催化剂级配技术在加氢改质装置上的工业应用

基于柴油产品质量升级的需求,中国石化大连石油化工研究院(FRIPP)开发了FHIDW加氢改质技术及配套FF-46加氢精制催化剂、FC-14B加氢改质催化剂和FDW-3临氢降凝催化剂.该技术及催化剂级配体系在格尔木炼油厂0.8 Mt·a-1加氢改质装置上的工业应用情况表明,级配催化剂体系对原料适应性强,装置运转平稳,操作灵活性高,催化剂失活速率慢,产品分布合理且质量优异,在降低柴油产品凝点的同时可以明显改善其密度和十六烷值,解决了改造前装置生产重柴油密度偏低的问题,为炼油厂生产符合国V质量标准的柴油提供了技术保障.     

Designing Semipermeable Hydrogel Shells with Controlled Thickness through Internal Osmosis in Triple-Emulsion Droplets

Hydrogel shells that compartmentalize the water core from the aqueous surrounding provide molecular selectivity on size and charge in transmembrane transport. It is highly demanding to produce thin hydrogel shells to minimize diffusion length and maximize core volume. Here, internal osmosis in water-in-oil-in-water-in-oil (W/O/W/O) triple-emulsion droplets is used to produce thin hydrogel shells enclosing a large water core. The triple-emulsion droplets are prepared to have an ultrathin middle oil layer using a capillary microfluidic device. The innermost water droplet has a higher osmolarity than the outer water layer containing photopolymerizable hydrogel precursors, which pumps water from the outer layer to the core through the ultrathin oil layer by the osmosis. Therefore, the outer layer gets thinner and hydrogel precursors are enriched while the size of the triple-emulsion droplets remains unchanged. Through photopolymerization of precursors and phase transfer from oil to water, hydrogel shells enclosing water core are produced in the water environment; the oil layer is ruptured for molecular exchange through the shells. The thickness and composition of the hydrogel shells are precisely controllable by the osmotic conditions. The shells show a high permeation rate due to the thinness as well as controlled cut-off threshold of permeation for neutral and charged molecules.     

基于图神经网络的网络性能智能预测

传统网络性能预测技术存在网络状态获取不够全面及网络性能评估准确性欠佳等问题，利用图神经网络学习推理网络关系数据的特点，结合捕获的网络全局信息，提出了一种基于图神经网络的网络性能智能预测方法。通过网络系统抽象及网络性能建模，将复杂的网络信息转化为形式化的图数据进行描述，利用图空域卷积处理图网络节点的消息传递过程，实现网络信息之间的关系推理，研究了实现网络性能预测的图神经网络模型，提出了一种能处理流量矩阵、网络拓扑、路由策略、节点配置的图神经网络体系结构，最后通过实验论证了模型能更好地实现包括时延、抖动和丢包率的网络性能的准确预测。