机车牵引电机故障诊断系统中的数据融合技术应用研究

在分析机车牵引电机常见故障类型及其故障诊断方法的基础上,得出机车牵引电机故障类型的主要特点。通过对数据融合技术的研究,将自适应加权理论的数据级融合技术应用到机车牵引电机故障诊断系统中,并对该技术的可行性进行实验仿真。仿真结果表明,基于数据融合技术的故障诊断方法适合于机车牵引电机这一特殊的系统,并能在机车牵引电机故障诊断系统中取得良好的效果。     

基于二级信息融合的装备故障诊断方法

针对武器装备故障诊断的多故障问题和单一诊断方法推理能力弱、匹配能力差的缺点,采用RBF神经网络数据级融合和DS证据理论特征级融合相结合的方法应用在二级故障诊断模型中,给出了模型实现步骤,并结合某型导弹制导电子箱故障诊断进行了实验验证.该法使诊断不确定性大大减小,克服了单一方法的缺陷与不足,并使武器装备故障诊断的准确度得到提高.     

斯米尔诺夫检验法在数据级信息融合数据关联中的应用

在信息融合系统中，为了确保进入融合中心的信息是满足融合需要的有效信息，在融合之前要对来自多个信息源的数据进行数据关联。作为信息融合前的信号预处理过程，利用数理统计中的斯米尔诺夫检验法对多个同类传感器测得的加速度响应信号进行数据关联，去掉故障传感器或受到严重干扰的通道数据，确保进入融合中心的数据能够有效反映被测对象的状态。通过某二级齿轮箱实验台实测数据的验证，证明该方法能够有效地检验出同类传感器信号是进行故障诊断的有效信号。     

旋转机械故障诊断中的主从混合神经网络模型研究

针对ＢＰ网络在旋转机械故障诊断应用中的不足，借助Ｈｏｐｆｉｅｌｄ网络的优良特性，建立了以反馈式Ｈｏｐｆｉｅｌｄ网络为主控网络、前馈式ＢＰ网络为从网络的主从混合神经网络模型。通过这个网络模型的设计、动力学行为分析、学习算法的描述和测试以及它在旋转机械故障诊断中的应用，结果表明：该网络模型具有收敛速度快、稳定性好、最小系统误差等优点，是一种实现旋转机械故障诊断的优良网络模型。     

基于伪标签深度学习的半监督滚动轴承故障诊断模型

针对实际工程应用中被标记的滚动轴承故障样本收集困难,传统诊断模型精度较低的问题,提出一种伪标签学习融合参数迁移深度学习网络的半监督滚动轴承故障诊断模型。首先将ImageNet数据集上预训练的残差网络（Residual Network,ResNet）模型参数迁移至本文模型中作为初始参数,并使用不同学习率微调网络层参数以加快模型收敛速度;随后引入伪标签半监督学习,使用标签数据训练模型并对无标记数据进行预测以生成伪标签;最后使用标签数据以及伪标签数据训练参数迁移后的ResNet模型,并测试诊断效果。对两种滚动轴承故障数据进行半监督下故障诊断实验及跨域故障诊断实验。实验结果表明,在具有大量未标记样本集下,所提出模型可迁移至不同设备完成诊断,具有较强的鲁棒性,可用于处理复杂工业环境中的故障诊断问题。     

基于数据增强的滚动轴承智能故障诊断方法

目的 针对包装机械设备中滚动轴承应用场景多且有效故障数据难采集而导致的智能诊断方法诊断准确率较低的问题,提出一种基于数据增强的滚动轴承智能诊断方法.方法 首先根据轴承振动信号的故障特征,提出一种数据增强方法,有效扩充训练数据样本多样性.然后采用卷积神经网络对原始样本和增强样本进行故障诊断训练,从而大幅度提高诊断模型的诊断性能.为了验证所提方法的有效性,建立滚动轴承故障试验台并采集轴承故障数据.结果 实验结果表明,在标签训练样本不充足的情况下,提出的方法与不使用数据增强方法相比,模型在诊断准确率方面取得了较大的提高,能够准确地识别各类轴承故障.结论 该方法实现了准确地对稀缺标记样本下滚动轴承故障的诊断,为保证包装机械滚动轴承故障诊断的诊断精度提供了可靠的方法.     

基于迁移学习与加权多通道融合的齿轮箱故障诊断

针对齿轮箱单一传感器故障识别精度波动大、数据利用率低、可靠性低及故障诊断模型在多工况下泛化能力不足等问题，提出了一种加权融合多通道数据与深度迁移模型的齿轮箱故障诊断方法。首先，为了充分挖掘齿轮箱多通道数据的信息，提出了基于信息熵加权的多通道融合方法，采用信息熵法计算各通道数据的融合权重，并对各通道的采样数据进行加权融合。其次，利用源域的融合数据对深度迁移模型进行预训练，将预训练得到的模型参数作为目标域模型的初始化参数，同时冻结目标域模型特征提取器的参数，并利用目标域的融合数据对目标域模型分类器的参数进行微调，实现深度迁移模型从源域到目标域的迁移以适应新的目标样本识别任务。最后，齿轮箱多工况迁移诊断试验结果表明，所提方法可有效用于齿轮箱的故障诊断，相比传统迁移学习方法平衡分布自适应算法(balanced distribution adaptation, BDA)、迁移成分分析(transfer component analysis, TCA)、联合分布自适应算法(joint distribution adaptation, JDA)、统计分布和几何空间联合调整算法(joint geome...     

