D-2-DenseNet噪音鲁棒的城市音频分类模型

为了提高噪音环境下城市音频分类系统的鲁棒性,提出了一种双特征2阶密集卷积神经网络（D-2-DenseNet）噪音鲁棒的城市音频分类模型.首先介绍了噪音添加和噪音鲁棒处理,阐述了一种双特征互补偿的算法;然后结合2阶密集卷积神经网络与自适应机制提出了一种噪音鲁棒音频分类模型：双特征2阶密集卷积神经网络.模型采用双特征互补偿自适应算法,可在特征提取与模型训练中更有针对性地提取有效音频信息,降低噪音干扰,以提高噪音鲁棒性.最后,基于Dcase2016数据集开展噪音环境下城市音频分类测试.实验结果表明,模型分类准确率分别可达77.12%、75.52%,与基线模型相比,平均分类准确率分别提高了8.51%和10.38%,验证了模型良好的噪音鲁棒性.     

绳牵引刚柔式波浪补偿并联机构的设计与建模

为了减少海上集装箱货物运输中风浪造成的损坏,提出刚柔混合驱动主动式波浪补偿并联机构. 基于刚柔混合并联机构的动定平台的矩阵旋转原理和几何封闭法建立位置逆解数学模型;利用空间几何原理构建机构位置正解数学模型;利用求导法则对位置逆解进行求导,建立速度雅可比矩阵与加速度的二阶影响矩阵;基于绳子是柔性变形体推导出系统刚度矩阵,探究影响系统刚度的因素和增加系统刚度的原则. 利用数值模拟仿真对运动学和系统刚度进行验证,结果表明位置正逆解的输入输出误差不超过实际误差的2.25%;发现理论仿真曲线和样机仿真曲线较吻合,误差不超过7.4%,验证了运动学模型的正确性;根据刚度矩阵发现刚度影响因素对系统刚度的影响规律. 通过对绳驱动刚柔混合驱动波浪补偿并联机构的实验验证,发现该机构的补偿效果均高于90%. 证明研究结果为刚柔混合主动式波浪补偿并联机构的运动和机构设计提供了理论支持.     

深海带电插拔连接器力学特性分析

以水下7 000 m工况的深海带电插拔连接器为研究对象,旨在获得湿插拔连接器在深海双重高压下的受力特性和插拔过程中的动密封性能。介绍连接器的插头结构与压力补偿技术;推导压力补偿器工作容积的计算公式;运用Ansys Workbench建立插头内壳体组件与插孔组件的动密封有限元模型,对内壳体、压力平衡膜及内胆进行强度计算;分析由连接器插拔引起的平衡膜内外瞬时压差对膜的应力分布及变形情况;讨论密封件径向压缩量、密封接触面摩擦因数、插拔速度对插拔过程中动密封性能的影响。结果表明:插头插合时,平衡膜外表面应力基本不变,其变形减小;拔出时,平衡膜外表面的应力及变形都逐渐增大。密封件之间的接触压力受密封接触面过盈配合量的影响明显,增大过盈配合量可显著提高动密封区域的接触压力及密封性能,同时也会加剧密封件失效的可能性,密封接触面的摩擦因数、插拔速度对接触压力的影响则较小。     

GICL型鼓形齿联轴器的研制

分析了国内外鼓形齿式联轴器技术的发展和特点，介绍了GICL型鼓形齿式联轴器的研制过程。     

含小水电分布式电源的配网线路串联补偿优化

以小水电接入电网为例进行分析研究,深入分析小水电给电网带来的不利影响,重点分析对变压器损耗和线路损耗这两方面的影响。最后,举例某地区小水电接入配电网进行分析研究,对丰水期的小水电加入串联补偿装置后线路电压的变化情况展开研究。     

无功补偿技术在智能电网中的应用

无功补偿技术是一项提高电网的功率因数,降低供电变压器及输送电路损耗,从而改善供电环境的技术,其在如今的电力供电系统中占据着重要的位置。本文主要针对无功率补偿技术的应用特点与基本原理进行分析,并分析其在智能电网中的应用。     

Tool deflection compensation in peripheral milling of curved geometries

This paper presents compensation of surface error due to cutting force-induced tool deflections in a peripheral milling process. Previous research attempts on this topic deal with error compensation in machining of straight geometries only. This paper is concerned with peripheral milling of variable curvature geometries where the workpiece curvature changes continuously along the path of cut. In the case of curved geometries, both process geometry and the cutting forces have shown to have strong dependence on workpiece curvature and hence variation of surface error along the path of cut. This calls for a different error compensation strategy than the one which is normally used for machining straight geometries. The present work is an attempt to improve accuracy in machining of curved geometries by use of CNC tool path compensation. Mechanistic model for cutting force estimation and cantilever beam model for cutter deflection estimation are used. The results based on machining experiments performed on a variety of geometries show that the dimensional accuracy can be improved significantly in peripheral milling of curved geometries.     

Langmuir probe RF plasma compensation using a simulation method

The problem of Langmuir probe data deformation due to RF pickup by the probe is treated through a computer simulation method. It is pointed out that proper RF compensations can be obtained by treatment of the Langmuir probe raw data through the use of computer software. It is demonstrated that correct, RF unaffected probe I–V$I\text{–}V$ characteristics can be accurately reproduced from the RF contaminated data. This eliminates the need for the use of any filters or other hardware procedures. User friendly matlab based software is presented. The software automatically retrieves the correct RF I–V$I\text{–}V$ characteristics for single Langmuir probe data which consequently allows for proper evaluation of plasma parameters such as the plasma electron temperature, electron number density and the electron energy distribution function (EEDF)     

我国轮胎产业的状况与展望

着重叙述了我国轮胎产业规模与结构,十大名牌企业代表我国轮胎业最好水平,国内企业形式是国有、国有控股、民营、台资、外资各种形式并存.提出我国企业应形成合理的经济规模,才能实现更大的效益增长.我国轮胎产量逐年提高,子午化率也提高较快.轮胎品质还有待提高,应增强自主开发能力.轮胎的进口价格远高于我国轮胎出口价格,指出只有不断创新,才能使我国成为轮胎出口强国.只要轮胎企业抓住机遇,未来的中国轮胎业将会有更广淘的发展空间.     

Fall detection for multiple pedestrians using depth image processing technique

A fall detection method based on depth image analysis is proposed in this paper. As different from the conventional methods, if the pedestrians are partially overlapped or partially occluded, the proposed method is still able to detect fall events and has the following advantages: (1) single or multiple pedestrian detection; (2) recognition of human and non-human objects; (3) compensation for illumination, which is applicable in scenarios using indoor light sources of different colors; (4) using the central line of a human silhouette to obtain the pedestrian tilt angle; and (5) avoiding misrecognition of a squat or stoop as a fall. According to the experimental results, the precision of the proposed fall detection method is 94.31% and the recall is 85.57%. The proposed method is verified to be robust and specifically suitable for applying in family homes, corridors and other public places.