基于残差决策的自动飞行控制系统远程实时故障诊断

针对自动飞行控制系统结构复杂、关联部件众多,发生故障时诊断时间长,从而影响飞机运行效率的问题,提出一种基于飞机通信寻址报告系统(ACARS)的远程实时故障诊断方案。首先,分析自动飞行控制系统的故障特点,设计搭建检测滤波器;然后,利用ACARS数据链实时发送的自动飞行控制系统的关键信息进行相关部件的残差计算,并根据残差决策算法进行故障诊断及定位;最后,针对不同故障部件残差间的差异大、决策门限无法统一的缺点,提出基于二次差值的残差决策改进算法,减缓了检测对象的整体变化趋势,降低了随机噪声和干扰的影响,避免了将瞬态故障诊断为系统故障的情况。实验仿真结果表明,基于二次差值的改进残差决策算法避免了多决策门限的复杂性,在采样时间为0.1 s的情况下,故障检测所需时间大约为2 s,故障检测时间大幅降低,有效故障检测率大于90%。     

Supporting primary students' learning of fraction conceptual knowledge through digital games

With the advent of mobile technologies, well-designed fraction apps can be used to help children gain fraction knowledge, a challenging topic for both teachers and students. The present pilot study adopted a quasi-experimental design to investigate whether children can learn fraction concepts equally well if half of the lesson time (20 min) is replaced with game-based learning. Keeping the total lesson time (40 min) identical, the control group (N = 33) received traditional instruction, and the experimental group (N = 32) was presented with a blended learning approach spending half of the class time (20 min) playing tablet-based fraction games, where each of the learners had their own tablet. The results suggested that in the posttest, the experimental group achieved similar learning gains to the control group and appear to have achieved better performance in the transfer test than the control group. This paper also discusses the efficiency of game-based learning, the mechanism of how fraction games might enhance learning, and the potential of integrating game-based learning in educational settings.     

植物基人造肉的营养特性与食用安全性

植物肉的出现为动物肉类食品供应短缺和养殖业环境污染问题带来希望。食品加工技术的快速发展弥补了植物基肉制品外观和口感的不足, 但其营养价值和安全性也应引起重视。本文综述了以植物蛋白为主要原料所制肉类替代品的营养价值, 包括植物肉中碳水化合物、蛋白质、脂肪、水分、维生素和矿物质的相对含量和营养性质。分析了植物肉在生产及食用过程中可能出现的物理、化学、生物因素在内的安全问题, 并探讨目前植物肉发展所存在的局限性与挑战, 以期为我国植物蛋白肉制品的研发与推广提供理论参考。     

（S）-噻吗洛尔半水合物的合成及表征

为了开发β受体阻断剂新药（S）-噻吗洛尔半水合物，采用3-吗啉-4-氯-1，2，5-噻二唑为起始原料，经水解反应得到中间体1（3-吗啉-4-羟基-1，2，5-噻二唑）。中间体1与R-环氧氯丙烷发生醚化反应，经后处理及重结晶得到中间体2 {（R）-4-［4-（环氧乙烷-2-基甲氧基）-1，2，5-噻二唑-3-基］吗啉}。中间体2经胺化反应、马来酸成盐及重结晶得到（S）-马来酸噻吗洛尔。（S）-马来酸噻吗洛尔经游离、纯水转晶得到符合药典标准的（S）-噻吗洛尔半水合物，总收率14.05%且e.e.值为99.66%。最终成品经IR、1H-NMR、13C-NMR、MS、TGA、DSC表征，并优化各步反应条件。结果表明:以三乙胺为醚化反应缚酸剂75 ℃反应最佳；以乙醇为胺化反应溶剂46 ℃反应16 h最佳；S-噻吗洛尔的转晶拆分以水作溶剂，比传统不对称合成工艺安全稳定，操作简单，适合工业化生产。     

Polyethylene oxide-polypropylene oxide -polyethylene oxide derived porous carbon materials with different molecular weights as ORR catalyst in alkaline electrolytes

Polyethylene oxide (PEO)-polypropylene oxide (PPO)-polyethylene oxide block copolymer having different molecular weights are used as precursors of carbon materials to prepare Hollow -Derivatives carbon material as an electrocatalyst through block copolymer self-assembly. The composition and microstructure of the prepared catalysts are shown by Raman spectroscopy, X-ray diffraction (XRD), Test of nitrogen adsorption and desorption curves, High resolution transmission electron microscopy (HR-TEM) and scanning electron microscopy (HR-SEM). Oxygen was passed into alkaline electrolyte solution until the solution reached saturation state. With molecular weight increasing, the obtained sample gradually changed from block to hollow and spherical. When the molecular weight was 12600 g mol^?1, the evenly hollow carbon nanocages was acquired (C-12600). In O₂ saturated alkaline electrolyte (0.1 M KOH solution), C-12600's limited current density，half-wave potential and initial potential are 5.23 mA cm^?2@0.4 V, 0.72 V and 0.81 V, respectively. And most important is that half-wave potential and onset potential have barely change after 2000 cycles of cyclic voltammetry. As a result, the porous carbon materials exhibited excellent electrocatalytic activity while maintaining high stability in alkaline KOH solution.     

改进HSR-FCN的服装图像识别分类算法研究

目前网络上的服装图像数量增长迅猛，对于大量服装图像实现智能分类的需求日益增加。将基于区域的全卷积网络（Region-Based Fully Convolutional Networks，R-FCN）引入到服装图像识别中，针对服装图像分类中网络训练时间长、形变服装图像识别率低的问题，提出一种新颖的改进框架HSR-FCN。新框架将R-FCN中的区域建议网络和HyperNet网络相融合，改变图片特征学习方式，使得HSR-FCN可以在更短的训练时间内达到更高的准确率。在模型中引入了空间转换网络，对输入服装图像和特征图进行了空间变换及对齐，加强了对多角度服装和形变服装的特征学习。实验结果表明，改进后的HSR-FCN模型有效地加强了对形变服装图像的学习，且在训练时间更短的情况下，比原来的网络模型R-FCN平均准确率提高了大约3个百分点，达到96.69%。