基于相对变换的ICA故障检测方法

针对传统独立主元分析方法(independent component analysis,ICA)在标准化处理后导致特征值大小近似相等,难以提取有代表性变量等问题,提出了一种基于相对变换的独立主元分析(relative transformation ICA,RTICA)故障检测方法。该方法引入欧氏距离相对变换理论,将原始空间数据变换得到相对空间,然后在相对空间进行独立主元分析,降低相对空间的数据维数,使提取的独立主元特征具有更大的适应性,建立故障检测模型,最终实现在线故障检测。该方法通过田纳西-伊斯曼过程仿真加以验证,并应用到电主轴裂纹故障的状态监测中,实验结果表明该方法能有效减少独立主元个数,简化故障检测模型的复杂度,增强状态检测性能。     

基于改进DEMD和ICA的海上风机传动系统早期故障诊断

针对传动系统早期故障振动信号较弱的情况,提出基于改进微分经验模式分解(DEMD)和独立分量分析(ICA)的海上风机传动系统早期故障诊断方法。为克服传统的DEMD算法在分解低阶本征模态函数(IMF)时存在失真现象,提出改进的微分经验模式算法将原始振动信号分解成若干个独立的IMF信号,结合ICA进一步进行原始振动信号故障特征分量的提取,并基于标准数据和风机动力传动故障诊断实验平台进行了仿真研究,最后选取海上风电机组传动系统常出现的发电机轴承故障进行诊断分析。结果表明,相对于传统的故障诊断方法,该方法能更好地放大故障分量,减少噪声和其他振动干扰信号的影响,提高了海上风电机组传动系统早期故障诊断的准确性。     

基于解析冗余式的风机传动系统故障诊断

为了以系统有序的方法对风机的具体元件参数进行故障诊断,给出一种基于键合图模型的故障诊断方法。键合图模型是一种跨能域的元件级模型,因此能够定位到具体的故障元件。同时键合图模型能够清晰地表明各元件之间的关系,因此适用于推导解析冗余式。通过解析冗余式可以对系统进行故障检测和隔离。为了能够系统地得到尽量多的解析冗余式,提高故障的可隔离性,该方法首先建立风机的键合图模型,然后由键合图模型导出时间因果图,由时间因果图导出变量关系图,最后由变量关系图消除键合图中结点方程的未知变量得到解析冗余式。实验结果验证了采用该方法推导的解析冗余式能够用于风机参数故障的诊断。     

基于BP算法的电站燃气轮机故障诊断

针对传统故障诊断方法在燃气轮机系统中应用的局限性,研究了基于ＢＰ算法的神经网络方法在电站燃气轮机故障诊断中的应用,通过选择足够的故障样本来训练神经网络,将代表故障的信息输入训练的神经网络后,由输出结果,就可以判断发生的故障种类,这样不仅减小了用于诊断的知识库,而且加快了计算速度,满足了实时在线诊断的要求。     

基于区间预测的风力机桨距执行器LPV故障诊断

针对风力机桨距系统故障导致的桨距角输出变化的问题,在建立风力机桨距系统线性参数变化(Linear Parameter Vary, LPV)模型的基础上,提出了基于区间预测方法的故障诊断方法。首先,以液油含量为调度变量,将风力机桨距系统非线性模型转化为LPV模型,使模型更加精确。其次,考虑到模型不确定性描述的边界问题,引入区间预测算法,根据桨距角输出是否处于区间预测输出上下限内判断故障发生与否。最后,将所提出的算法在风力机系统中进行仿真。仿真结果表明,所提出的算法能够很好地估计出桨距执行器故障,提高了故障诊断的鲁棒性。     

定性仿真技术在故障诊断中的应用

本文根据定性仿真理论，提出了一个故障定性诊断方法。该方法通过建立定性约束方程，分析故障对系统变量的影响，并据此构造出故障决策表。所提出的方法可用于系统知识部分未知情况下的故障诊断，文末以电阻网络的故障诊断为例证明了共有效性。     

基于软件传感器和FMECA的调速系统故障诊断

针对设备进行状态检修的需求,在建立故障诊断的信息模型基础上,设计了一种软件传感器,用于获取在动态过程中不能直接测量的特征参数,结合故障模式、影响及危害分析故障诊断模型实现了状态监测、性能检测和故障诊断的协同工作;提出了水轮机调速器故障诊断系统的体系结构和诊断流程。实例表明,该方法能够解决硬件传感器获取系统性能特征指标的问题,为实现调速器状态检修打下基础。     

Sensor fault detection and isolation for wind turbines based on subspace identification and Kalman filter techniques

This paper aims at the blade root moment sensor fault detection and isolation issue for three‐bladed wind turbines with horizontal axis. The underlying problem is crucial to the successful application of the individual pitch control system, which plays a key role for reducing the blade loads of large offshore wind turbines. In this paper, a wind turbine model is built based on the closed loop identification technique, where the wind dynamics is included. The fault detection issue is investigated based on the residuals generated by dual Kalman filters. Both additive faults and multiplicative faults are considered in this paper. For the additive fault case, the mean value change detection of the residuals and the generalized likelihood ratio test are utilized respectively. For multiplicative faults, they are handled via the variance change detection of the residuals. The fault isolation issue is proceeded with the help of dual sensor redundancy. Simulation results show that the proposed approach can be successfully applied to the underlying issue. Copyright © 2009 John Wiley & Sons, Ltd.     

