Stochastic geometry of network of randomly distributed moving vehicles on a highway

Vehicular ad hoc networks (VANETs) have become an extensively studied topic in contemporary research. One of the fundamental problems that has arisen in such research is understanding the network statistical properties, such as the cluster number distribution and the cluster size distribution. In this paper, we analyze these characteristics in the case in which vehicles are located on a straight road. Assuming the Rayleigh fading model and a probabilistic model of intervehicle distance, we derive probabilistic distributions of the aforementioned connectivity characteristics, as well as distributions of the biggest cluster and the number of disconnected vehicles. All of the results are confirmed by simulations carried out for the realistic values of parameters.     

Supply chain network design under the risk of uncertain disruptions

Facility disruptions in the supply chain often lead to catastrophic consequences, although they occur rarely. The low frequency and non-repeatability of disruptive events also make it impossible to estimate the disruption probability accurately. Therefore, we construct an uncertain programming model to design the three-echelon supply chain network with the disruption risk, in which disruptions are considered as uncertain events. Under the constraint of satisfying customer demands, the model optimises the selection of retailers with uncertain disruptions and the assignment of customers and retailers, in order to minimise the expected total cost of network design. In addition, we simplify the proposed model by analysing its properties and further linearise the simplified model. A Lagrangian relaxation algorithm for the linearised model and a genetic algorithm for the simplified model are developed to solve medium-scale problems and large-scale problems, respectively. Finally, we illustrate the effectiveness of proposed models and algorithms through several numerical examples.     

The thermodynamic characteristics of a Stirling/pulse tube hybrid cryocooler

A Stirling/pulse tube hybrid cryocooler (SPC), comprised of a Stirling cryocooler as the first stage and a pulse tube cryocooler as the second stage, features the ability of shifting cooling capacity between stages by adjusting the movement of the displacer in the first stage. Such an ability allows an SPC to accommodate itself to time-varying heat loads at different temperatures, which makes it a competitive candidate in space applications. However, due to the gas coupling, there exists a significant mutual effect between stages which endows an SPC with special thermodynamic characteristics and has a significant effect on the SPC’s capability of shifting cooling capacity between stages. With the phasor analysis and the thermodynamic analysis, this paper establishes an idealized model of an SPC. The model is then used to study the effect of the second stage on the first stage and reveal the condition that an SPC is able to shift cooling capacity between stages. Also, the model is compared with a Sage numerical model and the two models are consistent on the overall trend. Though it is unable to reflect reality precisely, the idealized model can interpret the mechanism and highlight some of the essential nature of an SPC, which will eventually benefit the appropriate design of an SPC.     

基于数值微分核脉冲信号数字处理方法

核反应堆核测量系统测量探测器输出的核脉冲信号，该信号后沿拖尾很长，在计数率较高时容易产生信号堆积和基线漂移等问题，导致源区计数率测量上限仅能达到105 Hz左右。文中基于数值微分方法，采用数字处理技术，在时域上分析了核脉冲信号经过前置放大、信号成形、低通滤波和脉冲甄别后的输出，并利用探测器实测信号进行了仿真。仿真结果表明，基于数值微分的数字处理方法可以实现相邻0.4 μs脉冲信号的识别和测量，将源区测量计数率上限提升到2×106 Hz以上。      

Deploy Efficiency Driven k-Barrier Construction Scheme Based on Target Circle in Directional Sensor Network

Journal of Computer Science and Technology - With the increasing demand for security, building strong barrier coverage in directional sensor networks is important for effectively detecting...     

改进Elman神经网络在短期热负荷预测中的应用

热电厂的短期热负荷预测在城市集中供暖中起着至关重要的作用,直接影响热电厂的经济效益和热能利用率。电厂的短期热负荷一般采用神经网络预测模型进行预测,而BP神经网络应用最为广泛。Elman神经网络算法在BP神经网络基础上加入了承接层,作为一步延时算子,实现记忆能力,使系统具备适应时变能力,增强系统全局稳定性。但Elman神经网络算法模型的构造依然需要大量样本的支撑,而且输入层的变量多,导致预测时间依然很长,收敛速度慢。该文在Elman神经网络预测前,进行了相关系数预处理和对样本中异常值的平均化预处理,通过数据归一化运算,使Elman神经网络输入层变量大幅减少。仿真实验表明,改进的Elman神经网络算法使预测模型快速寻优,减少预测时间的同时明显提高预测精度。     

人工监督神经网络在煤层裂缝发育区识别中的应用

针对在地震属性应用中难以精细识别煤层裂缝发育区边界的难题,开展了基于人工监督神经网络技术的煤层裂缝发育区应用研究。从地震资料中获取倾角导向体以提取高质量的地震属性,以多属性为指导进行人工拾取样点,并基于多层感知器进行神经网络机器训练学习,建立裂缝的最优属性集,拓展整体数据后获得裂缝概率体,从而识别划分出煤层裂缝发育区。该技术在山西阳泉新元矿区进行了应用,煤层裂缝发育区的识别结果明显优于属性直接识别,但勘探区内暂无钻井资料,预测效果还有待进一步验证。     

基于双层修正无迹卡尔曼的水下节点定位算法* .txt

摘要：针对水下无线传感网络中运动节点定位精度低的问题，提出了一种新的基于双层修正无迹卡尔曼的水下节点定位算法(DLMUKF)。该算法利用下层无迹卡尔曼滤波算法对节点状态进行预测，根据各信标节点的测距传播时延对预测的节点状态进行修正。运用上层无迹卡尔曼滤波算法对修正后的状态进行新的预测与修正。仿真实验中，DLMUKF算法的平均定位误差约为传统多边定位算法的15%，约为基于无迹卡尔曼滤波（UKF）定位算法的16%，受节点运动时间与速度的影响最小。通过实验证明DLMUKF算法能更充分利用实际距离值，可以有效减小运动节点的定位误差。 .txt     

基于循环一致性对抗网络的室内火焰图像场景迁移

基于深度学习的视频火灾探测模型的训练依赖于大量的正负样本数据,即火灾视频和带有干扰的场景视频。由于很多室内场合禁止点火,导致该场景下的火灾视频样本不足。本文基于生成对抗网络,将其他相似场景下录制的火焰迁移到指定场景,以此增广限制性场合下的火灾视频数据。文中提出将火焰内核预先植入场景使之具备完整的内容信息,再通过添加烟雾和地面反射等风格信息,完成场景与火焰的融合。该方法克服了现有多模态图像转换方法在图像转换过程中因丢失信息而造成的背景失真问题。同时为减少数据采集工作量,采用循环一致性生成对抗网络以解除训练图像必须严格匹配的限制。实验表明,与现有多模态图像转换相比,本文方法可以保证场景中火焰形态的多样性,迁移后的场景具有较高的视觉真实性,所得结果的FID与LPIPS值最小,分别为119.6和0.134 2。