Simulation of DC spacecraft power systems

A modular state-variable approach is presented for DC spacecraft power system modeling and simulation. Each modular component is treated as a multiport network, and a state model is written with the port voltages as the inputs. The state model of a component is solved independently of the other components, using its state transition matrix. The state variables of each component are updated, assuming that the inputs are constant. Network analysis principles are then utilized to calculate the component inputs     

OLS算法在无人机传感器故障诊断中的应用

在故障诊断领域,神经网络故障诊断方法以其优良的特性正得到越来越广泛的应用。本文针对无人机的特点提出一种基于RBF神经网络故障诊断方法,通过建立神经网络预测器来实现无人机机载传感器的故障诊断,其中网络学习算法的选取将直接影响神经网络故障诊断的性能。正交最小二乘算法(OLS)以其在设定网络参数方面的优点常用来作为RBF神经网络学习算法。本文将介绍OLS算法的原理和实现步骤,通过VC 6.0编程实现OLS算法,并利用无人机机载传感器数据来验证OLS算法的有效性。     

机器学习辅助湍流建模在分离流预测中的应用

数据驱动湍流建模是近年来发展的提高雷诺平均N-S方程预测精度的有效手段,通过机器学习算法能够从高置信度数据中自动提取特征,建立准确的从平均流动特征到雷诺应力的预测模型。针对高雷诺数下积冰翼型绕流这一类典型的复杂流动分离问题,基于此前研究者提出的机器学习预测框架,从输入输出特征选择和翼型绕流中数据分布特性两个方面出发,对机器学习预测结果的光滑性和准确性进行改善。提出了基于雷诺应力张量分析和流动特征辨识的输入特征选择准则;提出了局部区域建模方法以及基准模型和机器学习预测模型混合的代入计算方法。将改进方法应用于积冰翼型绕流问题之中,结果表明改进的方法能够准确给出训练集和预测集上的雷诺应力结果,并且代入平均流计算可以得到和真实分布更加接近的流动和机翼表面压力分布。     

基于RBF神经网络的航空发动机故障诊断研究

采用改进算法优化了径向基函数(RBF)网络.针对航空发动机工作条件和结构的复杂性,提出了用RBF网络进行故障诊断的方法,构建了基于RBF网络的多参数航空发动机故障智能诊断模型,并对典型发动机故障进行了诊断.结果表明,RBF网络具有优秀的故障学习能力,采用它进行航空发动机故障诊断是行之有效的,具有较好的应用前景.     

Robust fault detection and optimization for a network of unmanned vehicles with imperfect communication channels

The observer-based robust fault detection and optimization for a network of unmanned vehicles with imperfect communication channels and norm bounded modeling uncertainties are addressed. The network of unmanned vehicles is modeled as a discrete-time uncertain Markovian jump system. Based on the model, a residual generator is constructed and the sufficient condition for the existence of the desired fault detection filter is derived in terms of linear matrix inequality. Furthermore, a time domain optimization approach is proposed to improve the performance of the fault detection system. The problem of detecting small faults can be formulated as an optimization problem and its solution is given. For preventing false alarms, a new adaptive threshold function is established. The combined fault detection and optimization algorithm and the adaptive threshold are then applied to a network of highly maneuverable technology vehicles to illustrate the effective- ness of the orooosed aooroach.     

Method of energy efficiency assessment for power supply of autonomous drive systems for perspective mobile objects

Based on an integrated approach, we solve the optimization problem for energy content and weight-dimensional characteristics of the electric servo drive by synthesis of structure and analysis of characteristics for different power supplies. The generalized power supply model, developed for different power supply of drive control systems, is given.     

