Predictive CFD modeling of whey protein denaturation in skim milk spray drying powder production

A new model of whey protein thermal inactivation has been combined with a CFD model developed for skim milk spray drying. Extensive evaporation and particle formation models were used to calculate particle moisture contents, temperatures and residence times. Calculated parameters were then used as input data for an experimentally developed quality model based on Williams-Landel-Ferry (WLF) equations for inactivation kinetics. The developed quality model was implemented into the CFD code and calculated in parallel to simulations of skim milk droplets evaporation based on the characteristic drying curve approach. The quality model and the simulation procedure were validated by comparison of protein activity levels obtained from the CFD with data obtained from differential scanning calorimetry (DSC) of milk powder samples collected during skim milk spray drying experiments. The simulations for different feed rates fit well with measurement results and show that the loss of whey protein activity is lower at higher feed rates, due to lower temperature fields in this case.     

壳聚糖-聚乳酸接枝共聚物的制备与表征

以L-乳酸和壳聚糖为原料,辛酸亚锡为催化剂,通过原位聚合法制备了含聚乳酸链段的壳聚糖接枝共聚物,讨论了原料投料比、催化剂浓度对接枝率的影响。通过红外光谱对共聚物化学结构进行分析,并通过热重分析研究其热失重行为。与壳聚糖相比,聚乳酸支链的引入有效削弱了壳聚糖分子的氢键作用,共聚物起始分解温度和热稳定性下降。     

UV nano embossing for polymer nano structures with non-transparent mold insert

In recent years, UV nano embossing (or imprinting) process has been widely used for mass replication of nano structures. Most of the UV embossing machines use a UV transparent mold insert (e.g. quartz or glass). However, when a master of nano structures of interest could be realized only in a UV non-transparent mold material, it would be desirable to have a UV embossing machine which could be operated with such a UV non-transparent mold insert. In this regard, we have designed and manufactured a new UV embossing system in such a way that UV non-transparent (e.g., metal or ceramic) mold inserts can also be used as the master for the mass replication of nano structures. For the new UV embossing system, we fabricate several metal mold inserts: a nickel electroformed mold insert having grating nano structures and two AAO (anodic aluminum oxide) mold inserts having dimple nano structures and high-aspect-ratio nanopores. Finally, corresponding nano structures are successfully replicated via the UV nano embossing machine developed in this study.     

基于注意力机制和集成学习的网页黑名单判别方法

搜索引擎作为互联网主要应用之一,能够根据用户需求从互联网资源中检索并返回有效信息。然而,得到的返回列表往往包含广告和失效网页等噪声信息,而这些信息会干扰用户的检索与查询。针对复杂的网页结构特征和丰富的语义信息,提出了一种基于注意力机制和集成学习的网页黑名单判别方法,并采用本方法构建了一种基于集成学习和注意力机制的卷积神经网络（EACNN）模型来过滤无用的网页。首先,根据网页上不同种类的HTML标签数据,构建多个基于注意力机制的卷积神经网络（CNN）基学习器;然后,采用基于网页结构特征的集成学习方法对不同基学习器的输出结果执行不同的权重计算,从而实现EACNN的构建;最后,将EACNN的输出结果作为网页内容分析结果,从而实现网页黑名单的判别。所提方法通过注意力机制来关注网页语义信息,并通过集成学习的方式引入网页结构特征。实验结果表明,与支持向量机（SVM）、K近邻（KNN）、CNN、长短期记忆（LSTM）网络、GRU、结合注意力机制的卷积神经网络（ACNN）等基线模型相比,所提模型在所构建的面向地理信息领域的判别数据集上具有最高的准确率（0.97）、召回率（0.95）和F₁分值（0.96）,验证了EACNN在网页黑名单判别工作中的优势。     

A simple GPU-based approach for 3D Voronoi diagram construction and visualization

In this paper we propose a simple GPU-based approach for discrete incremental approximation of 3D Voronoi diagram. By constructing region maps via GPU. Nearest sites, space clustering, and shortest distance query can be quickly answered by looking up the region map. In addition, we propose another representation of the 3D Voronoi diagram for visualization.     

计算机输入设备的频谱检测与分析

利用频谱分析仪对计算机输入设备工作时发射的辐射信号进行测试,获取输入设备辐射信号的频谱数据并进行相应的频谱特性分析。找出其泄漏的主要频段和特点,为输入设备的防护提供依据。     

基于外观的复合属性学习的细粒度识别

由于语义可理解性及共享性,视觉属性作为刻画对象的中间特征表示在众多领域得到了广泛应用。视觉属性学习中,大量的人工成本用于属性定义和标注,因此难以避免地引入了主观偏见,属性表示的类别判别性难以保证,尤其面临对判别性要求较高的细粒度识别任务时更为明显。复合属性符合人类认知规律以及对象复杂多模分布的事实,从刻画对象的分布入手,以较低廉的代价建立兼具一定描述能力及较好判别能力的特征表示,以应对细粒度识别任务对判别特征和判别模型的较高要求。在细粒度识别代表性公开数据集CUB上验证了所提方法的有效性。在细粒度识别任务中,复合属性表现出比人工定义的属性以及类别判别属性更优的性能。     

HFC网络MAC协议的三叉树冲突化解算法分析

对HFC网络中普遍采用的三叉树冲突化解算法进行分析，介绍树冲突化解算法的两种栈管理模式：FIFO(先进先出)模式和LIFO(后进先出)模式，并举例说明FIFO和LIFPO的冲突化解过程。     

Ionic Modulation of Interfacial Magnetism in Light Metal/Ferromagnetic Insulator Layered Nanostructures

Ferromagnetic insulator thin film nanostructures are becoming the key component of the state‐of‐the‐art spintronic devices, for instance, yttrium iron garnet (YIG) with low damping, high Curie temperature, and high resistivity is explored into many spin–orbit interactions related spintronic devices. Voltage modulation of YIG, with great practical/theoretical significance, thus can be widely applied in various YIG‐based spintronics effects. Nevertheless, to manipulate ferromagnetism of YIG through electric field (E‐field), instead of current, in an energy efficient manner is essentially challenging. Here, a YIG/Cu/Pt layered nanostructure with a weak spin–orbit coupling interaction is fabricated, and then the interfacial magnetism of the Cu and YIG is modified via ionic liquid gating method significantly. A record‐high E‐field‐induced ferromagnetic resonance field shift of 1400 Oe is achieved in YIG (17 nm)/Cu (5 nm)/Pt (3 nm)/ionic liquid/Au capacitor layered nanostructures with a small voltage bias of 4.5 V. The giant magnetoelectric tunability comes from voltage‐induced extra ferromagnetic ordering in Cu layer, confirmed by the first‐principle calculation. This E‐field modulation of interfacial magnetism between light metal and magnetic isolator may open a door toward compact, high‐performance, and energy‐efficient spintronic devices.     

Neural networks with higher-order nonlinearity

Neural networks for associative memory based on the Hopfield relaxation model and matched filtering techniques with higher-order nonlinearity are proposed. These high-order models show dramatic improvement in memory storage capacity and error-correction capability in comparison to conventional binary autoassociative models