantifying optimal mesh and ring design costs

During the last decade telecommunication operators have been deploying WDM (Wavelength Division Multiplexing) technology to satisfy the exponential growth in global communication. While facilitating the advanced information society of today, this has also led to a higher dependency on the networks, and furthermore the high capacity utilization of optical fibers means that a single link failure will influence many users and enterprises. For these reasons, protection of network connections has become a major competitive parameter for the operators. Currently, the most popular protection method is ring protection, due to its simplicity, requiring only basic management functionality and operating with local restoration control. While many optical rings have been deployed, little work has been published on exactly what the cost of ring networks are, compared to general mesh networks. In this article we perform a quantitative comparison between ring protection and mesh protection, using real world network data and realistic prices for network components. Extending classic LP flow models to take rings and node costs into account, and using a link‐path based mesh network LP model, we are able to perform a total cost comparison of the two architectures, and of manual ring network design. The results suggest that the price of mesh network components must be reduced significantly to be competitive with ring based networks, and also that manual network design does not necessarily lead to the most cost‐efficient designs. © 2004 Wiley Periodicals, Inc. Naval Research Logistics, 2005     

Discrete stochastic optimization using variants of the stochastic ruler method

We present two random search methods for solving discrete stochastic optimization problems. Both of these methods are variants of the stochastic ruler algorithm. They differ from our earlier modification of the stochastic ruler algorithm in that they use different approaches for estimating the optimal solution. Our new methods are guaranteed to converge almost surely to the set of global optimal solutions under mild conditions. We discuss under what conditions these new methods are expected to converge faster than the modified stochastic ruler algorithm. We also discuss how these methods can be used for solving discrete optimization problems when the values of the objective function are estimated using either transient or steady‐state simulation. Finally, we present numerical results that compare the performance of our new methods with that of the modified stochastic ruler algorithm when applied to solve buffer allocation problems. © 2005 Wiley Periodicals, Inc. Naval Research Logistics, 2005.     

汽油调合优化模型

为了提高炼油厂制订成品油调合方案的科学性,研究了汽油调合优化的非线性规划模型,给出了目标函数和约束条件的具体形式。根据模型特征,选用模拟退火算法对模型求解。最后,通过某炼油厂的一个应用实例验证了上述成品油调合优化模型的有效性。     

长向量处理器高效RNN推理方法

模型深度的不断增加和处理序列长度的不一致对循环神经网络在不同处理器上的性能优化提出巨大挑战。针对自主研制的长向量处理器FT-M7032,实现了一个高效的循环神经网络加速引擎。该引擎采用行优先矩阵向量乘算法和数据感知的多核并行方式,提高矩阵向量乘的计算效率;采用两级内核融合优化方法降低临时数据传输的开销;采用手写汇编优化多种算子,进一步挖掘长向量处理器的性能潜力。实验表明,长向量处理器循环神经网络推理引擎可获得较高性能,相较于多核ARM CPU以及Intel Golden CPU,类循环神经网络模型长短记忆网络可获得最高62.68倍和3.12倍的性能加速。