基于域对抗网络和BERT的跨领域文本情感分析

跨领域文本情感分析时,为了使抽取的共享情感特征能够捕获更多的句子语义信息特征,提出域对抗和BERT(bidirectional encoder representations from transformers)的深度网络模型。利用BERT结构抽取句子语义表示向量,通过卷积神经网络抽取句子的局部特征。通过使用域对抗神经网络使得不同领域抽取的特征表示尽量不可判别,即源领域和目标领域抽取的特征具有更多的相似性;通过在有情感标签的源领域数据集上训练情感分类器,期望该分类器在源领域和目标领域均能达到较好的情感分类效果。在亚马逊产品评论数据集上的试验结果表明,该方法具有良好的性能,能够更好地实现跨领域文本情感分类。     

基于CNN的突发事件预警系统的设计与实现

为提高城市恐怖威胁突发事件的监测预警效率,设计了一套基于改进型卷积神经网络（CNN）的恐怖威胁预警系统。系统对采集的恐怖威胁突发事件进行威胁度评估。相对于现有文本评估方法,提出结合改进型词频-逆文本频率的卷积神经网络用于评估威胁和监测预警。并通过对研判期间内事件的威胁度分析,划分了恐怖威胁等级。通过数据的可视化实现为相关部门提供监测预警信息。模型对比测试和实际运行结果表明,该系统相比于已有的CNN模型及区域型卷积神经网络（RCNN）模型,综合评估的精确度分别提升了5.4%和3%。     

Data‐based multiscale modeling for blast furnace system

The operation of a blast furnace system involves strong multiscale features to which enough attention should be paid, when describing its complex dynamics. For this purpose, a data‐based multiscale modeling algorithm that can extract and pick out some subscale variables most relevant to the output from the original input variables set is presented. These selected subscale variables acting as inputs together with the output are introduced into a general linear or nonlinear model framework to form the corresponding multiscale model. Through model validation, the constructed multiscale models are found to exhibit large advantage compared with those traditional models based on the averaging idea over a fixed scale, especially in the cases of nonlinear models, in which high agreement between the predicted values and the real ones is observed. These results indicate that the proposed multiscale modeling algorithm on the one hand, can provide a kind of thought to develop a data‐based multiscale model from the viewpoint of methodology, and conversely, can serve as a potential blast furnace modeling tool from the viewpoint of engineering applications. © 2014 American Institute of Chemical Engineers AIChE J, 60: 2197–2210, 2014     

Inventory pinch based,multiscale models for integrated planning and scheduling‐part II: Gasoline blend scheduling

Integration of planning and scheduling optimizes simultaneous decisions at both levels, thereby leading to more efficient operation. A three‐level discrete‐time algorithm which uses nonlinear models and integrates planning and detailed scheduling is introduced: first level optimizes nonlinear blend models via multiperiod nonlinear programming (NLP), where period boundaries are initially determined by the inventory pinch points; second level uses fixed recipes (from the first level) in a multiperiod mixed‐integer linear program to determine first an optimal production plan and then to optimize an approximate schedule which minimizes the total number of switches in blenders and swing tanks; third level computes detailed schedules that adhere to inventory constraints computed in the approximate schedule. If inventory infeasibilities appear at the second or the third level, the first‐level periods are subdivided and blend recipes are reoptimized. Algorithm finds the same or better solutions and is substantially faster than previously published full‐space continuous‐time model. © 2014 American Institute of Chemical Engineers AIChE J, 60: 2475–2497, 2014     

Effect of Drying Temperature on Mechanical Properties of Dried Corn

The effect of drying air temperature on the mechanical properties of corn kernels was investigated. Corn was dried at drying temperatures of 40, 50, 60 and 70°C and air flow rate of 1.8 kg/min in a convective dryer. The kernels were then loaded uniaxially in a material testing machine at a loading rate of 3 mm/min, up to the rupture point. An increase in drying temperature from 40 to 70°C increased kernel deformation at the rupture point by an average of 12%. Moreover, values of force, stress, toughness, and modulus of elasticity of corn decreased on average by 21, 26, 36, and 38%, respectively.     

A multi-model approach for soft sensor development based on feature extraction using weighted kernel fisher criterion简

Multi-model approach can significantly improve the prediction performance of soft sensors in the proc- ess with multiple operational conditions. However, traditional clustering algorithms may result in overlapping phe- nomenon in subclasses, so that edge classes and outliers cannot be effectively dealt with and the modeling result is not satisfactory. In order to solve these problems, a new feature extraction method based on weighted kernel Fisher criterion is presented to improve the clustering accuracy, in which feature mapping is adopted to bring the edge classes and outliers closer to other normal subclasses. Furthermore, the classified data are used to develop a multiple model based on support vector machine. The proposed method is applied to a bisphenol A production process for prediction of the quality index. The simulation results demonstrate its ability in improving the data classification and the prediction performance of the soft sensor.     

