Output-relevant Variational autoencoder for Just-in-time soft sensor modeling with missing data

Main challenges for developing data-based models lie in the existence of high-dimensional and possibly missing observations that exist in stored data from industry process. Variational autoencoder (VAE) as one of the deep learning methods has been applied for extracting useful information or features from high-dimensional dataset. Considering that existing VAE is unsupervised, an output-relevant VAE is proposed for extracting output-relevant features in this work. By using correlation between process variables, different weight is correspondingly assigned to each input variable. With symmetric Kullback–Leibler (SKL) divergence, the similarity is evaluated between the stored samples and a query sample. According to the values of the SKL divergence, data relevant for modeling are selected. Subsequently, Gaussian process regression (GPR) is utilized to establish a model between the input and the corresponding output at the query sample. In addition, owing to the common existence of missing data in output data set, the parameters and missing data in the GPR are estimated simultaneously. A practical debutanizer industrial process is utilized to illustrate the effectiveness of the proposed method.     

APP软件性能效率研究

性能效率是APP软件的重要质量属性，但目前缺乏APP软件性能效率的通用模型。分析了APP软件的性能特征，基于ISO/IEC 25010标准提出了APP软件的性能效率模型，定义了APP软件性能效率的子特性和度量指标。基于提出的APP软件性能效率模型，通过实验对APP软件的性能效率进行了度量及相关分析。     

Formation of ST12 phase Ge nanoparticles in ZnO thin films

In this work, we focus on the Ge nanoparticles (Ge-np) embedded ZnO multilayered thin films. Effects of reactive and nonreactive growth of ZnO layers on the rapid thermal annealing (RTA) induced formation of Ge-np have been specifically investigated. The samples were deposited by sequential r.f. and d.c. sputtering of ZnO and Ge thin film layers, respectively on Si substrates. As-prepared thin film samples have been exposed to an ex-situ RTA at 600 °C for 60 s under forming gas atmosphere. Structural characterizations have been performed by X-ray Diffraction (XRD), Raman scattering, Secondary Ion Mass Spectroscopy (SIMS), and Scanning Electron Microscopy (SEM) techniques. It has been realized that reactive or nonreactive growth of ZnO layers significantly influences the morphology of the ZnO: Ge samples, most prominently the crystal structure of Ge-np. XRD and Raman analysis have revealed that while reactive growth results in a mixture of diamond cubic (DC) and simple tetragonal (ST12) Ge-np, nonreactive growth leads to the formation of only DC Ge-np upon RTA process. Formation of ST12 Ge-np has been discussed based on structural differences due to reactive and nonreactive growth of ZnO embedding layer.     

A design of a robust discrete-time controller

In this paper, we proposed a robust discrete-time controller. This control system, which is derived from the idea of the normalized plant, does not include plant parameters. Thus, we obtain a control system independent of plant parameters and that has the same structure as a conventional optimal servo control system. Simulation and experimental results show that the proposed method is fairly robust to plant parameter variations and external disturbances.     

基于FBG传感器网络的煤矿巷道在线监测系统设计

在掘进和采煤的过程中,巷道围岩的受力情况复杂多变,因此,巷道离层位移和围岩压力的实时监测是保证巷道正常工作的一个重要因素。设计了一种基于FBG传感器网络的煤矿巷道在线监测系统。该系统实现井下700 m数据采集,井上分析处理,可提供时序趋势变化、结构分布变化和阈值报警等功能。根据朱仙庄煤矿巷道的特点,系统安装4只FBG位移传感器、4只FBG土压力传感器和4只FBG温度传感器,对巷道掘进区域的离层位移、围岩压力和温度变化进行在线监测。监测结果表明:系统对离层位移、围岩压力和巷道温度变化反映灵敏,能够有效保证巷道掘进的正常进行,且为后续巷道施工提供数据支持和理论依据。     

融合文本特征的汉老双语句子相似度计算方法

双语句子相似度旨在计算不同语言句子间的语义相似程度,在信息检索、平行语料库构建、机器翻译等领域有重要作用。由于汉语、老挝语平行语料稀少,且老挝语在语义表达、句子结构上与汉语有明显差异,导致汉老双语句子相似度研究的难度较大。该文提出了一种融合文本特征的汉老双语句子相似度计算方法,并构建了句子相似度模型。首先,在句子相似度模型中将汉语、老挝语的词性、数字共现等文本特征与GloVe预训练词向量融合,以此丰富句子特征,提升模型计算准确率。其次,由基于自注意力的双向长短时记忆网络组成多层孪生网络来提取长距离上下文特征和深层次语义信息,其中自注意力机制可以保证语义信息的有效利用。最后,采用迁移学习的方法将通用模型参数初始化,并使用不同的微调参策略增强模型的泛化能力。实验表明,该文提出的方法,其召回率、准确率和F₁值分别达到了82.5%、85.78%和84.00%。     

基于改进完全局部二值模式火焰特征提取的 转炉炼钢终点碳含量预测

转炉炼钢火焰图像特征的准确提取是预测终点碳含量的关键,针对于火焰图像相似性高进而难以区分碳含量相近的火焰图像,导致无法准确精准预测碳含量的问题,提出一种改进完全局部二值模式(improved complete local binary pattern,ICLBP)的彩色纹理特征提取方法,用于提取不同碳含量下更具区分性的炉口火焰图像特征并进行终点碳含量的预测.首先,在不同颜色通道下采用局部相位量化(local phase quantization,LPQ)提取图像相位信息,与CLBP提取的图像幅值信息组合成融合特征ICLBP_MP,以增强CLBP算法结构的鲁棒性;然后,通过改进的颜色信息加权策略对其进行加权,以增强火焰图像的颜色对比度信息;最后,使用K近邻回归模型对碳含量进行预测.实验结果表明,碳含量预测在0.02％误差范围内的准确率为83.9％.     

借助汉-越双语词对齐语料构建越南语依存树库

由于对越南语的研究工作相对较少,因此还没有建立规模相对较大的依存树库。相对于已经拥有了形态丰富、语料成熟的汉语,越南语的依存句法分析要困难得多,所以该文提出了一种借助汉-越双语词对齐语料构建越南语依存树库的方法。首先对汉语-越南语句子对进行词对齐处理,然后对汉语句子进行依存句法分析。最后结合越南语本身的语言特点和有关的语法规则将汉语的依存关系通过汉-越双语词对齐关系映射到越南语句子中,从而生成越南语的依存树库。实验表明,该方法简化了人工收集和标注越南语依存树库的过程,节省了人力和构建树库的时间。实验结果表明,该方法相比采用机器学习的方法准确率明显提高。