浓盐溶液的膜蒸馏机理研究

考察了高浓度氯化钠溶液在直接接触式膜蒸馏操作中的热侧温度，冷侧温度，流速和料液浓度对膜渗透通量的影响。发现膜渗透通量随氯化钠溶液浓度的高而降低。当氯化钠溶液浓度约为25％时，膜渗透通量急剧下降。氯化钠溶液达到饱和时，膜渗透通量逐渐趋于稳定。研究认为此现象是由于膜面上有NaCl结晶形成而导致的。     

新膜及膜过程的研究现状及发展动向

本文综述了双极性膜及其电渗析、膜蒸馏、膜萃取、无机膜等新膜及膜过程，简要说明其研究状况，应用前景和发展动向。     

Precipitation and stability of reversed austenite in 9Ni steel

A new method was used to analyze the factors affecting the precipitation of reversed austenite during tempering. The samples were kept at various tempering temperatures for 10 min and their length changes were recorded. Then, the precipitation of reversed austenite which led to the length reduction was shown by thermal expansion curves. The results show that the effects of process parameters on the precipitation of reversed austenite can be determined more accurately by this method than by X-ray diffraction...     

The assessment of microorganism growth in the membrane distillation system

Growth of microorganism in the membrane distillation (MD) system has been evaluated, and their presence on the membrane surface was studied using scanning electron microscopy (SEM) coupled with energy dispersion spectrometry (EDS). The membrane samples were collected from MD modules used for the following applications: concentration of salts and acids, production of demineralized water from tap water, separation of ethanol from fermentation broth in a membrane bioreactor and treatment of saline wastewater containing Pseudomonas and Streptococcus faecalis bacteria and Penicillium and Aspergillus fungi. The MD experiments were performed with polypropylene capillary membranes. SEM examinations of the membranes used for the treatment of saline wastewater indicated a significant amount of S. faecalis bacteria and Aspergillus fungi. The polypropylene membranes used did not reject S. faecalis bacteria, which were detected on the membrane surface on the distillate side. The presence of fungi in the membrane pores was observed only on the feed side. The contamination of the membrane surface by yeast cells was not observed in the case of the membrane bioreactor operated over one year. The running conditions of the MD process (such as elevated temperature, significant salt concentrations and low pH values) inhibited the growth of microorganism in the MD system.     

Milled fibre reinforced epoxy resin — an engineering polymer composite

A milled fibre reinforced epoxy resin composite (MFRE) has been developed for low temperature use. It is resistant to fracture at temperatures down to at least 77 K in direct contact with liquefied gases. It can be used as an adhesive, a gap-filling mastic or a moulding compound for difficult cryogenic applications. The characteristics of the material and some applications are described.     

柏木油回顾

本文简要回顾中国柏木油之产量、质量和价格。笔者以为中国柏木油价格在国际市场走低多年 ,其主因是香气质量。为改进柏木油香气质量 ,主张以水蒸汽蒸馏技术取代水上蒸馏法 ,不赞成用干馏法或“红锅”法来制取柏木原油。对现有柏木宝贵的天然资源 ,主张由林业部门主导统筹规划和进行科学开发及综合利用。     

Comparison of solar thermal technologies for applications in seawater desalination

This paper deals with a global analysis of the use of solar energy in seawater distillation under Spanish climatic conditions. Static solar technologies as well as one-axis sun tracking were compared. Different temperature ranges of the thermal energy supply required for a desalination process were considered. At each temperature range, suitable solar collectors were compared in some aspects as: (1) fresh water production from a given desalination plant; (2) attainable fresh water production if a heat pump is coupled to the solar desalination system; (3) area of solar collector required for equivalent energy production. Results showed that direct steam generation (DSG) parabolic troughs are a promising technology for solar-assisted seawater desalination.     

基于知识回顾与特征解耦的目标检测蒸馏

当前的知识蒸馏算法均只在对应层间进行蒸馏,为了解决这一问题,提高知识蒸馏的性能,首先分析了教师模型的低层特征对学生模型高层特征的指导作用,并在此基础上提出了基于知识回顾解耦的目标检测蒸馏方法。该方法首先将学生模型的高层特征与低层特征对齐、融合并区分空间和通道提取注意力,使得学生的高层特征能够渐进式地学到教师的低层和高层知识;随后将前背景解耦,分别蒸馏;最后通过金字塔池化在不同尺度上计算其与教师模型特征的相似度。在不同的目标检测模型上进行了实验,实验表明,提出的方法简单且有效,能够适用于各种不同的目标检测模型。骨干网络为ResNet-50的RetinaNet和FCOS分别在COCO2017数据集上获得了39.8%和42.8%的mAP,比基准提高了2.4%和2.3%。     

最优知识传递宽残差网络输电线路螺栓缺陷图像分类

目的 输电线路螺栓图像具有分辨率低和视觉信息较差的特点,针对螺栓缺陷图像分类时教师网络（大模型）参数量大、学生网络（小模型）分类精度低的问题,提出了一种最优知识传递宽残差网络输电线路螺栓缺陷图像分类方法,来弥补使用大小模型进行分类任务的局限性。方法 首先改变大模型宽度,即拓宽网络特征表达维度来增加向小模型传递的螺栓缺陷知识和简化小模型结构至3个残差块;然后为了选出传递螺栓缺陷知识性能最优的大模型,提出知识偏差的概念来可视化大模型向小模型螺栓缺陷知识传递的程度,综合分析不同宽度大模型利用知识蒸馏算法与注意力转移算法分别指导小模型训练后的分类准确率,并用知识偏差来确定最优知识传递模型;最后将最优知识传递模型利用知识蒸馏算法与注意力转移算法相结合指导小模型训练,尽可能提升小模型的螺栓缺陷分类性能。结果 在自建螺栓缺陷图像分类数据集上进行了验证,结果表明宽度为5的大模型向小模型传递螺栓缺陷知识性能最优,使小模型的螺栓缺陷分类准确率提高了5.56%,小模型与最优大模型的分类准确率只差2.17%,知识偏差为0.28,且小模型的参数量仅为大模型参数量的0.56%。结论 本文提出的最优知识传递宽残差网络输电线路螺栓缺陷图像分类方法,弥补了大小模型螺栓缺陷图像分类的局限性,实现了精度与资源消耗的平衡。     

基于自编码和知识蒸馏的表面缺陷检测方法

针对传统的表面缺陷检测方法只能对具有高对比度或低噪声的明显缺陷轮廓进行检测的问题,提出了一种基于自编码和知识蒸馏的表面缺陷检测方法来准确定位和分类从实际工业环境捕获的输入图像中出现的缺陷。首先,设计了一种级联自动编码器（CAE）架构用于分割和定位缺陷,其目的是将输入的原始图像转换为基于CAE的预测蒙版;其次,利用阈值模块对预测结果进行二值化以获得准确的缺陷轮廓;然后,把缺陷区域检测器提取并裁剪出来的缺陷区域视为下一个模块的输入;最后,将CAE分割结果的缺陷区域通过知识蒸馏进行类别分类。实验结果表明,与其他几种表面缺陷检测方法相比,所提出的方法综合性能最好,其缺陷检测平均准确率为97.00%。该方法能够有效地对较小的、边缘不清晰的缺陷进行分割,满足对物品表面缺陷实时分割检测的工程要求。