La1.89Ce0.11CuO4/La0.7Ca0.3MnO3异质结的原位制备及输运性质研究

采用脉冲激光沉积方法,在SrTiO3(STO)基片上原位制备了电子型掺杂高温超导体La1.89Ce0.11CuO4(LCCO)与超大磁电阻(CMR)材料La0.7Ca0.3MnO3(LCMO)异质结.使用X射线衍射和透射电镜分析,表明异质结是c轴取向外延生长.LCCO层的超导转变温度TC0是18.5K,LCMO层的金属-绝缘转变温度为195K.我们研究了这种异质结的界面微分电导,得到LCCO的能隙大小是4.2meV.     

深度复卷积递归网络模型的师生学习语声增强方法

基于深度神经网络的方法已经在语声增强领域得到了广泛的应用,然而若想取得理想的性能,一般需要规模较大且复杂度较高的模型。因此,在计算资源有限的设备或对延时要求高的环境下容易出现部署困难的问题。为了解决此问题,提出了一种基于深度复卷积递归网络的师生学习语声增强方法。在师生深度复卷积递归网络模型结构中间的复长短时记忆递归模块提取实部和虚部特征流,并分别计算帧级师生距离损失以进行知识转移。同时使用多分辨率频谱损失以进一步提升低复杂度学生模型的性能。实验在公开数据集Voice Bank Demand和DNS Challenge上进行,结果显示所提方法相对于基线学生模型在各项指标上均有明显提升。     

基于平均模型和误差削减网络的语音转换系统*

现阶段用于语音转换的深度学习方法多是通过使用大量的训练数据来生成高质量的语音。本文提出了一种基于平均模型和误差削减网络的语音转换框架,可用于有限数量的训练数据。首先,基于CBHG网络的平均模型使用排除源说话人和目标说话人的多说话人语音数据进行训练;然后,在有限数量的目标语音数据下对平均模型执行自适应训练;最后,提出一种误差削减网络,可以进一步改善转换后语音的质量。实验表明,所提出的语音转换框架可以灵活地处理有限的训练数据,并且在客观和主观评估方面均优于传统框架。     

基于BP神经网络的多传感器信息融合研究

BP网络是应用最广的一种人工神经网络,将BP神经网络应用到压力检测领域的温度等非线性补偿,具有重要的实用价值,对压力检测精度的改进效果显著。从传感器信息融合的角度看,神经网络就是一个融合系统。通过对神经网络基本理论的阐述,针对研究对象将BP神经网络原理与多传感器信息融合技术有机集合起来,提出了基于BP神经网络的二传感器信息融合模型及改进算法,建立了BP神经网络训练标准样本库,并对该网络模型进行主要技术指标的测试和仿真工作,测试结果表明构建的模型及其改进算法能很好地满足了高精度压力检测仪的指标要求。     

Development of the liquid level meters for the PandaX dark matter detector

The two-phase xenon detector is at the frontier of dark matter direct search. This kind of detector uses liquid xenon as the sensitive target and is operated in two-phase （liquid/gas） mode, where the liquid level needs to be monitored and controlled in sub-millimeter precision. In this paper, we present a detailed design and study of two kinds of level meters for the PandaX dark matter detector. The long level meter is used to monitor the overall liquid level while short level meters are used to monitor the inclination of the detector. These level meters are cylindrical capacitors that are custom-made from two concentric metal tubes. Their capacitance values are read out by a universal transducer interface chip and are recorded by the PandaX slow control system. We present the developments that lead to level meters with long-term stability and sub-millimeter precision. Fluctuations （standard deviations） of less than 0.02 mm for the short level meters and less than 0.2 mm for the long level meter were achieved during a few days of test operation.     

洗涤时pH值对纳米级β-Ni(OH)2的团聚程度和电化学性能的影响

Nano-scale nickel hydroxide was prepared by precipitate transformation method in the paper. Effect of rinse pH on the agglomeration degree and electrochemical performance of nano-scale Ni(OH)₂ was investigated. The mea-surement results of XRD and TEM indicate that the prepared nano-scale Ni(OH)₂ is β(Ⅱ)-phase,the grain size is in the rang of 10～50nm,and rinse pH exerts a great influence on the agglomeration degree of nano-scale Ni(OH)₂. The agglomeration of material becomes very obvious when rinse pH=11, and the density of nano-scale Ni(OH)₂ is enhanced obviously.Cyclic voltammetry(CV) and simulate cells experiment show that nano-scale Ni(OH)₂ with suitable agglomeration degree have better electrochemical CV performance than those with ideal disperse Ni(OH)₂ and micron Ni(OH)₂, and its proton diffusion coefficient is also the highest. It can elevate the discharge potential platform and prolongs discharge time, so the utilization ratio of Ni(OH)₂ is raised.     

