基于串联灰色多元回归模型的大坝蓄水期变形性态分析

基于某双曲拱坝蓄水期坝顶水平位移的特点,在对其监测资料进行时空分析基础上,构建了蓄水期坝顶水平位移的多元线性回归模型和串联灰色多元线性回归模型,分析预测了该坝蓄水期变形性态。实例分析结果表明,串联灰色多元线性回归模型预测效果优于多元线性回归模型,更合理地反映大坝蓄水期水平位移的变化趋势。     

基于多元线性回归的地下水开采可靠度模型

针对各种不确定性因素直接影响地下水资源评价结果的可靠性问题,基于多元线性回归模型的统计分析,以地下水开采可靠度为核心,提出了基于多元线性回归模型的地下水开采可靠性分析方法,结合传统多元回归分析与现代工程可靠性分析实现了基于多元线性回归的可靠度分析,并以淮北市地下水开采为例进行了计算分析.结果表明,该方法简单、实用、可行.     

基于逐步回归分析的用电量预测

文中针对用电量预测的多元线性回归模型，提出了逐步回归分析方法。与传统的多元线性回归模型相比较，逐步回归分析方法无需建立全部变量的回归方程，而是在全部自变量中按对因变量的作用大小，边进行显著性检验，边入选或剔除变量，不重要变量始终不进入回归方程，最后形成重要变量的最优回归方程。实例计算分析表明，与多元线性回归模型预测方法相比较，逐步回归预测模型对用电量的预测是有效的。     

多元线性回归在经济技术产量预测中的应用

传统以技术为中心的油田产量预测方法考虑因素单一,不能直观反映原油价格、生产成本等经济因素和开发投资、产能建设等决策因素对原油产量的变化。利用多元线性回归定量分析方法的基本原理,在综合考虑生产时间、百万吨产能投资、新建产能、单位生产成本、开发投资、生产成本、油价等经济技术因素的基础上,优选出了对油田产量影响显著的生产时间、百万吨产能投资、新建产能、生产成本和油价5个因素,并建立了相应的多元线性回归产量预测模型。研究了多元线性回归方法在EXCEL办公软件中实现快速分析与回归预测的过程。通过某油田实际参数与产量关系预测与验证,新建立的多因素经济技术多元线性回归预测模型简单、实用、可操作性强,预测误差小于2.3%;同时,预测了该油田2011~2015年产量变化趋势,对油田中长期科学规划决策具有一定指导意义。     

大伙房水库水质预测模型研究

水质为水库运行管理的重要因素,水质变量包括溶解氧、总磷、叶绿素a及透明度等,是水库富营养化判定的重要指标。在线性回归法分析的基础上,优化环境因子参数,分别采用多元线性回归模型、径向基函数模型与自适应模糊神经网络推理模型对辽宁省大伙房水库的水质进行预测,并通过平均绝对误差、均方根误差及相关系数判定水质模型的预测效果。结果表明,自适应模糊神经网络推理模型预测效果明显优于多元线性回归模型和径向基神经网络模型,因此自适应模糊神经网络推理模型更适合于大伙房水库的水质预测。     

不同机器学习模型对混凝土坝位移预测性能的影响

针对机器学习模型的合理选择问题,基于Gauss、Linear和Sigmoid核函数,对某拱坝径向位移利用支持向量机、关联向量机、极限学习机和传统多元线性回归法分别建立预测模型,并对比分析拟合均方差、复相关系数、最大绝对误差和预测置信带宽等模型性能评价指标.结果 表明,机器学习模型的性能整体优于多元线性回归模型,但受核函...     

垃圾焚烧烟气中二噁英类浓度的支持向量回归预测

文章利用垃圾焚烧烟气的在线监测数据,探索建立二噁英类浓度与烟气中其他污染物浓度、工况参数的多元回归预测模型,从而实现对烟气中二噁英类浓度的间接实时监测,但受限于二噁英类浓度的监测成本高,可供回归预测训练的样本少的问题,线性回归预测模型的泛化能力较弱、稳定性较差。使用非线性的支持向量回归方法建模,有助于解决这一问题。文章使用3种核函数构建支持向量回归预测模型,应用华南地区某垃圾焚烧厂的10组监测数据作为训练集和测试集,比较了支持向量回归预测和多元线性回归预测的相对误差。研究结果表明,训练集为8组数据时,支持向量回归预测的相对误差明显小于多元线性回归,尤其是模型使用1阶多项式核函数和径向基核函数时的最大百分比误差较小,泛化能力较强。     

小范围区域内初始地应力场的多元线性回归

针对采用多元线性回归法分析初始地应力场进行工况组合时仅有选取足够大的计算范围方能避免计算与实测的结果偏差过大,利用多元线性回归拟合精度高的优点,在较小区域范围内通过合成初始应力场并施加使模型内部应力保持平衡的边界条件,采用ABAQUS软件进行地应力平衡分析计算,结果表明,该方法基本消除了边界条件对应力场的影响,可获得拟合精度高、直接应用于工程分析的初始应力场,可供借鉴.     

