基于神经网络的感应电动机特性辨识新方法

提出并论证了一种基于神经网络的感应电动机特性辨识新方法,只需测得电机两相电流数值便可以辨识出电动机转矩和转速,用改进的Levenberg-Marquardt算法对神经网络进行学习和训练,构建了适合电动机转矩转速观测的BP神经网络。由于RBF神经网络无论是在逼近能力、函数拟合和学习速度方面都优于BP网络,也利用RBF网络进行了辨识。该方法较已经提出的方法相比,需要的检测量少,辨识方法简单。仿真研究表明,RBF神经网络辨识效果优于BP神经网络。     

智慧高速公路设计案例分析

本文以某智慧高速的设计案例为依托,首先对国内外智慧高速发展和应用现状进行归纳总结,分析认为电动化、智能化的新型运载工具,信息化和功能专业化的基础设施,开放、共享、协同的管理与服务将是未来高速公路三大发展趋势。通过对高速公路使用者和管理者的需求针对性地精准分析,构建面向高速公路使用者和管理者的创新型应用服务。结合新技术的发展趋势,以服务需求为导向,建设"三网合一"智慧高速基础设施和智慧云控平台构成、数据流为驱动的、"开放、包容、创新"的智慧高速公路。基于案例分析的智慧高速规划设计全过程,对于国内智慧高速的整体设计和建设规划具有重要参考价值,探索开辟全新的高速公路智能化建设方向和思路。     

A latent class accelerated hazard model of social activity duration

Over the past decade, activity scheduling processes have gained increasing attention in the field of transportation research. However, still little is known about the scheduling of social activities even though these activities account for a large and growing portion of trips. This paper contributes to this knowledge. We analyze how the duration of social activities is influenced by social activity characteristics and characteristics of the relationship between the respondent and the contacted person(s). To that end, a latent class accelerated hazard model is estimated, based on social interaction diary data that was collected in the Netherlands in 2008. Chi-square tests and analyses of variance are used to test for significant relations between the latent classes and personal and household characteristics. Findings suggest that the social activity characteristics and the characteristics of the relationship between the socializing persons are highly significant in explaining social activity duration. This shows that social activities should not be considered as a homogenous set of activities and it underlines the importance of including the social context in travel-behavior models. Moreover, the results indicate that there is a substantial amount of latent heterogeneity across the population. Four latent classes are identified, showing different social activity durations, and different effects for both categories of explanatory variables. Latent class membership can be explained by household composition, socio-economic status (education, income and work hours), car ownership and the number of interactions in 2 days.     

The value of options for time charterparty extension: an artificial neural networks (ANN) approach

The most frequently associated options in the physical shipping market are options to extend the charter period on time charters and additional shipment options on contracts of affreightment. The value of freight options, in practice, is estimated mostly by referring to forward curves. An option on freight has different properties from its financial counterparts, and the straightforward adoption of theoretical models does not produce promising results. In this paper, extension options, which have the property of options on futures, were transformed into regular European options before the application of the Black-Scholes model (BSM). The efficient market hypothesis, which justifies the parity of the performance of a long-term charter to that of repetitive short-term charters, worked as the basis for the transformation. The option values determined by the BSM were compared with actual realized values. Additionally, the artificial neural networks (ANN) was employed to derive the option values. This study is meaningful as the first-time application of both the closed-form solution and the ANN to the valuation of physical freight options. The research results can contribute to the quality of chartering decisions. The results could also be used in quantifying credit risk, as extension options tend to be granted to charterers with more creditability.     

接触网检测车振动补偿研究

针对接触检测车实际工况,分析了检测车车体振动的动力学特性,建立了基于运动规律的数学模型,论述了训练样本得出检测车拉出值振动补偿函数的原理和方法。     

用人工神经网络预测沥青路面材料的回弹模量

将人工神经网络应用于沥青路面材料的回弹模量的预测 ,提出了一种有效的预测方法 ,并构造了预测回弹模量的神经网络模型 .预测结果表明 ,该模型具有较高的预测精度 ,为预测路面材料的回弹模量等力学指标提供了一种新的方法     

动态管理模式下路侧停车泊位占有率预测方法

城市停车已逐步实现信息化和动态化管理,本文对动态管理模式下大范围路侧泊位占有率预测方法进行研究.在收集美国旧金山492万条停车交易数据的基础上,利用可同时提取数据空间关联和时序趋势特征的卷积长短时记忆神经网络(Convolutional LSTM Network,ConvLSTM),分别构建考虑停车费率和时限动态变化的有政策模型,和没有动态管理信息输入的无政策模型.结果显示,有政策模型的训练效率和预测精度会显著提升.在政策平稳阶段,两种模型均能够有效预测泊位占有率;在政策发生变化时段,无政策模型的预测误差出现激增,但有政策模型的预测误差依然保持平稳,表明本文提出的方法能够很好地应对动态管理模式下停车需求的变化.     

兰渝铁路兰州至广元段精密工程控制测量技术体系及特点

兰渝铁路精密工程控制测量技术体系的建立贯穿了我国铁路精密工程控制测量标准从建立到逐步完善的全过程。为给类似铁路项目测量方案的设计与实施提供参考和借鉴,结合兰渝铁路兰州至广元段建立的精密工程控制网技术体系,从全线统一的"三网合一"的测量技术体系、基于CGCS2000的平面坐标基准、工程独立坐标系、符合工程实际的水准基点平差方案、地震对测量控制网造成影响的评估、长大隧道洞内CPⅡ控制网建网及其分段测量方法等方面进行研究和论述。研究结论及实践经验对于丰富和完善铁路精密工程控制测量标准具有一定意义。     

基于贝叶斯网络的声呐系统战场损伤评估及修复模型

为了解决战场环境下的声呐系统损伤评估及修复（SSBDAR）问题,本文研究了战场损伤评估及修复模型。针对舰艇声呐系统战场损伤评估的因果推理特性,建立了基于贝叶斯网络的战场损伤评估模型,提出了基于贝叶斯网络的评估流程,给出了基于贝叶斯网络的评估决断算法。该模型利用贝叶斯网络对各评估修复节点进行分析,能够解决损伤信息不确定性,而且充分利用舰艇声呐系统各节点的损伤信息,提高了模型评估及修复效率,是解决舰艇声呐系统战场损伤评估及修复问题的一个有效途径。     

基于LM算法的组合导航系统的故障诊断

LM (Levenberg-Marquardt)算法是梯度下降法与高斯-牛顿法的结合,在训练次数与精度上明显优于标准的BP算法.介绍了BP神经网络,采用LM算法对标准BP算法进行改进,并应用于船舶组合导航系统的故障诊断.仿真结果表明,LM算法可有效地提高学习速度,缩短训练时间,诊断效果好.