Rare earth elements (REEs) doping technology can effectively control the microstructure and improve the quality and performance of materials. This paper summarizes the research progress of REEs in metal additive manufacturing (MAM) in recent years and briefly introduces the effects of REEs on the molten pool fluidity, purified structure, and interfacial bonding between the molten cladding layer and substrate. It focuses on the mechanism of the role of REEs in the refinement and homogenization of microstructures, including grain growth, columnar to equiaxed transition (CET), and elemental segregation. The reasons for the influence of REEs on the homogenization of the structure and elemental segregation are analyzed. The effects of REE type, content, and dimension on hardness and wear resistance are investigated. Finally, tribological applications of REEs in biological and high-temperature environments are summarized, and the impact of REEs-modified alloys is summarized and prospected.

针对可再生能源发电和电动汽车并网的不确定性给微电网运行的经济性和稳定性带来挑战,本文在设定不同充电场景的基础上,构建了考虑光伏发电和出租车充电不确定性的微电网经济运行模型,以最小化微电网总发电成本为目标,各分布式电源的功率限值与功率平衡为约束,采用混合粒子群优化算法进行求解,得到不同充电场景下各分布式电源的最优功率输出和系统在调度周期内的最低运行成本.优化结果表明,综合优化充电行为能提高孤立微电网的可调度性,使微电网的运行成本降低21.2％.

为探究UspA蛋白在光核桃(Amygdalusmira Koehne)中的表达条件及其在逆境胁迫防御反应中的调控作用,利用数据库中碧桃(Prunus persica)的基因序列设计一对特异性上下游引物,通过PCR扩增技术从光核桃叶片中克隆出普遍胁迫蛋白基因AmUs?pA.生物信息学分析结果表明:该基因ORF全长528bp,编码175个氨基酸,无跨膜区域,为亲水性稳定蛋白.蛋白序列分析和系统进化分析表明,AmUspA具有Usp家族典型的UspA结构域,并且与其他植物的Usp蛋白具有较高的同源性.随后构建了pET-21a-AmUspA融合表达载体,优化了蛋白表达条件,即在37℃、IPTG浓度为0.1mmol·L-1,诱导8h时可获得较高浓度的重组蛋白,经纯化脱盐后共得到浓度为0.91mg·mL-1的融合蛋白25mg并制备抗体.最后,通过模拟各种非生物胁迫条件考察AmUspA转化菌株在大肠杆菌中的表达模式,发现各种胁迫条件下转AmUspA的菌株抗逆性均优于空载体,并且在400mmol·L-1 NaCl、40mmol·L-1 NaHCO3、10μmol·L-1 H2O2、100mmol·L-1甘露醇胁迫下优势显著,进一步表明UspA可能在光核桃的防御反应中发挥作用.该研究为探索AmUspA蛋白的生物学作用奠定了基础,同时也为优化桃属种质资源提供了新的思路.

失败知识共享是大数据企业提高创新绩效、降低失败概率的有效途径,能够促进大数据企业的协同创新发展.针对大数据产业联盟成员失败知识共享问题,本文运用演化博弈理论,构建大数据产业联盟成员间失败知识共享演化博弈模型,分析各因素对大数据企业失败知识共享策略演化路径的影响机理.结果表明:通过提高大数据企业间信任水平、容错度、惩罚成本、失败知识共享量、失败知识互补程度和降低失败知识共享难度系数可促进大数据产业联盟成员间的失败知识共享.

In this research,a retina image automatic recognition system based on Convolutional Neural Network (CNN) is proposed for the disadvantages of the traditional retina image processing process which is cumbersome and poor in robustness. First, image preprocessing includes noise removal, numerical normalization, and data volume amplification; then, a new neural network model, XNet, is designed. XNet inherits the advantages of LeNet and Inception networks. The network parameters are based on training. The samples were adjusted adaptively. Finally, the comparison of accuracy and number of iterations was performed for different network structures. The experimental results show that the XNet network structure is superior to LeNet and Inception, and the accuracy rate can reach 91%; and the necessity of data amplification is confirmed through experiments.

