城乡空置住房的识别、特征与成因——基于华北平原X县用电量等多源数据

对空置住房进行精准感知在国土空间规划中具有重要意义。基于华北X县2019年全县居民用电大数据构建了一种精准的空置识别和空置率计算模型。结果显示：该县出现过空置的住房占比高达33.81%,年空置率为21.64%;月空置率在一年中随月份小幅规律波动,波谷出现在2月和8月,波峰出现在5月和11月;村空置率具有空间自相关属性,县城为空置率高值集聚区;全年性空置为1931种空置节律模式中最主要的一种。将空置成因归纳为举家迁移和入城团聚等新流动模式促使全年空置和春节空置增多;扩展家庭内部住房过滤引发老旧住房空置;县城住房城镇化导致中心城区出现空置。应重视空置识别在国土空间规划中的基础作用,客观看待乡村住房空置现象,高度警惕县城的高空置率问题。     

火工品温湿度加速老化模型和算法研究

目的 研究对火工品适用的温湿度加速模型,以及加速系数计算程序,获取准确的温湿度加速系数外推公式.方法 通过分析国内外使用的各种双因素加速模型的适用性,确定火工品适用的温湿度加速模型,采用免费自由的开源统计分析软件——R软件,对通过加速试验获取的某点火头2种温湿度加速条件下的试验数据,进行单组数据拟合和多组数据拟合两种温湿度加速老化算法的开发,并且对2种算法、3种寿命分布下寿命外推的准确性进行对比分析.结果 确定了Peck模型适合用于火工品的温湿度加速模型,获取了某电点火头对应的温湿度加速系数和湿度项反应速率常数.结论 基于多组数据拟合的温湿度加速老化算法,比基于单组数据拟合的温湿度加速老化算法对试验数据的兼容性好.3种寿命分布中,Weibull分布的加速系数计算结果最保守.     

机器学习在材料服役性能预测中的应用

针对材料服役性能预测存在误差大、计算复杂、适用性差等问题,提出了基于数据挖掘的机器学习预测方法。首先阐述了机器学习的应用流程,并总结了常用模型原理及其在材料性能预测中的应用。然后采用多种机器学习模型对RPV钢的辐照性能进行预测,并通过Stacking集成方法提高了模型的预测精度。结果表明,机器学习可用于材料服役性能预测,具有较高的预测精度和可靠性。根据材料服役数据的不同特征选择合适的学习模型,同时进行模型融合和参数优化,可有效提高模型的预测精度及运算速度。     

环境因素数据质量审核方法研究

目的 解决大气环境试验站的气象数据报表存在数据缺失、数据超过边界、数据内部不一致、数据时间不一致等问题,采用数据质量审核技术找出缺失、可疑或错误的数据,对科研人员进行警示,使其能够追根溯源,分析存在问题的原因。方法 参照相关规程和标准,建立环境因素数据质量审核算法和审核流程,并使用计算机语言编程固化流程和算法,利用多线程算法对数据进行扫描,实现大批量气象数据的并行审核。结果 利用Python语言设计了环境因素质量控制软件,通过该软件的应用证明了该方法合理、有效,能够快速、准确地标注气象数据报表中存在质量问题的数据。利用软件统计了存在问题数据的数量和正确率。结论 通过加载数据测试,证明了环境因素数据质量审核方法和审核流程合理有效,利用设计的审核软件实现了气象数据报表快速审核,提高了审核效率,减少了科研人员的劳动强度,为进一步填补、修正和利用数据奠定了基础。     

列车运行监控装置主机的超高斯加速振动试验研究

目的解决LKJ主机的超高斯振动试验问题。方法基于LKJ主机的实测振动环境数据,提出一种超高斯振动数据归纳方法,并应用该方法归纳出LKJ主机x、y、z向的实测PSD,同时按5.66的加速因子构建出超高斯加速振动试验剖面,其中x、y、z向PSD的RMS分别为3.34、6.28、3.85 m/s^(2),峰度分别为6.48、6.58、6.81,频率范围均为2~350 Hz。最后,采用超高斯加速试验剖面分别对2个LKJ主机进行超高斯振动试验验证。结果试验共激发出7类故障模式,这与特定线路运行LKJ主机现场故障模式高度吻合。结论该试验方法验证了LKJ主机超高斯加速振动试验的有效性。     

