样本属性重要度的支持向量机方法

支持向量机(SVM)是一种新的机器学习方法,已经广泛应用于模式识别和函数估计等问题中.针对现有的加权支持向量机(WSVM)和模糊支持向量机(FSVM)只考虑样本重要性而没有考虑属性重要性对分类结果的影响的缺陷,提出了基于样本属性重要度的支持向量机方法,该方法首先利用信息论中的信息增益技术计算各个样本特征属性对分类属性的重要度,然后对所有样本的同一特征属性的值分别用对应的属性重要度进行加权,最后所得数据集用于训练和测试SVM.数值实验的结果表明,该方法提高了分类器的分类精度.     

Water Formation in Non-Hydrolytic Sol–Gel Routes: Selective Synthesis of Tetragonal and Monoclinic Mesoporous Zirconia as a Case Study

Several non-hydrolytic sol–gel syntheses involving different precursors, oxygen donors, and conditions have been screened aiming to selectively produce mesoporous t-ZrO₂ or m-ZrO₂ with significant specific surface areas. The in situ water formation was systematically investigated by Karl Fisher titration of the syneresis liquids. XRD and nitrogen physisorption were employed to characterize the structure and texture of the ZrO₂ samples. Significant amounts of water were found in several cases, notably in the reactions of Zr(OnPr)₄ with ketones (acetone, 2-pentanone, acetophenone), and of ZrCl₄ with alcohols (benzyl alcohol, ethanol) or acetone. Conversely, the reactions of Zr(OnPr)₄ with acetic anhydride or benzyl alcohol at moderate temperature (200 °C) and of ZrCl₄ with diisopropyl ether appear strictly non-hydrolytic. Although reaction time and reaction temperature were also important parameters, the presence of water played a crucial role on the structure of the final zirconia: t-ZrO₂ is favored in strictly non-hydrolytic routes, while m-ZrO₂ is favored in the presence of significant amounts of water. ¹H and ¹³C NMR analysis of the syneresis liquids allowed us to identify the main reactions responsible for the formation of water and of the oxide network. The morphology of the most interesting ZrO₂ samples was further investigated by electron microscopy (SEM, TEM).     

纯电动汽车磷酸铁锂电池组的建模及优化

鉴于传统神经网络和支持向量机机理复杂、计算量大的缺陷,很难实时跟踪磷酸铁锂电池组复杂快速的内部反应,影响电池荷电状态的估算精度,提出应用一种简单、有效的极限学习机对一额定容量为100Ah、额定电压为72V的纯电动汽车磷酸铁锂电池组建模,并分别与BP神经网络、RBF神经网络、支持向量机进行对比。随后,以学习时间和泛化性能为优化目标,应用粒子群方法寻找最佳隐层节点个数。结果表明,基于极限学习机的磷酸铁锂电池组模型的学习时间、泛化性能优于BP神经网络、RBF神经网络、支持向量机;隐层节点优化后,模型的学习时间和泛化性能达到最优。     

基于PSO-SVR算法的氧乐果合成过程建模研究

针对氧乐果合成过程中温度控制具有参数时变、时滞后、非线性的特点,提出了一种基于改进粒子群算法的支持向量回归的建模方法。对于支持向量回归模型,3个参数(ε,C,γ)的选取很大程度上决定了其拟合的精度和泛化能力的好坏,采用改进的粒子群算法对参数(ε,C,γ)进行同时寻优,建立了改进的氧乐果合成过程PSO-SVR回归模型,该模型具有很好的学习能力和推广能力。实验结果表明,模型较好地体现了系统的动态特性,可用于氧乐果合成过程的模型预估控制,提高系统的控制品质。     

装备使用与保障研究数据的生成方法研究

装备使用与保障数据是开展装备科学研究的重要依据,而由于管理不善、保密、样本过少等原因常常导致相关数据缺少应有的积累,无法满足实际研究的需要。为了克服这一困难,文章提出通过算法生成数据予以解决。首先根据数据的应用特性,将装备使用与保障数据看作简单数据、规律已知数据和规律未知数据三类,分别制定生成策略。对于简单数据和规律已知数据,通过已有的商用软件或简单的自定义方法生成;对于规律未知数据,采用模式注入或特性继承的方法生成。在某大型装备管理综合信息系统的设计开发中,对这些生成方法进行了工程实践,结果表明了方法的正确性和可行性。     

粘土担载的钼基催化剂的制备、结构及CO加氢合成醇性能

采用溶胶凝胶法与等体积浸渍相结合制备了一系列以粘土为载体的K-Co-Mo催化剂. 采用XRD、N₂等温吸脱附、H_{26+的还原,但对Mo⁴⁺和Co²⁺的还原没有明显的影响. 催化剂经还原后,在其表面生成了一种更低价态的Mo^δ+(1<δ<4)物种,被认为是合成醇的活性中心. 与非负载催化剂相比,粘土担载的K-Co-Mo具有更高的合成醇性能. 负载型催化剂具有较高的活性物种分散度,并且其介孔结构在一定程度上延长了合成醇反应中间体的滞留时间,从而促进了低碳醇的生成. 经773 K还原的催化剂具有较高的活性,其原因可为催化剂表面具有较高含量的Mo^δ+物种.}     