基于多种信号风力发电机在线故障检测装置设计*

风力发电机在线检测系统可以实时了解风力发电机运行工况及时发现故障隐患。针对检测数据种类过少导致风力发电机故障诊断出现误判问题,文中设计了一种多通道数据采集实时在线故障检测装置。以DSP TMS320F28335(简称F28335)作为系统主控芯片采集现场数据,Lab VIEW软件系统和MCGS组态软件进行同步显示。通过模拟实验表明,与传统的风力发电机在线检测装置相比,能更全面的采集多种运行工况下的数据,为故障诊断提供可靠依据。     

应用深度自编码网络和XGBoost的风电机组发电机故障诊断

针对风电机组现场故障样本难获取的问题,为实现风电机组发电机部件的故障诊断,通过分析风机监控与采集(SCADA)数据,设计了基于深度自编码(DAE)网络和XGBoost的故障诊断算法。该算法包含两部分:第一部分是DAE故障检测算法,通过DAE获取SCADA数据的重构值,分析重构误差的变化趋势与其超越阈值的情况以预测风机故障和提取故障样本;第二部分是XGBoost故障识别算法,用贝叶斯优化搜索XGBoost的最优超参数,建立XGBoost多分类故障识别模型。算例结果表明,DAE算法能够捕获风电机组发电机早期故障,XGBoost比其他算法更精确地识别不同故障类型。     

基于多源信息融合-相关向量机的风力发电机故障诊断

为了能够全面准确地识别风力发电机的故障类别,考虑信号源振动和电流之间的相关性,提出了一种基于信息融合和改进相关向量机相结合的故障诊断方法。通过直驱风力发电机试验台实测数据,提取具有较高敏感度的特征参数作为诊断样本,建立基于振动和电流的改进相关向量机诊断模型进行初步故障诊断。利用信息融合建立多信号源故障诊断模型,获得最终风机故障诊断结果。试验表明,与基于单一信号的故障诊断方法相比,该方法具有更高的准确性,能很好地识别具有机电耦合特性的风力发电机故障类型。     

汽轮发电机组故障智能诊断方法研究

该文针对汽轮发电机组故障特点，研究了用于故障智能诊断的神经网络结构，对传统的BP神经网络进行了改进，提出了神经网络的逐层优化法，优化后的神经网络更加适用于故障诊断。通过对汽轮发电机组常见故障的分析，提出了适用于汽轮机故障诊断的原因-征兆表，在此基础上对某热电厂的汽轮发电机组故障进行了诊断，诊断结果表明，该方法能够正确地诊断出存在的故障，具有实用价值。     

融合粗糙集与神经网络的燃气轮发电机组振动故障诊断方法

针对燃气轮发电机组振动故障诊断中可测参数难以直接反映机组故障状态的问题,提出一种融合粗糙集理论和神经网络的燃气轮发电机组振动故障诊断方法。结合粗糙集对燃气轮发电机组振动信号原始特征数据进行约简,减少冗余信息。将粗糙集与神经网络有机结合,用优化了的神经网络诊断燃气轮发电机组振动故障。试验结果表明了所述方法的有效性,为燃气轮发电机组振动故障的快速诊断提供了可参考的新思路。     

Contribution to the Fault Diagnosis of a Doubly Fed Induction Generator for a Closed-loop Controlled Wind Turbine System Associated with a Two-level Energy Storage System

—In this article, a contribution to the fault diagnosis of a doubly fed induction generator for a closed-loop controlled wind turbine system associated with a two-level energy storage system using an on-line fault diagnostic technique is proposed. This technique is proposed to detect the rotor fault in the doubly fed induction generator under non-stationary conditions based on the spectral analysis of stator currents of the doubly fed induction generator by an adaptive fast Fourier transform algorithm. Furthermore, to prevent system deterioration, a fractional-order controller with a simple design method is used for the control of the whole wind turbine system. The fractional-order controller ensures that the system is stable in both healthy and faulty conditions. Additionally, to improve the production capacity under wind speed fluctuations and grid demand changes, a two-level energy storage system consisting of a supercapacitor bank and lead-acid batteries is proposed. The obtained simulation results show that the objectives of the fault diagnosis procedure and control strategy are reached.     

