Non-crossing Quantile Regression Neural Network as a Calibration Tool for Ensemble Weather Forecasts

Despite the maturity of ensemble numerical weather prediction(NWP),the resulting forecasts are still,more often than not,under-dispersed.As such,forecast calibration tools have become popular.Among those tools,quantile regression(QR)is highly competitive in terms of both flexibility and predictive performance.Nevertheless,a long-standing problem of QR is quantile crossing,which greatly limits the interpretability of QR-calibrated forecasts.On this point,this study proposes a non-crossing quantile regression neural network(NCQRNN),for calibrating ensemble NWP forecasts into a set of reliable quantile forecasts without crossing.The overarching design principle of NCQRNN is to add on top of the conventional QRNN structure another hidden layer,which imposes a non-decreasing mapping between the combined output from nodes of the last hidden layer to the nodes of the output layer,through a triangular weight matrix with positive entries.The empirical part of the work considers a solar irradiance case study,in which four years of ensemble irradiance forecasts at seven locations,issued by the European Centre for Medium-Range Weather Forecasts,are calibrated via NCQRNN,as well as via an eclectic mix of benchmarking models,ranging from the naïve climatology to the state-of-the-art deep-learning and other non-crossing models.Formal and stringent forecast verification suggests that the forecasts post-processed via NCQRNN attain the maximum sharpness subject to calibration,amongst all competitors.Furthermore,the proposed conception to resolve quantile crossing is remarkably simple yet general,and thus has broad applicability as it can be integrated with many shallow-and deep-learning-based neural networks.

Holographic alloy positioning design system and holographic network phase diagrams of Au-Cu system

Taking Au?Cu system as an example, three discoveries and two methods were presented. First, a new way for boosting sustainable progress of systematic metal materials science (SMMS) and alloy gene engineering (AGE) is to establish holographic alloy positioning design (HAPD) system, of which the base consists of measurement and calculation center, SMMS center, AGE center, HAPD information center and HAPD cybernation center; Second, the resonance activating-sychro alternating mechanism of atom movement may be divided into the located and oriented diffuse modes; Third, the equilibrium and subequilibrium holographic network phase diagrams are blueprints and operable platform for researchers to discover, design, manufacture and deploy advanced alloys, which are obtained respectively by the equilibrium lever numerical method and cross point numerical method of isothermal Gibbs energy curves. As clicking each network point, the holographic information of three structure levels for the designed alloy may be readily obtained: the phase constitution and fraction, phase arranging structure and properties of organization; the composition, alloy gene arranging structure and properties of each phase and the electronic structures and properties of alloy genes. It will create a new era for network designing advanced alloys.

Alloy gene Gibbs energy partition function and equilibrium holographic network phase diagrams of AuCu_3-type sublattice system

Taking AuCu3-type sublattice system as an example, three discoveries have been presented: First, the third barrier hindering the progress in metal materials science is that researchers have got used to recognizing experimental phenomena of alloy phase transitions during extremely slow variation in temperature by equilibrium thinking mode and then taking erroneous knowledge of experimental phenomena as selected information for establishing Gibbs energy function and so-called equilibrium phase diagram. Second, the equilibrium holographic network phase diagrams of AuCu3-type sublattice system may be used to describe systematic correlativity of the composition?temperature-dependent alloy gene arranging structures and complete thermodynamic properties, and to be a standard for studying experimental subequilibrium order-disorder transition. Third, the equilibrium transition of each alloy is a homogeneous single-phase rather than a heterogeneous two-phase, and there exists a single-phase boundary curve without two-phase region of the ordered and disordered phases; the composition and temperature of the top point on the phase-boundary curve are far away from the ones of the critical point of the AuCu3 compound.