In order to solve the “soft field” and uncertainty problems in electrical capacitance tomography, a homotopy extension regularization and selfscaling metric image reconstruction algorithm for electrical capacitance tomography is presentedOn the basis of ECT theory, the algorithm combines homotopy continuation with regularization method, and derives the corresponding correction formula combined with selftuning ratio Finally, the mathematical model of inverse problem of ECT Image reconstruction is derived.The algorithm is used in digital simulation experiment to verify its effectiveness The simulation results are compared with the classical Landweber algorithm, SD algorithm and other imaging algorithms The results show that the algorithm has the advantages of high image quality, fast convergence speed and less iteration times in ECT Image reconstruction It is an effective algorithm to solve ECT imaging problems

由于传统语音识别算法识别耗时长且准确率低,该文提出了一种基于双向循环神经网络来进行语音识别的方法.循环神经网络能够进行记忆,是一种特殊的神经网络,它在NLP领域取得了很大的成功.相比于单向神经循环网络,双向循环神经网络在识别的正确率上有着更大的优势.实验证明,相比于单独的SGMM,DNN等语音识别算法,双向循环神经网络算法对语音识别的错误率更低,对语音识别的研究具有重大意义.

为了改善传统语音识别算法识别不够准确且消耗时间较大的问题,本文提出了一种基于Kaldi的子空间高斯混合模型与深度神经网络相结合的算法进行语音识别.针对声音频率信号识别率较低的问题,本文采用了快速傅立叶变换和动态差分的方法进行MFCC特征提取.实验证明,相比于单独的SGMM、SGMM+MMI等语音识别算法,该算法对语音识别的错误率更低,对语音识别的研究具有重大意义.

无人驾驶技术能够有效的提高交通效率并减少交通事故率,是未来智能汽车技术的重要发展方向,但其面临的问题难点也是前所未有的.本文介绍了无人驾驶技术的构成与功能,并对其在涉水行驶时可能发生的安全问题进行探讨,最后针对该问题提出一种解决方案.

In electrical capacitance tomography technology, to solve the ‘soft-field’ nature and the typical ill-posed problem, a Broyden Correction image reconstruction algorithm is proposed for electrical capacitance tomography system. After the analysis of the basic principles of the ECT system, to solve the problem of electrical capacitance tomography system, we deduced a mathematical model of Broyden Correction algorithm and analyzed the convergence of the algorithm using the monotony of the iterative error. The feasibility of this algorithm solving ECT problems is also discussed. This algorithm meets the convergence condition and the error of image reconstruction is small. Simulation data and experimental results indicate that this algorithm can provide favorable stabilization and high quality images compared with SD, LBP, Landweber and CG and this new algorithm provides a feasible and effective way for ECT image reconstruction field.

To solve the ‘soft-field’nature and the ill-posed problem in electrical capacitance tomography technology，a Huang clan correction image reconstruction algorithm for electrical capacitance tomography is presented. Firstly，according to the basic principles of the Electrical Capacitance Tomography system，the formula of Huang clan correction in the problem of capacitance tomography is derived. Secondly，the iterative formula for simulation experiment is given after the correction. Finally，the validity of the proposed method is verified by digital simulation. The simulation experiment results show that the error of the image for extremely low layer flow，low layer flow and core flow dropped to 24. 39% ，25. 81% and 40. 91% respectively. Ｒesults were lower than Linear Back Projection method，Landweber method，Steepest Descent method and Conjugate Gradient method. In addition，the number of iterations were maintained at 12，12 and 27times，also less than the Landweber method and the Steepest Descent method. The results of the analysis show that the effect of the Huang Clan Correction Image Ｒeconstruction Algorithm are good

A modified implicit type Landweber image reconstruction algorithm is presented for solving the'soft-field'nature and the ill-posed problem in Electrical Capacitance Tomography (ECT) technology.On analysis of the basic principles of the ECT system,the paper proposes a formula of the modified implicit type Landweber algorithm to solve the problem of electrical capacitance tomography,and finishes the iterative modification,and analyzes the convergence of the algorithm.Simulation experimental results show that compared with Landweber,LBP methods,both precision and speed of the modified implicit formula Landweber method are very good,and it is simple and stabilized.