Numerical investigation for the suitable choice of bubble diameter correlation for EMMS/bubbling drag model

Mesoscale bubbles exist inherently in bubbling fluidized beds and hence should be considered in the constitutive modeling of the drag force. The energy minimization multiscale bubbling (EMMS/bubbling) drag model takes the effects of mesoscale structures (i.e., bubbles) into the modeling of drag coefficient and thus improves the coarse-grid simulation of bubbling and turbulent fluidized beds. However, its dependence on the bubble diameter correlation has not been thoroughly investigated. The hydrodynamic disparity between homogeneous and heterogeneous fluidization is accounted for by the heterogeneity index,H_d, which can be affected by choice of bubble diameter correlation. How this choice of bubble diameter correlation influences the model prediction calls for further fundamental research. This article incorporated seven different bubble diameter correlations into EMMS/bubbling drag model and studied their effects onH_d. The performance of these correlations has been compared with the correlation used previously by EMMS/bubbling drag model. We found that some of the correlations predicted lower H_d by order of a magnitude than the correlation used by the original EMMS/bubbling drag. Based on such analysis, we proposed a modification in the EMMS drag model for bubbling and turbulent fluidized beds. A computational fluid dynamics (CFD) simulation using two-fluid model with the modified EMMS/bubbling drag model was performed for two bubbling and one turbulent fluidized beds. Voidage distribution, time averaged solid concentration and axial solid concentration profiles were studied and compared with the previous version of the EMMS/bubbling drag model and experimental data. We found that the right choice of bubble diameter correlations can significantly improve the results for CFD simulations.     

Modeling the Interaction between Dendrimers and Nucleic Acids: a Molecular Perspective through Hierarchical Scales

Cationic dendrimers are promising nanocarriers for gene delivery thanks to their ability to establish strong interactions with oppositely charged strands of DNA and siRNA and to promote their aggregation. The binding between dendrimers and nucleic acids is typically a complex process that involves various types of interactions at different scales. To design efficient dendrimer candidates for DNA and siRNA binding it is necessary to have a detailed understanding of their interactions with oligonucleotides in the solvent. Molecular simulation can support experimental work, providing a privileged point of view on the aggregation process. This Minireview discusses recent computational efforts to unravel dendrimer–oligonucleotide binding, and proposes a perspective of the multiscale aggregation process based on hierarchy and on the transformations of the interacting “molecular units” following intermolecular interactions.     

Approaching Ultrastable High‐Rate Li–S Batteries through Hierarchically Porous Titanium Nitride Synthesized by Multiscale Phase Separation

Porous architectures are important in determining the performance of lithium–sulfur batteries (LSBs). Among them, multiscale porous architecutures are highly desired to tackle the limitations of single‐sized porous architectures, and to combine the advantages of different pore scales. Although a few carbonaceous materials with multiscale porosity are employed in LSBs, their nonpolar surface properties cause the severe dissolution of lithium polysulfides (LiPSs). In this context, multiscale porous structure design of noncarbonaceous materials is highly required, but has not been exploited in LSBs yet because of the absence of a facile method to control the multiscale porous inorganic materials. Here, a hierarchically porous titanium nitride (h‐TiN) is reported as a multifunctional sulfur host, integrating the advantages of multiscale porous architectures with intrinsic surface properties of TiN to achieve high‐rate and long‐life LSBs. The macropores accommodate the high amount of sulfur, facilitate the electrolyte penetration and transportation of Li⁺ ions, while the mesopores effectively prevent the LiPS dissolution. TiN strongly adsorbs LiPS, mitigates the shuttle effect, and promotes the redox kinetics. Therefore, h‐TiN/S shows a reversible capacity of 557 mA h g^?1 even after 1000 cycles at 5 C rate with only 0.016% of capacity decay per cycle.     

PSO优化多核RVM的模拟电路故障预测

针对模拟电路健康管理的特点，提出了一种基于PSO优化多核RVM的模拟电路故障预测方法。利用参数分析得到电路的输出频域响应作为特征，计算其与电路无故障标准响应的欧氏距离来表征电路元件健康值，将多个核函数线性组合，并用PSO优化多核RVM参数后的模型实现对各个时间点元件的健康值变化轨迹进行预测。仿真结果表明，该方法在小样本情况下，预测效果优于单一核函数的RVM模型，适用于健康管理中实时预测，具有较好的实用性。