Alloy gene Gibbs energy partition function and equilibrium holographic network phase diagrams of Au_3Cu-type sublattice system

Taking Au3Cu-type sublattice system as an example, three discoveries have been presented. First, the fourth barrier to hinder the progress of metal materials science is that today’s researchers do not understand that the Gibbs energy function of an alloy phase should be derived from Gibbs energy partition function constructed of alloy gene sequence and their Gibbs energy sequence. Second, the six rules for establishing alloy gene Gibbs energy partition function have been discovered, and it has been specially proved that the probabilities of structure units occupied at the Gibbs energy levels in the degeneracy factor for calculating configuration entropy should be degenerated as ones of component atoms occupied at the lattice points. Third, the main characteristics unexpected by today’s researchers are as follows. There exists a single-phase boundary curve without two-phase region coexisting by the ordered and disordered phases. The composition and temperature of the top point on the phase-boundary curve are far away from those of the critical point of the Au3Cu compound; At 0 K, the composition of the lowest point on the composition-dependent Gibbs energy curve is notably deviated from that of the Au3Cu compounds. The theoretical limit composition range of long range ordered Au3Cu-type alloys is determined by the first jumping order degree.

Study of macroscopic fundamental diagram on Shanghai urban expressway network in China

The macroscopic fundamental diagram（ MFD） is studied to obtain the aggregate behavior of traffic in cities. This paper investigates the existence and the characteristics of different types of daily MFD for the Shanghai urban expressway network. The existence of MFD in the Shanghai urban expressway network is proved based on two weeks＇ data.Moreover, the hysteresis phenomena is present in most days and the network exhibits different hysteresis loops under different traffic situations. The relationship between the hysteresis phenomena and the inhomogeneity of traffic distribution is verified. The MFDs in the years of 2009 and 2012 are compared. The hysteresis loop still exists in 2012, which further verifies the existence of the hysteresis phenomenon. The direct relationship between the length of the hysteresis loop（ ΔO） and the congestion is proved based on sufficient data. The width of the hysteresis loop, i. e., the drop in network flow（ ΔQ） has no relationship with the congestion, and it varies from day to day under different traffic situations.

Application of Artificial Neural Network in Indicator Diagram

Indicator diagram plays an important role in identifying the production state of oil wells. With an ability to reflect any non-linear mapping relationship, the artificial neural network (ANN) can be used in shape identification. This paper illuminates ANN realization in identifying fault kinds of indicator diagrams, including a back-propagation algorithm, characteristics of the indicator diagram and some examples. It is concluded that the buildup of a neural network and the abstract of indicator diagrams are important to successful application.

Effect of Chromium on CCT Diagrams of Novel Air-Cooled Bainite Steels Analyzed by Neural Network

The quantitative effects of chromium content on continuous cooling transformation （CCT） diagrams of novel air-cooled bainite steels were analyzed using artificial neural network models. The results showed that the chromium may retard the high and medium-temperature martensite transformation.

MOLTEN SALT PHASE DIAGRAMS CALCULATION USING ARTIFICIAL NEURAL NETWORK OR PATTERN RECOGNITION-BOND PARAMETERS PART 3.ESTIMATION OF LIQUIDUS TEMPERATURE AND EXPERT SYSTEM

１ＩＮＴＲＯＤＵＣＴＩＯＮＴｈｅｅｘｐｅｒｉｍｅｎｔａｌｄａｔａｏｎｔｈｅｌｉｑｕｉｄｕｓｌｉｎｅｓｏｒｓｕｒｆａｃｅｓｉｎｂｉｎａｒｙｏｒｔｅｒｎａｒｙｓｙｓｔｅｍｓｆｒｏｍｒｅｆｅｒｅｎｃｅｓａｒｅａｌｗａｙｓｆｉｎｉｔｅ．Ｓｏｍｅｔｉｍｅｓｔ...

Application of Windows Socket Technique to Communication Process of the Train Diagram Network System Based on Client/Server Structure

This paper is focused on the technique for de si gn and realization of the process communications about the computer-aided train diagram network system. The Windows Socket technique is adopted to program for the client and the server to create system applications and solve the problems o f data transfer and data sharing in the system.

A Study on Selecting the Shortest Routes by Voronoi Diagram in Route Networks of GIS

The problems of fast determining shortest paths through a polygonal subdivision planar with n vertices are considered in GIS. Distances are measured according to an Euclidean metric. A geographical information system (GIS) has a collection of nearest neighborhood operations and this collection serves as a useful toolbox for spatial analysis. These operations are undertaken through the Voronoi diagrams. This paper presents a novel algorithm that constructs a' shortest route set' with respect to a given source point and a target point by Voronoi diagrams. It will help to improve the efficiency of traditional algorithms, e. g., Djkstra algorithm, on selecting the shortest routes. Moreover, the novel algorithm can check the connectivity in a complex network between the source point and target one.

