Decomposition and replacement: Spatial knowledge distillation for monocular depth estimation

Knowledge distillation has become a key technique for making smart and light-weight networks through model compression and transfer learning. Unlike previous methods that applied knowledge distillation to the classification task, we propose to exploit the decomposition-and-replacement based distillation scheme for depth estimation from a single RGB color image. To do this, Laplacian pyramid-based knowledge distillation is firstly presented in this paper. The key idea of the proposed method is to transfer the rich knowledge of the scene depth, which is well encoded through the teacher network, to the student network in a structured way by decomposing it into the global context and local details. This is fairly desirable for the student network to restore the depth layout more accurately with limited resources. Moreover, we also propose a new guidance concept for knowledge distillation, so-called ReplaceBlock, which replaces blocks randomly selected in the decoded feature of the student network with those of the teacher network. Our ReplaceBlock gives a smoothing effect in learning the feature distribution of the teacher network by considering the spatial contiguity in the feature space. This process is also helpful to clearly restore the depth layout without the significant computational cost. Based on various experimental results on benchmark datasets, the effectiveness of our distillation scheme for monocular depth estimation is demonstrated in details. The code and model are publicly available at : https://github.com/tjqansthd/Lap_Rep_KD_Depth.     

基于CiteSpace的资源型产业研究知识图谱分析

资源型产业发展为推进国家经济增长和工业化进程提供了重要保障。为深入了解资源型产业的研究情况，以CNKI数据库为数据源，搜集2000—2020年关于资源型产业的核心及以上期刊论文，利用CiteSpace软件从发文作者与研究机构分布、关键词共现网络和时区图谱等方面，绘制知识图谱，进行可视化分析。研究发现：资源型产业领域的研究成果愈加丰富，但研究群体间联系合作较少，且现有的合作研究主要集中在所处地域资源富集和具有学科优势的研究机构及学者；资源型产业领域的研究热点可概括为产业发展、资源型城市、产业集群、产业结构、产业链和产业集聚等方面；针对资源型产业领域未来可从资源型产业相关理论研究、创新发展模式和可持续发展等方面深入展开。     

A tool for mass transfer evaluation in packed-bed columns for chemical engineering students

This paper presents a Microsoft Excel tool to calculate liquid-gas mass transfer coefficients in packed towers to support numerical design activities in the courses of Unit Operations for Industrial Process and Sustainable Process Design for the Master’s degree in Chemical Engineering of the University of Naples Federico II (Italy).The Mass Transfer Solver Tool (MT Solver Tool) uses several available models to estimate, separately, the values of liquid and gas mass-transfer coefficients and the wet surface area for 144 random and structured packings of interest for absorption/stripping and distillation processes. In addition, a separate spreadsheet can be used in a user-defined mode, to evaluate the mass transfer coefficients with new packing types or to interpret experimental data when the geometrical and physical characteristics of the packing are known. Eventually, the tool is supplied with a data library, where packing geometry and model fitting parameters can be retrieved.The software is aimed to support students and educators in the Unit Operations for Industrial Process and Sustainable Process Design courses. In particular, this is meant to be an example on how the accuracy of design algorithms adopted in unit operation processes is affected by the use of the underpinning correlations for mass transfer rate or pressure drops. Besides, this is aimed to encourage comparison of different correlations when exact field data are not available. Besides, chemical engineers and researchers interested in packed columns design and modelling data may also benefit from the utilization of the software. The MT Solver Tool was introduced to students in a dedicated tutorial lesson after lecturers on packed column design algorithms for distillation, absorption and stripping. Most of the students of the course participated to a group training aimed to simulate the design of an absorption column supported by the MT Solver Tool providing feedback on its application.After the training, an anonymous survey was proposed to the students to monitor the approval rating of the proposed activity and the use of the MT Solver Tool software to support numerical calculations.     

气相色谱法测定低酒精度饮料中乙醇的含量

分别采用静态顶空气相色谱法与传统蒸馏后进样的气相色谱法测定低酒精度饮料中乙醇含量。选用TGWAXMS色谱柱。2种方法在乙醇体积分数为0.01%~5.00%时均具有良好的线性关系。传统蒸馏法乙醇的体积分数检出限为0.003%,静态顶空法检出限为0.001%。试验结果表明传统蒸馏后进样的气相色谱法由于前处理步骤繁琐,导致回收率相对较低(83%~91%)。而静态顶空法直接测定饮料中乙醇的方法具有前处理简单,待测物质乙醇不易损失等优点,回收率为97%~100%。     