Parallel operation of shunt active power filters with capacitylimitation control

This paper presents a capacity-limitation technique to enlarge the power capacity that can be handled by the shunt active power filters (APFs). The capacity limited includes the reactive power as well as the amplitude of the distortion current supplied by the APFs. Under capacity-limitation control m+1 APFs supply the load reactive power with m APFs supplying their rated power and one APF supplying a fraction of its rated value. Similarly, n+1 APFs supply the load current harmonic with n APFs supplying the distortion current limited in amplitude and one APF supplying a fraction of its limited amplitude. Advantages of the proposed approach include high flexibility for extending system capacity, high reliability due to no control interconnection, reducing power capacity demand of APF, high modularity due to identical APFs, stable reactive power and harmonic current sharing and its performance is insensitive to parameters mismatch of APFs, cost-effective due to modularization, and so on. Three single-phase 1 KVA APFs are designed and implemented. The effectiveness is demonstrated by some experimental results     

基于LMI和离散模型的航空发动机压气机传感器鲁棒故障诊断

针对航空发动机压气机健康监测提出了一种基于线性矩阵不等式(LMI)和H_∞优化理论的航空发动机压气机传感器鲁棒故障诊断的方法.在航空发动机具有模型不确定性和外界噪声的情况下,应用基于神经网络的线性拟合方法实现航空发动机压气机离散模型的建立;并通过LMI和H_∞优化问题的求解得到未知输入观测器的设计参数,实现具有强鲁棒性的传感器故障诊断.该方法比以前研究中未知输入观测器故障诊断方法的优点在于能够同时处理模型不确定性和外界噪声.应用ALSTOM公司提供的燃气涡轮压气机模型进行了仿真验证,在压气机具有白噪声模型误差和正弦外界干扰的情况下,实现对小于测量范围2%的传感器故障的检测和诊断.      

基于遗传算法的大型空间结构模型降阶

 采用遗传算法求解大型空间结构的模型降阶问题。给出了模型降阶的数学表示和所用遗传算法的有效形式，即编码方式、适值方式、交叉方式和变异方式等。仿真实例说明，遗传算法优于标准内平衡法和Hopfield神经网络方法。     

Monitoring power supply current and using a neural network routineto diagnose circuit faults

As a circuit is tested, the current drawn from a power supply can vary as different functions are invoked by the test. The current draw can be plotted against time, showing a characteristic trace for the test performed. Sensors in the ATS power supply can be used to monitor the current flow during test execution. Defective components can be classified using a Neural Network according to the pattern of variation from the “trace” of a good card. This can be performed as a background function, with the network gaining in accuracy over time. This paper discusses the Neural Network Routine for diagnosing circuit faults using monitored power supply current     

过程功率谱熵在转子振动定量诊断中的应用

针对旋转机械振动过程的复杂性和振动故障产生的随机性,提出了1种以信息熵理论为基础,通过多测点、多转速下的功率谱信息熵(功率谱熵)差矩阵来描述旋转机械振动过程变化规律的故障定量诊断方法。采用在转子试验台模拟转子振动的4种典型故障,得到4个测点多转速下的振动故障数据;对这些故障数据进行分析和处理,求其功率谱熵矩阵。结果表明:通过对转子振动故障信号进行实例计算和分析,该方法在转子振动故障分类和故障严重程度判断方面效果良好。     

人工神经网络在飞机电源系统故障诊断中的应用

分析总结飞机电源系统故障经验后建立故障数据库 ,以该数据库为基础设计实现了一个基于神经网络的飞机电源故障诊断系统。由于诊断网络的权值和结构随着实际的排故结果不断进行更新 ,因此它具有很强的故障诊断能力。经过大量的实际验证 ,该方法故障定位的准确性很高     

基于机匣信号的滚动轴承故障卷积神经网络诊断方法

针对在滚动轴承故障激励下的机匣微弱故障特征,提出了基于卷积神经网络(CNN)的故障诊断方法。利用矩阵图法、峭度图法以及小波尺度谱法3种振动信号的预处理方法,将一维原始信号转换为图像信号;利用卷积神经网络对故障进行识别。通过比较分析发现:通过连续小波尺度谱更易提取滚动轴承的故障特征,其故障识别率达到95.82%,均高于其他几种振动信号预处理方法;由于卷积神经网络可以利用深层网络结构自适应地提取滚动轴承故障特征,比传统支持向量机(SVM)方法的故障识别率高约7%。结果证明了该方法的有效性与可行性,且具有较好的泛化能力和稳健性。      