人工增雨（雪）效果评估方法探索

人工增雨（雪）常用的评估方法为统计检验和物理检验,统计检验需要大量试验样本,且随机化试验在我国难以实现,大量样本无从获得。物理检验目前主要是雷达观测演变状况和飞机直接观测云内微物理参数,由于技术和能力有限,这些方法都难以实现。为了在人工影响天气作业后尽快得出评估结果供政府及环境保护等部门参考,利用雨滴谱记录数据对一次人工增雪作业后的效果进行了评估,发现这些数据在没有常规降雪记录情况下具有一定的参考价值。从评估结果看,人工增雪作业后,降水强度有明显增大的趋势。     

A Multi‐Scale Analysis of Single‐Family Residential Water Use in the Phoenix Metropolitan Area

Studies that evaluate determinants of residential water demand typically use data from a single spatial scale. Although household‐scale data are preferred, especially when econometric models are used, researchers may be limited to aggregate data. There is little, if any, empirical analysis to assess whether spatial scale may lead to ecological fallacy problems in residential water use research. Using linear mixed‐effects models, we compare the results for the relationship of single‐family water use with its determinants using data from the household and census tract scales in the city of Phoenix. Model results between the household and census tract scale are similar suggesting the ecological fallacy may not be significant. Common significant determinants on these two spatial scales include household size, household income, house age, pool size, irrigable lot size, precipitation, and temperature. We also use city/town scale data from the Phoenix metropolitan area to parameterize the linear mixed‐effects model. The difference in the parameter estimates of those common variables compared to the first two scales indicates there is spatial heterogeneity in the relationship between single‐family water use and its determinants among cities and towns. The negative relationship between single‐family house density and residential water use suggests that residential water consumption could be reduced through coordination of land use planning and water demand management.     

基于BP神经网络的天线识别

本文比较研究了BP神经网络中的几种常用算法,针对这些不同算法下的BP神经网络进行训练,并得出了各自网络的性能。在此基础上,针对经典BP算法和LM算法进行对比研究,找到LM算法的改进之处。此外,在实际的应用中表明,不仅不同的BP算法对网络的运算速度、泛化能力等有较大的影响,而且BP神经网络对隐含层神经元数目也很敏感。我们希望在BP神经网络的基础上,构建一种合适的天线模型,来应用于天线的分类识别,这将具有很大的现实意义。     

Effect of Monitoring Technique on Quality of Conservation Science

Monitoring free‐ranging animals in their natural habitat is a keystone of ecosystem conservation and increasingly important in the context of current rates of loss of biological diversity. Data collected from individuals of endangered species inform conservation policies. Conservation professionals assume that these data are reliable—that the animals from whom data are collected are representative of the species in their physiology, ecology, and behavior and of the populations from which they are drawn. In the last few decades, there has been an enthusiastic adoption of invasive techniques for gathering ecological and conservation data. Although these can provide impressive quantities of data, and apparent insights into animal ranges and distributions, there is increasing evidence that these techniques can result in animal welfare problems, through the wide‐ranging physiological effects of acute and chronic stress and through direct or indirect injuries or compromised movement. Much less commonly, however, do conservation scientists consider the issue of how these effects may alter the behavior of individuals to the extent that the data they collect could be unreliable. The emerging literature on the immediate and longer‐term effects of capture and handling indicate it can no longer be assumed that a wild animal's survival of the process implies the safety of the procedure, that the procedure is ethical, or the scientific validity of the resulting data. I argue that conservation professionals should routinely assess study populations for negative effects of their monitoring techniques and adopt noninvasive approaches for best outcomes not only for the animals, but also for conservation science. Efecto de la Técnica de Monitoreo en la Calidad de la Ciencia de la Conservación