针对近来车辆涉水事件频发所造成的不安全现象, 研究了一种基于MC9S12G单片机的车用智能涉水感应预警系统, 该系统结构合理、成本较低、占用空间小、安装方便, 智能化程度高, 当车辆行驶与停驶时, 遇到暴雨、积水的情况, 该系统可以及时提示车主采取措施, 来降低安全风险以及财产损失.

通过研究多种神经网络模型对视网膜图像的分类效果,提出一种基于迁移学习的Inception V3结构的视网膜图像分类方法.该方法所用的Inception V3神经网络模型经过适当地调整与训练,不仅继承原网络模型结构稀疏性和高性能运算的优点,而且能对只经过简单预处理的图像数据样本做到高精度分类.将该神经网络模型与自己训练的LeNet、AlexNet模型进行对比实验,结果表明:基于迁移学习的Inception V3神经网络模型不仅收敛速度快,而且分类准确率可以达到89％,整体性能优于另外两种神经网络模型.

为了改善传统脑电信号分类时间长、精度不够准确且分类难度较大的问题,利用脑电传感器(MindWave传感器)及RealTerm软件从串口抓取数据获取脑电波TGAM数据包,并对采集的脑电信号数据进行分解计算处理,得到各个波段数据,使用基于负熵的独立分量分析的固定点算法(FastICA)提取脑电信号特征,并用深度学习分类算法对脑电信号进行分类.传统机器学习算法不能准确分类复杂的脑电信号,运用卷积神经网络(Convolutional Neural Network,CNN)提取数据进行训练,构建分类器,实现了对脑电信号更高效更准确的分类.实验结果表明,与Fisher线性判别、BP神经网络、朴素贝叶斯相比,此算法可以更准确地区分是否清醒的状态,对脑电信号分类的研究具有重大意义.

Group connectivity is one of the major research challenges in large-scale industrial wireless sensor networks (IWSNs). Critical nodes (CNs) are mainly responsible to maintain group-connectivity. This article focuses on prioritize these CNs to sleep more than other nodes in order to save their energy resulting prolong global connectivity in group-based IWSNs.

Aiming at the problem of low categorization accuracy and uneven distribution of the traditional text classification algorithms,a text classification algorithm based on deep learning has been put forward.Deep belief networks have very strong feature learning ability,which can be extracted from the high dimension of the original feature,so that the text classification can not only be considered,but also can be used to train classification model.The formula of TF-IDF is used to compute text eigenvalues,and the deep belief networks are used to construct the classifier.The experimental results show that compared with the commonly used classification algorithms such as support vector machine,neural network and extreme learning machine,the algorithm has higher accuracy and practicability,and it has opened up new ideas for the research of text classification.

为了改善传统脑电信号分类不够准确且分类难度较大的问题,研究一种基于方差和深度学习的模型对脑电信号进行分类.针对脑电信号图像识别率较低的问题,采用方差对脑电信号进行特征提取,结合深度学习的一种典型方法——深度信念网络对提取数据进行训练,构建分类器,实现对脑电信号更高效的分类.实验证明,相比于SVM支持向量机、朴素贝叶斯等分类算法,该模型可以更准确地分类.

With the recent advancement in sensing devices, large-scale industries generates a massive amount of data which are generally processed in a centralized manner. In this poster, Fog Computing paradigm is employed for the industrial applications with an aim to reduce traffic overhead leveraging storage and computing services toward edge devices.

In large-scale industrial wireless sensor networks (IWSNs), some nodes are likely to be critical to maintain group-connectivity. Prior studies on topology control with critical nodes (CNs) mainly focus on network connectivity inside a group. It is non-trivial to maintain group-connectivity without considering any CNs in group-based IWSNs. Sleep scheduling is one of the approaches to save residual energy of wireless nodes in energy-constraint IWSNs while satisfying network connectivity and reliability. This article focuses on prioritize these CNs to sleep more than other nodes in order to save their energy resulting prolong global connectivity in group-based IWSNs. The proposed sleep scheduling scheme significantly outperforms the state-of-the-art sleep scheduling in terms of number of critical nodes below critical energy, group-connectivity, and ratio of always-awake critical nodes in the group-based IWSNs.