Counting of Oxygen Defects versus Metal Surface Sites in Methanol Synthesis Catalysts by Different Probe Molecules

Different surface sites of solid catalysts are usually quantified by dedicated chemisorption techniques from the adsorption capacity of probe molecules, assuming they specifically react with unique sites. In case of methanol synthesis catalysts, the Cu surface area is one of the crucial parameters in catalyst design and was for over 25 years commonly determined using diluted N₂O. To disentangle the influence of the catalyst components, different model catalysts were prepared and characterized using N₂O, temperature programmed desorption of H₂, and kinetic experiments. The presence of ZnO dramatically influences the N₂O measurements. This effect can be explained by the presence of oxygen defect sites that are generated at the Cu‐ZnO interface and can be used to easily quantify the intensity of Cu‐Zn interaction. N₂O in fact probes the Cu surface plus the oxygen vacancies, whereas the exposed Cu surface area can be accurately determined by H₂.     

Highly Stable,Water‐Dispersible Metal‐Nanoparticle‐Decorated Polymer Nanocapsules and Their Catalytic Applications

A facile synthesis of highly stable, water‐dispersible metal‐nanoparticle‐decorated polymer nanocapsules (M@CB‐PNs: M=Pd, Au, and Pt) was achieved by a simple two‐step process employing a polymer nanocapsule (CB‐PN) made of cucurbit[6]uril (CB[6]) and metal salts. The CB‐PN serves as a versatile platform where various metal nanoparticles with a controlled size can be introduced on the surface and stabilized to prepare new water‐dispersible nanostructures useful for many applications. The Pd nanoparticles on CB‐PN exhibit high stability and dispersibility in water as well as excellent catalytic activity and recyclability in carbon–carbon and carbon–nitrogen bond‐forming reactions in aqueous medium suggesting potential applications as a green catalyst.     

Preparation and characterization of Ti0.7Sn0.3O2 as catalyst support for oxygen reduction reaction

Sn-doped TiO_2 nanoparticles with high surface area of 125.7 m~2·g~(-1) are synthesized via a simple one-step hydrothermai method and explored as the cathode catalyst support for proton exchange membrane fuel cells.The synthesized support materials are studied by X-ray diffraction analysis,energy dispersive X-ray spectroscopy and transmission electron microscopy.It is found that the conductivity has been greatly improved by the addition of 30 mol%Sn and Pt nanoparticles are well dispersed on Ti_(0.7)Sn_(0.3)O_2 support with an average size of 2.44 run.Electrochemical studies show that the Ti_(0.7)Sn_(0.3)O_2 nanoparticles have excellent electrochemical stability under a high potential compared to Vulcan XC-72.The as-synthesized Pt/Ti_(0.7)Sn_(0.3)O_2 exhibits high and stable electrocatalytic activity for the oxygen reduction reaction.The Pt/Ti_(0.7)Sn_(0.3)O_2 catalyst reserves most of its electrochemically active surface area(ECA),and its half wave potential difference is 11 mV,which is lower than that of Pt/XC-72(36 mV) under 10 h potential hold at 1.4 V vs.NHE.In addition,the ECA degradation of Pt/Ti_(0.7)Sn_(0.3)O_2is 1.9 times lower than commercial Pt/XC-72 under 500 potential cycles between 0.6 V and 1.2 V vs.NHE.Therefore,the as synthesized Pt/Ti_(0.7)Sn_(0.3)O_2 can be considered as a promising alternative cathode,catalyst for proton exchange membrane fuel cells.     

抗污染薄层复合聚酰胺膜的结构设计及改性策略

由活性层和支撑层组成的薄层复合(TFC)聚酰胺(PA)膜,是目前广泛应用于纳滤、反渗透、正渗透和压力延迟渗透过程中的高性能脱盐膜,具有水通量大和截盐率高等优异性能。然而,由于TFC-PA膜存在活性层疏水性强、支撑层孔径大等特点,致使TFC-PA膜在实际使用过程中极易受到膜污染,制约了TFC-PA膜的进一步推广和使用。本文讨论分析了TFC-PA膜的结构特点和表面性质,总结归纳了在不同膜过程中TFC-PA膜污染形成的原因及特点,详细论述了国内外抗污染TFC-PA膜的研究进展。本文重点介绍了活性层抗污染改性和支撑层抗污染改性方法,并对其抗污染机理以及存在的问题进行了阐述与分析,最后对抗污染TFC-PA膜的结构设计与表面改性策略进行了总结及展望。