一种新的动态权重更新相关反馈方法*

提出了正负相关图像所占比例与特征值的标准方差相结合来调整对应权值的方法.为了避免相关反馈的权值调整陷入局部最优状态,引入一个干扰因子,使权值跳出局部最优区域.结果表明了该方法的优越性.     

基于变压器反馈的平坦噪声系数超宽带CMOS低噪声放大器

在系统中集成超宽带(UWB)收发机芯片用于支持室内定位正成为移动通信终端技术发展的一个重要趋势.在超宽带收发机中,低噪声放大器(LNA)是一个核心功能模块.超宽带的全频段(3.1～10.6 GHz)覆盖要求给低噪声放大器的设计带来了巨大挑战,尤其是需要在宽带匹配及在带内维持平坦的噪声系数的情况下.传统的低噪声放大器架构应用在超宽带设计时,噪声、增益和输入匹配之间存在较明显的性能折中关系,因此无法达到良好的综合性能指标要求.本文采用基于变压器反馈的输入匹配的第一级架构和多功能第二级输出驱动结构,实现了平坦的噪声系数和高增益等性能.基于TSMC 65 nm工艺设计的电路仿真结果表明,该低噪声放大器在3.1～10.6 GHz全频段内,可实现输入匹配S_(11)-10 dB,增益17 dB,噪声系数2.71±0.28 dB,1-dB压缩点-17.5 dBm等指标,电路整体功耗为32.8 mW.因此,综合性能Ⅰ(FoM-Ⅰ)和综合性能Ⅱ(FoM-Ⅱ)分别可达2.32和0.41.     

采用VSG策略的MMC阻抗建模及并网稳定性分析

随着分布式能源并网的增多,局部电网逐渐呈现出弱电网特性,虚拟同步发电机（VSG）控制由于可以模拟同步发电机的特性,能够为弱电网提供惯性和阻尼支撑。本文采用谐波线性化方法对采用VSG控制策略的模块化多电平换流器(MMC)进行阻抗建模。所得到的阻抗模型在低频段主要呈容性,在高频段主要呈感性,在复杂电力系统中,可能会与电网阻抗存在谐振点。分析了VSG主要控制参数对MMC阻抗的影响,发现有功-频率下垂系数、阻尼系数和虚拟转动惯量主要影响50Hz附近的阻抗。根据阻抗稳定性判据对MMC并网系统进行稳定性分析,发现并网系统在高频段有振荡的风险。     

直动式电反馈高压电气比例减压阀的研制

介绍一种新型的直动式电反馈高压电气比例减压阀的结构原理。实验结果表明，该减压阀调压范围宽，调压精度高，具有较好的静、动态特性。     

焊接与切割设备的使用和维修(五十二)--TIG焊机的使用与维修

讲述了并接电容排除WS-200钨极氩弧焊电流表减动作失灵的方法。详细分析了WSE-315氩弧焊机的原电路电流反馈和改进后的电流反馈原理；例举了WSE-500多功能脉冲钨极氩弧焊机故障排除的事例。     

国外现代铝带冷轧机和国产先进冷轧机

介绍了国外现代化铝带冷轧机的特点和典型轧机及国内铝带冷轧机的发展。     

气动位置伺服系统的高精度控制研究

提出采用电-气压力比例阀构成气动位置伺服系统,运用全局线性化的观点线性化描述系统动态特性的数学方程,补偿了气缸活塞行程变化对系统控制性能的不良影响,并采用离散线性二次最优控制原理设计了系统的控制补偿器,实现了气动位置伺服系统的高精度控制。     

SEQUENTIAL LOOP CLOSING IDENTIFICATION OF MULTIVARIABLE PROCESSES USING THE BIASED RELAY FEEDBACK METHOD

Each controller in multiloop control systems for multivariable processes can be tuned sequentially with the ultimate information for the paired input and output while former loops have been closed, and hence, single-input single-output autotuning methods can be applied. In this sequential autotuning for multiloop control systems, several iterations are usually required for better control performances. Especially when pairings are undesirable, the autotuning sequences should be repeated with correct pairings, which result in long field experiments. Here, to avoid this drawback, a simple method to identify process models while loops are being sequentially tuned is proposed. The identified models can be used to correct pairings of multiloop control systems and to improve tuning performances without several field iterations. In addition, they can be used to obtain model-based control systems such as decoupling control systems.     

Testing usability and trainability of indirect touch interaction: perspective for the next generation of air traffic control systems

This study aims to determine whether indirect touch device can be used to interact with graphical objects displayed on another screen in an air traffic control (ATC) context. The introduction of such a device likely requires an adaptation of the sensory-motor system. The operator has to simultaneously perform movements on the horizontal plane while assessing them on the vertical plane. Thirty-six right-handed participants performed movement training with either constant or variable practice and with or without visual feedback of the displacement of their actions. Participants then performed a test phase without visual feedback. Performance improved in both practice conditions, but accuracy was higher with visual feedback. During the test phase, movement time was longer for those who had practiced with feedback, suggesting an element of dependency. However, this ‘cost’ of feedback did not extend to movement accuracy. Finally, participants who had received variable training performed better in the test phase, but accuracy was still unsatisfactory. We conclude that continuous visual feedback on the stylus position is necessary if tablets are to be introduced in ATC.     

基于状态聚类的多站点CSPS系统的协同控制方法

单站点传送带给料加工站（Conveyor-serviced production station,CSPS）系统中,可运用强化学习对状态——行动空间进行有效探索,以搜索近似最优的前视距离控制策略.但是多站点CSPS系统的协同控制问题中,系统状态空间的大小会随着站点个数的增加和缓存库容量的增加而成指数形式（或几何级数）增长,从而导致维数灾,影响学习算法的收敛速度和优化效果.为此,本文在站点局域信息交互机制的基础上引入状态聚类的方法,以减小每个站点学习空间的大小和复杂性.首先,将多个站点看作相对独立的学习主体,且各自仅考虑邻近下游站点的缓存库的状态并纳入其性能值学习过程;其次,将原状态空间划分成多个不相交的子集,每个子集用一个抽象状态表示,然后,建立基于状态聚类的多站点反馈式Q学习算法.通过该方法,可在抽象状态空间上对各站点的前视距离策略进行优化学习,以寻求整个系统的生产率最大.仿真实验结果说明,与一般的多站点反馈式Q学习方法相比,基于状态聚类的多站点反馈式Q学习方法不仅具有收敛速度快的优点,而且还在一定程度上提高了系统生产率.