基于模糊神经网络的燃气轮机热力部件故障诊断研究

针对燃气轮机热力部件故障,提出了基于模糊神经网络的故障检测和诊断方法。在利用模糊规则描述系统故障状态的基础上,通过建立故障诊断目标函数,利用误差反向梯度算法实时修正神经网络连接权值和阈值。仿真结果证明与传统BP神经网络相比,模糊神经网络在对燃气轮机热力部件故障的识别中,具有更高的准确率。     

概率因果网络在汽轮机故障诊断中的应用

在分析了汽轮机振动故障特点的基础上,提出了用遗传算法进行汽轮机等旋转机械故障诊断问题,定义了遗传算法求解故障诊断问题的概率因果网络,给出了求解故障诊断的数学表达式和适合汽轮机等旋转机械的故障集、征兆集、因果强度和先验概率表。建立了汽轮机故障诊断模型指出表达式的最小值的集合对应于故障集和征兆集,该模型能有效地识别出汽轮机的多故障,弥补了专家系统和神经网络等诊断方法不能正确诊断多故障的不足。     

A new aerodynamic decoupled frequential FDIR methodology for satellite actuator faults

This paper presents new results regarding the development of a supervision scheme for a nonlinear satellite model. The main issue concerns the handling of frequency faults affecting the reaction wheels of a spacecraft attitude control system, that is, how to detect and isolate faults, how to determine the different frequencies characterising these faults through spectral analysis and lastly, how to prevent propagation into failures with potential mission abortion as a consequence. Thus, this work investigates the design of a scheme for fault detection, isolation and control reconfiguration applied to the reaction wheels of a spacecraft attitude control, based on the satellite model. This scheme is classifiable as active fault tolerant control. As the study focuses on a general satellite nonlinear model, where aerodynamic and gravitational disturbances, as well as measurement errors, are present, the robustness of the suggested strategy is achieved by exploiting an explicit disturbance decoupling method via a nonlinear geometric approach. To achieve accurate fault diagnosis, aerodynamic disturbance decoupling represents the key point because the aerodynamic model is often uncertain. Moreover, an improvement of the nonlinear geometric approach is presented, to realise both aerodynamic and manoeuvre decoupled fault diagnosis. To the best authors’ knowledge, this is the first works presenting a methodology for frequency fault diagnosis, which is based on the nonlinear geometric approach for fault and disturbance decoupling. The obtained results demonstrate that the proposed methodology can achieve better performances with respect to traditional fault detection and isolation schemes. Copyright © 2013 John Wiley & Sons, Ltd.     

基于堆叠自编码网络的风电机组发电机状态监测与故障诊断

为实现风力发电机的异常检测分析,提出了一种基于风电机组发电机正常状态下数据采集与监控(SCADA)样本数据的堆叠自编码网络深度学习方法。首先将多个自编码网络连接构成深度堆叠自编码网络,选取发电机SCADA状态变量数据作为网络的训练输入,使网络逐层智能提取数据间的分布式规则,从而构建发电机的堆叠自编码学习模型。依据故障状态下发电机SCADA数据内部动态平衡规则被破坏,利用发电机深度学习网络的输入与重构值计算重构误差,并作为整体状态的观测量。通过采用自适应阈值检测重构误差的状态趋势变化,并作为异常预警判定准则,从而实现对发电机故障的判定。当发电机发生异常时,变量的实际值与对应模型的重构值发生较大偏差,表现为状态变量的残差趋势将会偏离原有的动态稳定状态。因此利用状态变量的残差趋势变化对异常变量进行隔离,判定可能的故障原因达到故障诊断的目的。通过对发电机故障前后记录数据进行仿真分析,结果验证了堆叠自编码网络深度学习方法对发电机状态监测与故障诊断的有效性。     

基于加权综合损失优化深度学习和DGA的变压器故障诊断方法

为进一步提高变压器故障诊断效果,提出了一种基于加权综合损失优化深度学习和油中溶解气体分析(dissolved gas-in-oil analysis,DGA)的变压器故障诊断方法。该方法以DGA特征量为输入,以Softmax层各故障状态概率分布为输出,基于堆栈稀疏自编码深度学习理论构建了变压器故障诊断模型。针对常规交叉熵损失函数下,变压器故障诊断效果偏低,训练样本不平衡分布影响故障诊断水平的问题,采用加权综合损失函数对深度学习模型进行优化。案例分析结果表明:相比传统方法,本文方法可削弱训练样本不对称对变压器故障诊断的不利影响并提高变压器故障诊断水平,各训练集下,本文方法故障诊断准确率可保持在90%以上。     

Filter‐based fault detection and diagnosis using output PDFs for stochastic systems with time delays

In this paper, a fault detection and diagnosis (FDD) scheme is studied for general stochastic dynamic systems subjected to state time delays. Different from the formulation of classical FDD problems, it is supposed that the measured information for the FDD is the probability density function (PDF) of the system output rather than its actual value. A B‐spline expansion technique is applied so that the output PDF can be formulated in terms of the dynamic weights of the B‐spline expansion, by which a time delay model can be established between the input and the weights with non‐linearities and modelling errors. As a result, the concerned FDD problem can be transformed into a classic FDD problem subject to an uncertain non‐linear system with time delays. Feasible criteria to detect the system fault are obtained and a fault diagnosis method is further presented to estimate the fault. Simple simulations are given to demonstrate the efficiency of the proposed approach. Copyright © 2006 John Wiley & Sons, Ltd.