开窗自然通风热损失预测评估方法研究

根据某办公建筑开窗热损失实测结果,运用多元线性回归、BP神经网络、径向基函数神经网络(RBF)完成其热损失值预测,得到回归模型调整R~2为0.517,拟合度相对较低;BP神经网络预测结果最大误差为1.25×10~5,最小为0.1×10~5,较出现局部最低值,但不能满足整体最低值;RBF与实际值几乎相同,残差值波动小于BP神经网络,确定径向基函数神经网络预测效果最佳,BP神经网络次之,多元线性回归只适用于粗略预测。     

利用回归分析推算生料中SiO2、Al2O3的含量指导水泥生产

利用Excel程序中的回归分析功能,采用多元线性回归方法,由生料的发热量、CaO和Fe2O3的含量快速推算出生料中SiO2、Al2O3的含量,从而计算生料的率值,指导水泥生产配比的调整.     

Effect of pseudopotential choice on the calculated electron and phonon band structures of palladium hydride and its vacancy defect phases

Density functional theory is increasingly used to predict and understand the properties of hydrogen storage materials. Many such calculations have been performed for various real and hypothetical palladium hydrides, yet despite excellent agreement on electron band structures, significant disparities persist in relation to phonon band structures and critical matters such as dynamic stability of alternative structures. Some disparities may arise because of differing computation approaches between researchers. Therefore in this work a systematic approach was followed to compare calculated electron and phonon band structures for four palladium hydrides: PdH and Pd₃VacH₄ (the superabundant vacancy phase) assuming that octahedral (oct) or tetrahedral (tet) lattice interstices are occupied by H, with six commonly used calculation schemes based on the local density approximation and the generalised gradient approximation, within the harmonic approximation. Of the twenty-four combinations tested, seven are new to the literature. Excellent agreement was found between the calculation schemes for the electron band structures of all four crystal structures. The position regarding phonons is much less satisfactory, however, and highlights the sensitivity of phonon properties to the calculated lattice constants. None of the calculation schemes could reproduce the measured phonon energy gap of PdH(oct) and it is necessary to include anharmonicity of the H potential to obtain realistic results. The calculated lattice constants of PdH(tet) were larger than any observed in experiments, although the structure is dynamically stable. All six calculation schemes predicted dynamic instability for Pd₃VacH₄(oct), although the calculated lattice constant agreed with the estimated zero-temperature experimental value. This structure requires new calculations accounting for anharmonicity. The calculated lattice constant for Pd₃VacH₄(tet) was larger than any experimental value, so this alternative, while dynamically stable, is certainly not observed.     

Liquid hydrogen releases show dense gas behavior

The number of maritime initiatives with hydrogen as alternative fuel is increasing. While most of the early projects aim at using compressed hydrogen the use of liquid hydrogen (LH2) is more practical and is expected to become more attractive for implementation on larger vessels due to more efficient storage, bunkering and handling of the fuel. In the industry there seems to be some confusion regarding the behavior of LH2 releases, in particular whether a release into air will behave like a dense gas or a buoyant gas. The understanding of this aspect is critical to optimize design with regard to safety. This article will explain the expected behavior of LH2 releases and discuss expected hazard distances from LH2 releases relative to gaseous hydrogen releases and LNG. Some other safety concerns of LH2, like indoor releases, releases from vent masts, potential BLEVEs and RPTs are also discussed. The article explains why a higher safety standard may be required when designing hydrogen fueled vessels than for existing LNG fueled vessels.     

带粒气流对透平叶片冲蚀特性的数值分析与试验研究

基于欧拉－拉格朗日解基本思想，数值模拟了带粒双级跨音速燃气透平三元粘性湍流场，在此基础上以计入颗粒湍流扩散的颗粒随机轨道模型和PSIC法实现了气固双向耦合，跟踪和分析了10万个分布直径和石英砂、催化剂颗粒的运动轨迹及颗粒滑移程度，并计算了叶片的质量冲蚀率。结果表明：涂层后叶片的质量冲蚀率较基体可减小一个数量级。另外，分别采用石英砂和催化剂颗粒对2Cr12Mo－5基体材料和MCrAlY涂层进行了冲蚀试验，通过比较发现：在大部会冲击角下，涂层能有效地减小叶片的冲蚀；在同样的试验条件下，大石英砂颗粒造成的叶片冲蚀强于小石英砂颗粒。     

德阳平原区潜水含水层水动力弥散试验研究

科学、便捷的获得污染物在地下水中运移的时空变化规律,不仅有利于地下水污染防治,还能为区域规划提供科学可靠的依据。以四川省德阳平原地区为例,以胭脂红为示踪剂,通过野外径向弥散试验,采用标准曲线法求解弥散系数,并通过Matlab编程计算实测浓度与标准曲线浓度之间的误差,得到纵向弥散系数为0.53m2/h,均方根误差为0.066 3。可见获得的弥散系数较为合理,弥散试验结果可信。     

Emission spread from mass and energy exchange in the atmospheric surface layer: Two-dimensional simulation