金属增材制造技术研究热点聚类分析

为深入分析金属增材制造技术分类和发展状况,采集中国期刊全文数据库(CNKI)收录的核心期刊上的768篇科技文献,借助文献分析可视化软件CiteSpace,对关键词聚类进行了全景式描绘,构建金属增材制造技术知识图谱,揭示该技术研究分类以及演化趋势。     

基于主动学习的离群点集成挖掘方法研究

离群点检测任务通常缺少可用的标注数据，且离群数据只占整个数据集的很小一部分，相较于其他的数据挖掘任务，离群点检测的难度较大，尚没有单一的算法适合于所有的场景。因此，结合多样性模型集成和主动学习思想，提出了一种基于主动学习的离群点集成检测方法OMAL（Outlier Mining based on Active Learning）。在主动学习框架指导下，根据各种基学习器的对比分析，选择了基于统计的、基于相似性的、基于子空间划分的三个无监督模型作为基学习器。将各基学习器评判的处于离群和正常边界的数据整合后呈现给人类专家进行标注，以最大化人类专家反馈的信息量；从标注的数据集和各基学习器投票产生的数据集中抽样，基于GBM（Gradient BoostingMachine）训练一个有监督二元分类模型，并将该模型应用于全数据集，得出最终的挖掘结果。实验表明，提出方法的AUC有了较为明显的提升，且具有良好的运行效率，具备较好的实用价值。     

结合预训练模型和语言知识库的文本匹配方法

针对文本匹配任务,该文提出一种大规模预训练模型融合外部语言知识库的方法。该方法在大规模预训练模型的基础上,通过生成基于WordNet的同义—反义词汇知识学习任务和词组—搭配知识学习任务引入外部语言学知识。进而,与MT-DNN多任务学习模型进行联合训练,以进一步提高模型性能。最后利用文本匹配标注数据进行微调。在MRPC和QQP两个公开数据集的实验结果显示,该方法可以在大规模预训练模型和微调的框架基础上,通过引入外部语言知识进行联合训练有效提升文本匹配性能。     

利用蒸馏装置回炼系统降低炼厂挥发性有机物排放

中国石油独山子石化分公司炼油厂为降低VOCs排放对各生产装置采取操作优化措施。其中蒸馏装置降低取样频次，回收取样产生的轻污油，消除设备密封泄漏点；优化蒸馏脱丁烷塔顶及常压塔顶操作，将蒸馏装置石脑油初馏点由42 ℃提高至49 ℃，降低石脑油罐区VOCs排放；将炼油新区各装置送往中间轻污油罐区的轻污油通过蒸馏装置大循环线进入原油泵入口进行回炼，实现中间罐区不接收不储存轻污油，减少了全厂VOCs排放，有效改善厂区生产环境。     

甲基氯硅烷精馏流程的模拟与优化

采用Aspen plus软件对工业七塔精馏过程进行全流程建模与模拟，优化工艺参数，研究了新的精馏节能工艺。对一甲塔等7个精馏塔采用双因素水平的灵敏度分析，考察了塔釜采出率、回流比、进料位置和塔顶压力对产品浓度和热负荷的影响，确定一甲塔最优的工艺参数：塔釜摩尔采出率为0.92，摩尔回流比为130，塔顶压力为0.18 MPa，总理论板数为400，在210块理论板位置进料。在此基础上，针对高能耗的脱高塔/脱低塔，模拟研究了双效精馏新工艺，新工艺可节省39.70%的年总成本；针对一甲塔模拟研究了热泵精馏新工艺，新工艺可降低41.42%的年总成本。     

基于OAV和AEDA对工夫红茶的PLSR分析

采用顶空蒸汽蒸馏提取结合气相色谱-质谱联用技术及气相色谱-嗅闻技术,对4种中国工夫红茶进行定性、定量分析。根据香气成分的香气活力值和芳香萃取物稀释分析分别筛选出30、36种挥发性成分作为香气成分,结合2种方法确定22种香气成分为工夫红茶的香气关键成分,并将其与感官属性和红茶样品进行偏最小二乘法相关性分析。确定对工夫红茶香气起重要作用的22种关键香气物质为:1-戊醇、叶醇、1-辛烯-3-醇、反式氧化芳樟醇、芳樟醇、顺-6-壬烯醇、香叶醇、苯甲醇、苯乙醇、橙花叔醇、戊醛、己醛、反-2-己烯醛、糠醛、(反,反)-2,4-庚二烯醛、苯乙醛、(反,反)-2,4-癸二烯醛、水杨酸甲酯、丙位-壬内酯、2,3-丁二酮、2-庚酮和香豆素。