基于改进的卷积神经网络的步态识别

为了提高惯性传感器采集到的序列数据中步态识别的准确率,建立了一个激励层改进的卷积神经网络(CNN)模型。针对三轴加速度传感器对运动太过敏感导致步态周期划分不准确的问题,采用加速度传感器与弯曲度传感器组合获取人体运动信息。将CNN模型中激励层的线性整流函数(ReLU)改进为带泄露线性整流函数(Leaky ReLU),以解决遇到卷积输出数据小于0时神经元被抑制的问题,进而达到提高步态识别准确率的目的。实验结果表明激励层优化的CNN模型在行走、上下楼和上下坡五种步态模式下识别率达到了95.79%,与未采用弯曲度传感器的改进CNN模型和未进行激励层改进的CNN模型相比,步态识别率有所提高。     

SAR performance-based fault diagnosis for electro-hydraulic control system: A novel FDI framework for closed-loop system

Model-based fault diagnosis serves as an efficient and powerful technique in addressing fault detection and isolation (FDI) issues for control systems. However, the standard methods and their modifications still encounter some difficulties in algorithm design and application for complex higher-order systems. To avoid these difficulties, a novel fault diagnosis framework based on multiple performance indicators of closed-loop control system is proposed. Under this framework, a so-called performance residual vector is constructed to measure the differences between the real system and the nominal model in terms of system stability, accuracy, and rapidity (SAR) respectively. The criteria for quantification, normalization of the SAR residuals and the explicit mappings between the thresholds and the required performance are given. FDI can be easily achieved simultaneously by monitoring the normalized residual vector length and direction in the SAR performance residual space. A case study on electro-hydraulic servo control system of turbofan engine is adopted to demonstrate the effectiveness of the proposed method.     

求解具有多层试验数据成败型单元混联系统可靠性近似限的信息论方法

研究了求解具有多层成败型试验数据时,系统可靠性置信下限近似解的信息论方法,给出了各层试验数据综合的方法和步骤,推导了各层数据信息论方法折合及综合的计算公式。根据单元试验信息提供的信息量与系统折合试验应提供的信息量相等的原则,将多层试验数据逐级折合为系统的等效成败型数据,导出了等效成败型数据信息论方法折合的基本公式;结合可靠性工程的基本原理,根据单个成败型单元可靠性评定的经典精确置信限方法,给出了系统可靠性置信下限的近似解。并以实例形式说明了该方法的具体应用。     

某直升机专用交流地面电源车PSPICE模型仿真分析

针对某直升机专用交流地面电源车系统,建立了PSPICE软件的仿真模型,且进行了系统仿真,研究了基于独立源及受控源的SPWM波产生方法,分析了滤波器参数分布对于系统性能的影响。仿真结果表明该电源系统对于雷达、吊声类冲击性负载有较好的适应性。     

机载红外搜索跟踪系统关键技术分析

概述了红外搜索跟踪系统的组成及主要技术特点,对红外成像系统中的红外望远光学系统、红外探测器组件等进行了分析;探讨了稳定与瞄准系统实现方案;对图像跟踪器中的跟踪算法及工程实现中采取的措施进行了论述。     

基于ARCN模型的轴承故障诊断

提出注意力循环机制与胶囊网络融合的注意力循环胶囊网络（ARCN）的诊断模型。提取时序特征信息构建初级胶囊;自适应融合路由机制、注意力循环机制构建数字胶囊特征;基于西储大学轴承实验数据,验证了ARCN模型的准确率、鲁棒性、稳定性、收敛误差,其准确率相比Caps模型识别准确率提高1.2%、收敛误差达到0.2。基于实验仿真平台,采集正常、内环故障、外环故障和滚动体故障的振动信号,并通过小波基变换获取的时频图构建ARCN模型的数据集。仿真实验结果表明:ARCN模型下,每类故障被误诊的概率不超过总样本的1%。      

Self-perspiration garment for extravehicular activity improves skin cooling effects without raising humidity

Introduction

The current U.S. extravehicular activity (EVA) suit in space includes liquid cooling and ventilation garment (LCVG) to control thermal condition. Tubes knitted in LCVG for flowing water interrupt evaporation of perspiration, and astronauts feel discomfort. In the present study, we hypothesized that a self-perspiration garment would effectively lower the skin temperature without raising humidity in the garment. Thus, we developed and examined the effects of the garment.

Methods

Eight healthy subjects were studied with a cyclic ergometer of 30, 60 90 and 120 W loading for 3 min each. Skin temperature and humidity on the back were measured continuously. Subjects wore and tested three types of garments i.e., a spandex wear without any cooling device (Normal), a simulated LCVG (s-LCVG) or the spandex wear knitted a vinyl tube for flowing water, and the spandex wear with a tube, which flows water and self-perspiration with oozing water for evaporative cooling (SPEC).

Results

All measurements were reached to steady state 2–3 min after the setting. The s-LCVG decreased skin temperature 0.39±0.14 °C during 12 min of cooling. With SPEC, skin temperature did not decrease significantly until 6–9 min after starting the cooling. However, the temperature decreased rapidly and significantly after that, and finally decreased 1.59±0.32 °C. Humidity in the SPEC was significantly lower than that in s-LCVG.

Discussion

SPEC was effective for lowering skin temperature without raising humidity in the garment. The concept is expected to use as a better cooling system during EVA.     

带多孔板的APU进气系统气动性能实验

利用实验方法对一类带多孔板的辅助动力装置(APU)进气系统气动性能进行了研究.实验结果表明:两出口的总压恢复系数和总压畸变指数均随流量的增加不断降低.总压恢复系数在定总流量时略微下降,在定配比时几乎线性下降.总压畸变在周向上主要受多孔板无孔区影响,径向上主要受流道曲率影响.随开孔率增加,总压恢复系数不断增加.动力端不同开孔率下,总压畸变以径向畸变为主,负载端在小开孔率时以周向畸变为主,大开孔率时径向畸变成为主要因素.      

光纤捷联惯性测量组合八点减振与四点减振模式比较

对于光纤陀螺捷联惯性测量组合减振设计中常用的八点减振隔振模式和四点减振隔振模式进行分析比较,建立动力学模型,并进行试验验证。可为光纤捷联惯组减振设计和误差分析提供一定的指导。