MOLTEN SALT PHASE DIAGRAMS CALCULATION USING ARTIFICIAL NEURAL NETWORK OR PATTERN RECOGNITION-BOND PARAMETERS

ＭＯＬＴＥＮＳＡＬＴＰＨＡＳＥＤＩＡＧＲＡＭＳＣＡＬＣＵＬＡＴＩＯＮＵＳＩＮＧＡＲＴＩＦＩＣＩＡＬＮＥＵＲＡＬＮＥＴＷＯＲＫＯＲＰＡＴＴＥＲＮＲＥＣＯＧＮＩＴＩＯＮＢＯＮＤＰＡＲＡＭＥＴＥＲＳ①Ｐａｒｔ１．Ｔｈｅｐｒｅｄｉｃｔｉｏｎｏｆｔｈｅｐｈａ...

Quantitative analysis of Ni effect on CCT diagrams of novel air-cooled bainite steels using artificial neural network models

The quantitative effect of Ni content on continuous cooling transformation （CCT） diagrams of novel air-cooled bainite steels was analyzed using artificial neural network models. The results showed that Ni may retard the high- and medium-temperature transformation and martensite transformation. The results conform to the materials science theories.

Quantification of the Hysteresis of Macroscopic Fundamental Diagrams and Its Relationship with the Congestion Heterogeneity and Performance of a Multimodal Network

Recent studies have observed hysteresis loops in the macroscopic fundamental diagram (MFD). In particular, for the same network density, higher network flows occur during congestion onset than during congestion offset. To evaluate management strategies using the MFD, investigating the relationship between the size of these loops and network performance is needed. The existing literature has mainly discussed correlating loop width (difference in density) and height (capacity drop) with congestion heterogeneity, but has failed to prove a relationship between the capacity drop and traffic conditions. Moreover, quantification of the MFD loop in complex multimodal networks has not been investigated. The objective of this paper covers these aspects. We simulated the Sioux Falls network with different mode-share ratios (car and bus users) based on a multi-agent simulation, MATSim. We investigated the relationships between MFD loop size and congestion heterogeneity (standard deviation of density) and network performance (average passenger travel time), and found that both were directly correlated with loop width, while weakly correlated with loop height. Moreover, we divided the MFD loop into two parts according to congestion onset and offset periods and found that the heights of the two parts had opposite effects. Accordingly, we show why the relationship between capacity drop and congestion heterogeneity is not found in the literature. We also found that network performance inversely affected the height of part of the loop while the height of its other part increased with an increase in congestion heterogeneity. These results help to evaluate network performance in the presence of MFD hysteresis, leading to elaborated management decisions.

Travel Patterns and Street Networks: A Novel Visualization Methodology of City-Level Traffic Flows and Network Usability

Street Networks, knitted in the urban fabric, facilitate spatial movement and control the flow of urbanization. The interrelation between a city’s spatial network and how the residents travel over it has always been of high interest to scholars. Over the years, multifaceted visualization methods have emerged to better express this travel trend from small to large scale. This study proposes a novel approach to 1) visualize city-wide travel patterns with respect to the street network orientation and 2) analyze the discrepancies between travel patterns and streets to evaluate network usability. The visualizations adopt histograms and rose diagrams to provide several insights into network-wide traffic flows. The visualization of four New York City (NYC) boroughs including Queens, Brooklyn, Bronx, and Staten Island was generated for the daily traffic and the average hourly flows in the morning and evening rush hours. Then the contrasts between built-in street network topology and travel orientation were drawn to show where people travel over the network, travel demand, and finally which segments experience high or light traffic, revealing the true picture of network usability. The findings of the study provide an insight into the novel and innovative approach that can help better understand the travel behavior lucidly and assist policymakers in decision making to maintain a balance between urban topology and travel demands. In addition, the study demonstrates how to further investigate city street networks and urbanization from different diverse dimensions.