This article provides an original method for simulating the spread of hazardous emissions in the atmospheric surface layer in the event of emergencies The modeling process implies the development of a mathematical tool based on the partial differential equation (PDE) of turbulent diffusion. It is also based on a steady-state equation for determining the contaminant dispersion during an emergency with regard to the elapsed dispersion time, volumetric nature of the contamination source, wind speed. The model allows studying the changes in air composition that have occurred under various meteorological (wind speed and direction, temperature, fogs, precipitation, dust winds) and anthropogenous factors. An attempt was made to calculate the probable contamination area around the arbitrary system of sources with high contaminant concentration under unfavorable weather conditions. The original method for calculating the movement and dispersion of emissions allows forecasting the contaminant concentration level in the two-meter layer above the ground and the vertical concentration distribution with regard to the landscape, identified according to the meteorological data. Our model can be easily scaled on any territory; the entire line of complex landscape features was taken into account by introducing a model of weather statistics for the research region.     

Solvent effect in the synthesis of nanostructured Pt–Sn/CNT as electrocatalysts for the electrooxidation of ethanol

Pt and Pt–Sn nanoparticles were synthesized and supported onto carbon nanotubes (CNT), the electrocatalytic activity towards the ethanol oxidation reaction was analyzed. The effect of the solvent employed for the synthesis was evaluated. Metal nanoparticles synthesis was made using water (Pt–Sn/CNT-W) or ethanol (Pt–Sn/CNT-E) as a solvent. Pt–Sn/CNT-W material presented only Pt–Sn alloy nanoparticles homogeneously distributed on the carbon nanotubes support. Pt–Sn/CNT-E sample showed non well-dispersed nanoparticles forming agglomerates along the CNTs surface with predominantly Sn⁴⁺ superficial species (SnO₂) as show the XPS, FTIR and electrochemical results. These surface arrangements had important effects on the electrocatalytic properties. Pt–Sn/CNT-W shows higher ethanol electrooxidation activity than the Pt–Sn/CNT-E, which is attributed to: a) the double catalytic effect and the intrinsic electronic mechanism favored by the presence of Sn; b) the good particle dispersion of the bimetallic active phase on the CNT and; c) the absence of SnO₂ species.     

Liquid axial dispersion coefficient in a packed cooling tower

A dynamic method is used to predict the liquid axial dispersion coefficient in cooling towers. A KCl tracer was employed on the water inlet on the top of the tower. Input and response signals were measured. Cooling tower parameters were determined by predicting the response tracer signal and curve fitting technique in the time domain.     

Hydrogen gas dispersion studies for hydrogen fuel cell vessels I: Vent Mast releases

We report modeling results for hydrogen releases associated with deploying hydrogen fuel cell technology on vessels. This first paper (I) considers hydrogen releases through the vessel Vent Mast from 250-bar hydrogen gas storage tanks, the type of tanks being used for the first hydrogen vessels. A manifolded 10-tank hydrogen storage system, holding 278 kg of hydrogen, can be emptied in ～10 min for maintenance purposes, with a pressure reduction to half the original pressure (125 bar) realized in 2 min if a rapid pressure reduction is needed, for example in the event of a fire. The time profile for filling a tank is also of interest so as not to exceed the tank thermal limits. The calculations show that a manifolded 10-tank array can be filled with hydrogen to 250-bar pressure in ～2 h from a 350-bar hydrogen refueling trailer without exceeding the 85 °C temperature limit typical of Type IV hydrogen tanks.Computational fluid dynamic (CFD) modeling shows that when the hydrogen is released out of the 10-tank array and into the Vent Mast in a 5-mph wind blowing horizontally, the effect of the wind on the hydrogen dispersion strongly depends on the hydrogen exit speed. For high release speeds (～800–900 m/s), the hydrogen flow is strongly momentum-driven, and there is modest cross-wind influence. For low hydrogen exit speeds (～10 m/s), the hydrogen is readily entrained in the wind flow and blown sideways, with the downstream flammable envelope rising at a positive angle to the horizontal due to buoyancy. To capture the influence of a wind with a downward component (e.g., created by a downdraft near a building), a calculation of a low-velocity (8.6 m/s) hydrogen release was performed with a 5-mph wind pointed downward at a 45° angle. The results show that despite the buoyancy of hydrogen, the wind blows the hydrogen substantially downward for low hydrogen speeds exiting the Vent Mast.     

分析一维河流示踪试验数据的质量守恒法

应用质量守恒原理，通过适当的数学推导，将描述一维水团示踪试验的解析表达式转化为一直线方程在进行数据分析时，只要按文中提出的步骤进行数据转化，便可采用线性回归法或直线图解法计算直线常数，从而能够确定出河流纵向离散系数Ｄ．最后，通过算例说明了方法的具体应用步骤。     

CALPUFF模型在模拟核电站事故源项大气扩散中的应用

利用CALPUFF高斯烟团大气扩散模型,以放射性核素131I为例对大亚湾核电站一旦发生冷端双端断裂事故源项可能出现的大气扩散进行模拟,模拟放射性核素131I在典型气象条件下的空气时间积分浓度和地面沉积浓度分布,可以为核电站事故的环境影响评价方法提供参考。