基于多源诊断信息融合的发动机气路分析

针对气路可测信息的有限性及不确定性易导致航空发动机气路分析结果准确度和精度不高的问题,提出了一种基于贝叶斯网络的多源诊断信息融合机制,以提高气路部件故障诊断的准确度和精确度.介绍了基于贝叶斯网络的气路分析方法,在此基础上提出了改进的用于多源诊断信息融合的气路分析贝叶斯网络,即以常规的气路可测参数为主,而其他诊断信息则借助故障模式先验概率表引进贝叶斯网络,实现多源信息的融合.仿真实验表明:通过融合多源信息能够准确地诊断出传统的气路分析难以识别的故障,同时降低了健康参数估计值的方差,提高了诊断结果的精确度;在观测噪声放大一倍、健康参数变化量小于1%的情况下,通过融合多源信息仍能准确的估计出健康参数变化量,且标准差均于0.1%.      

Real-time expert systems for advanced power control (a statusupdate) [aerospace power system testbed]

The most recent developments in the Boeing Aerospace Autonomous Power System (APS) testbed are presented. The APS testbed is a 28-VDC system with 3-kW capability, assembled for use in developing improved control, techniques for aerospace electrical power systems. The main emphasis is on the development of a sophisticated programming environment to control concurrent execution of multiple autonomous algorithms coupled with a continuous input/output data flow. The integration of high-level control algorithms used for battery charge control into a real-time execution environment is discussed. This includes methods that allow several functions to respond to real-time input, affect/maintain expert system (shared) memory, and control the electrical power system configuration. Sophisticated schemes for scheduling these expert system control functions are required to allow real-time multitasking     

人工神经网络融合诊断航空发动机气路故障

为了提高诊断的准确性,提出了一种利用人工神经网络融合诊断航空发动机气路故障的新算法.由4个子系统有机结合起来建立了神经网络融合诊断系统,对从飞参数据中得到的气路故障数据进行预处理之后,分别输入广义回归神经网络子系统和BP神经网络子系统进行诊断,然后研究了一种新的信息融合算法对两者的诊断结果进行融合,使诊断结果的故障特征更加明显,提高了诊断的准确性.通过测试表明,该信息融合算法十分有效,具有较高的实用价值.      

测试点的优选方法

以系统或设备功能框图和信号流程图为基础,构造出故障信息矩阵,通过测试点对故障检测、隔离或定位贡献大小（权数）的分析,用逐步分解矩阵的方法,选出需要的测试点,确定出最佳诊断程序,达到用最少的测试点和最少的时间检测与隔离故障的目的。     

航空发动机故障诊断算法测试平台

在MATLAB/Simulink环境下搭建航空发动机故障诊断算法测试平台。利用已有的民用大涵道比涡扇发动机非线性模型进行仿真得到航空发动机故障数据,数据既可以用于故障诊断算法的开发,也可以用于故障诊断算法的验证。对用户提供的故障诊断算法按照统一的评价标准进行分析,得出诊断算法优良性的量化值,使得不同结构、不同平台的诊断算法能直接作出比较。通过故障诊断算法的诊断结果进行分析,得到的评价结果可全面反映故障诊断算法的性能。     

基于STPA的飞机交流系统供电转换安全性分析方法研究

飞机交流发电系统是整机的主要电力来源,应对其进行完善的安全性分析。传统安全性分析方法对系统组件间非线性交互引起的安全问题关注较少,当研制型号支持数据不足时,存在分析遗漏风险。根据典型交流发电系统供电转换过程基本特点,基于STPA方法构建安全控制结构图,识别不安全控制行为（UCA）,引入相似系统的失效模式及影响分析（FMEA）结果,分析UCA致因因素和致因场景,使用时间自动机理论的形式化工具进行系统建模与验证;通过专家评判及事故对比来验证该方法的正确性。结果表明：在传统分析方法的基础上引入STPA方法,能够有效识别出不安全控制行为和事故发生的原因,该方法可以作为传统方法的有效补充。     

基于人工免疫网络模型的航空发动机传感器故障诊断

提出了一种用于传感器故障诊断的免疫网络,对其结构和特点进行了分析,给出了相应的诊断算法。对传感器典型故障进行了故障诊断仿真,分析了免疫网络能检测出的最小故障偏差水平以及在不同噪声水平下的故障诊断效果。仿真结果表明,所研究的方法能有效检测到故障传感器,并具有良好的灵敏性及抗噪声干扰能力。