GhostConv轻量级网络设计及故障诊断研究

提出一种GhostConv轻量级网络模型并将其用于故障诊断。GhostConv利用常规卷积生成一小部分特征图,然后在生成的特征图上进行多次特征提取来生成其余特征图,最大程度地节约了常规卷积中生成冗余特征图的成本,减少了模型参数,保证了模型的性能。采用连续小波变换对振动信号进行时频变换生成二维时频图,之后利用设计的GhostConv搭建轻量级网络模型进行故障诊断。采用凯斯西储大学轴承数据集进行验证,并与其他卷积结构网络模型进行参数量、计算量以及识别准确率的对比。实验结果表明,与其他模型相比,所使用的网络模型在参数量和计算量较少的条件下依旧有较高的识别精度,且具有较好的鲁棒性和泛化能力,具有一定的工程应用价值。

基于改进残差网络的抽油机故障诊断研究

针对抽油机故障诊断的传统图像识别方法识别率高但速率较慢,或训练速度适宜但识别率较低等问题,提出一种基于改进残差网络模型的示功图图像识别算法。改进策略包括替换模型第1层卷积核,由更小卷积核代替;改变残差模块排列顺序;将传统ResNet50(残差网络)模型的全连接层替换成径向基函数(RBF:Radial Basis Function)网络作为额外的分类器;采用数据增强方式对数据集进行扩充,并利用迁移学习在改进的ResNet50-RBF模型得到ImageNet上预训练好的权重参数后进行训练。实验结果表明,改进的模型在示功图识别中得到了98.86%的准确率,与其他网络相比,鲁棒性进一步加强,并且速率得到一定提升,为抽油机故障诊断提供了一定参考。

基于d3.js的miRNA调控网络图绘制模块的设计

对d3.js类库绘图功能进行研究,设计并实现了miRNA调控网络图绘制模块。模块能够将存储在数据库中的miRNA及其调控的靶基因等相关数据读取出来并形成JSON文件,结合d3.js可视化库,绘制有向的网络图及权重网络图。通过网络图能够在有限空间内更加清晰地展示miRNA及其靶基因之间的关系。

基于改进ResNet的示功图分类算法研究

示功图是反映抽油机井工作状态的重要图示,通过分析示功图的闭合曲线形状,可以得出抽油机井的具体工作状态,从而可以判断出抽油机井是否发生故障以及具体的故障类型。随着深度学习的发展,基于深度神经网络的示功图分类也逐渐应用到了抽油机井工况检测当中。该文提出了基于改进ResNet的示功图分类算法,通过优化残差结构和引入SE子结构等措施,提高了分类准确性和鲁棒性。改进的残差结构嵌入了SE子结构,对输入特征进行降维的同时也减小了参数的数量,在降低计算量的同时也添加了更多非线性因素,通过不断增加有效特征的权重,不断减小无效特征的权重,进而完成了特征重标定,不仅起到加速网络收敛的作用,也使模型更加轻量化,从而提高了模型的性能。相较于其它模型,改进的ResNet模型可以更好地适应示功图分类任务,分类效果更好。实验结果表明,基于改进ResNet的示功图分类算法在精确率、召回率和F1值上均优于其它示功图分类算法。该研究为抽油机井工况检测系统提供了更好的理论支持。

基于变分图自动编码器与K均值聚类的虚拟网络嵌入算法应用

将虚拟网络映射到物理网络是网络功能虚拟化中一项重要的任务.为了有效地分配虚拟网络请求,需要将虚拟网络嵌入到物理网络拓扑中.然而,由于虚拟网络的复杂性和物理网络的限制,这一任务变得非常具有挑战性.鉴于此,研究在现有虚拟网络嵌入算法(Virtual Network Embedding, VNE)模型基础上进行改进,融入了变分图自动编码器(Variational Graph Auto-Encoders, VGAE),提出了一种新型虚拟网络嵌入算法模型.通过编码器对虚拟网络的嵌入特征进行提取,随后利用K-means聚类算法对所得到的嵌入特征进行分类,最终得到合适的嵌入分配方法.实验结果表明,该新模型相较于其他同类型的嵌入算法性能表现最佳,稳定性最好,其平均嵌入请求接受率为60%,长期平均CPU资源利用率最高达97%.综上所述,研究提出的新型虚拟网络嵌入算法在资源利用率和嵌入质量方面表现出色,能够有效应对复杂的网络环境和大规模的虚拟网络请求.