示波极谱同时测定金属锌中铅和镉

本文提出了在ＨＣｌ－Ｈ＿３ＰＯ＿４－ＫＣｌ底液中示波极谱同时测定金属锌中铅、镉含量的方法，检测下限铅为０．１ｍｇ／Ｌ，镉为０．０１ｍｇ／Ｌ铅、锡波形清晰，重视性好。试样中加标准回收，铅回收率为９９．２％～１００％，镉回收率为９５％～９８％，与标准方法比较，相对误差＜５％。方法准确、简便、快速，适用于金属锌中铅、镉的常规分析。     

结合数据场情感空间和混合蛙跳算法的连续语音情感变化趋势检测

针对现有的情感计算算法中存在着情感跟踪延迟的问题,且没有考虑到情感状态的连续性的情况,提出了一种结合数据场情感空间和混合蛙跳算法的连续语音情感变化趋势检测技术。首先构建数据场情感空间,利用情感特征量模拟数据场粒子,用势能函数描述粒子之间的相互作用。然后运用混合蛙跳算法技术,用青蛙个体来模拟情感状态变化过程中的情感特征量,得到情感变化的趋势。通过对变化趋势的分析,可以达到情感预测的目的。经实验证明,该算法性能比现有算法有较大改进。      

一种新型中高温热泵非共沸工质的理论与实验分析

针对工业生产中存在大量可资利用的余热资源,以能够利用余热资源的中高温热泵系统为研究对象,对绿色非共沸混合工质及其循环特性展开了理论和实验分析.通过理论计算筛选出非共沸混合工质ZM1(R125/R152a/R245fa),进行了多工况的实验研究,并与已经使用的由R22和R142b组成的混合工质ZM2在同等工况下进行了对比.结果表明,在满足用户需求这一层面上, ZM1和ZM2水平相当,加上其绿色环保的特性,达到了替代工质的要求,具有成为中高温热泵工质的潜质.     

PandaX实验的进展与展望

PandaX是位于四川的中国锦屏地下实验室（CJPL）内的,利用氙作为工作物质,探测暗物质等其它极稀有事件的大型地下粒子探测实验计划。PandaX探测器中的氙工作在气液两相的模式。通过检测液相和气相中闪烁光强的比例,实现对目标粒子包括WIMP和Axion等疑似信号的鉴别。PandaX已顺利完成了有效区体积约为120 kg氙的一期计划,目前正在运行实施有效区体积约为500 kg氙的二期计划。在WIMP质量为45 GeV/c2附近,探测灵敏度达2.97×10-45 cm2（90%置信度）,没有发现WIMP疑似信号。本工作介绍了PandaX一期和二期的进展,以及后续升级计划,其中包括升级PandaX的时间投影室至4吨,建造用于屏蔽岩石放射性本底的水屏蔽,以及136Xe双贝塔衰变的研究计划等。PandaX is a xenon-based dark matter and other rare events searching project located at China JinPing underground Laboratory (CJPL), Sichuan province, China. Dual phases, liquid and gas phase of xenon are used in the detector. Particles, including Weak Interaction Massive Particles (WIMPs) or axions etc., may induce scintillating signals in both liquid (S1 signal) and gas (S2 signal) phases, and then are identified by analyzing the ratio of S1 and S2. The PandaX had completed its first stage, the 120 kg in fiducial volume, and now is running in its stage Ⅱ with fiducial volume of 500 kg. The running results show that the WIMP-matter cross section detecting sensitivity of about 2.97×10-45 cm2(90% C.L.) has been achieved at a WIMP mass of 44.7 GeV/c2, and no WIMP candidates are identified. The upgrading plan of the PandaX is discussed, which include to build a larger Time project chamber (TPC) of about 4 tons, a water tank for radiation shielding, and the plan of 136Xe double beta decay studies.