基于PMS的输变电设备故障诊断系统设计

为提高输变电设备故障诊断能力,研究了基于PMS的输变电设备故障诊断系统设计方法.首先,构建系统总体设计构架,采用便携式设备,对故障信息进行采集;再利用传感器技术进行故障信息采集和热成像处理,进而构建输变电设备故障信息PMS特征提取模型,根据振动波形分析和红外图谱分析,实现故障特征的提取和诊断,并构建输变电设备故障诊断的PMS数据信息融合模型,实现输变电设备的多参量自诊断智能巡视方法设计;最后,利用高集成传感信息处理和实时在线检测方法,对输变电设备的健康状态评估及运行趋势进行预测,实现故障实时诊断.仿真结果表明,采用该方法进行输变电设备故障诊断的效率较高,最短可在3 s时间内检测完毕,故障诊断的精准度水平最高可达到89%,具有一定的应用价值.     

基于BP网络信息融合技术的电路板故障诊断

对基于BP神经网络的信息融合故障诊断技术进行了研究,将信息融合技术应用到某型反舰导弹俯仰综合放大器电路板的故障诊断中,并利用改进的BP神经网络进行数据融合,得到了较为理想的结果.研究表明,该方法能够较好地解决电路板元件故障诊断的不确定性问题.     

堆叠自编码器在样本不充足下的轴承故障诊断方法

深度学习作为一种实用的大数据处理工具,在机械智能故障诊断领域也受到广泛关注,许多研究者已经成功地将深度学习模型应用于故障诊断领域.但这些研究往往忽略了两个重要的问题:(1)当原始训练数据集不足时,模型训练过程不理想;(2)网络模型的学习内容不明确.为了克服上述不足,提出一种新的数据增强的堆叠自编码器(DESAE)框架,...     

Maximum-Likelihood Data Fusion of Phase-Difference and Threshold-Detection Techniques for Wind-Speed Measurement

Analyses of threshold-detection and phase-difference techniques for wind-speed measurement using ultrasonic transducers are presented. The influence of uncertainties that are associated with additive noise and attenuation of the ultrasonic signal on the wind-speed measurement uncertainty is analyzed. A data-fusion procedure based on the maximum-likelihood estimation (MLE) algorithm is developed for the determination of wind speed, with data gathered through threshold-detection and phase-difference techniques. The data-fusion procedure provides a lower measurement uncertainty than those obtained with the above techniques when taken separately. Practical design issues are considered, and an application example is shown to illustrate the proposed procedure.      

Performance evaluation of multi-sensor data-fusion systems in launch vehicles

In this paper, the utilization of multi-sensors of different types, their characteristics, and their data-fusion in launch vehicles to achieve the goal of injecting the satellite into a precise orbit is explained. Performance requirements of sensors and their redundancy management in a typical launch vehicle are also included. The role of an integrated system level-test bed for evaluating multi-sensors and mission performance in a typical launch vehicle mission is described. Some of the typical simulation results to evaluate the effect of the sensors on the overall system are highlighted     

基于模糊神经网络的数据融合结构损伤识别方法

为了有效利用结构健康监测系统中的多源传感器数据信息,提高损伤检测与评估的识别正确率,该文通过构造模糊神经网络分类器,提出了一种基于模糊神经网络的数据融合损伤识别方法并将之应用于结构健康诊断中。它先通过数据预处理,提取原始响应信号中的特征参数,接着将之作为模糊神经网络的输入,构造模糊神经网络模型进行识别决策,最后运用数据融合算法,计算出数据融合后的决策结果。为了验证所提方法的有效性,通过一个7自由度的建筑模型,分别用单一模糊神经网络决策器和数据融合损伤识别方法进行了损伤识别和比较。研究结果表明:该文所提方法比单一决策结果更准确、可靠。     

Composite versus multichannel binary phase-only filtering

Multichannel filtering and its inherent capacity for the implementation of data-fusion algorithms for high-level image processing, as well as composite filtering and its capacity for distortion-invariant pattern-recognition tasks, are discussed and compared. Both approaches are assessed by use of binary phase-only filters to simplify implementation issues. We discuss similarities and differences of these two solutions and demonstrate that they can be merged efficiently, giving rise to a new category of filters that we call composite-multichannel filters. We illustrate this comparison and the new filter design for the case of rotation-invariant fingerprint recognition. In particular, we show that the gain in terms of encoding capacity in the case of the composite-multichannel approach can be used efficiently to introduce multichannel-filter reconfigurability.     

扩展卡尔曼滤波方法在转子典型故障诊断中的应用

转子故障的早期诊断与预示是当前转子动力学的一个难点。文献[1]提出了一种基于多模型估计得到转子裂纹参数对转子进行故障诊断的方法,但是该方法需要构造大量的多模型估计器,存在着构造复杂,计算量大的问题。针对这个问题,提出基于扩展卡尔曼理论(EKF)的转子典型故障诊断方法。针对Jeffcott转子建立了不对中、裂纹和弯曲故障模型,还分别构建了基于扩展卡尔曼滤波-加权整体迭代(EKF-WGI)的参数估计方程。通过不对中、弯曲和裂纹故障的实验验证,表明该参数估计方法对于转子典型故障有着较高的诊断能力。与传统的诊断方法提取频谱特征相比,该方法不依赖于经验和故障事例,可以较精确地估计故障参数,在转子的故障诊断中更有针对性。     

Calibration of an integrated robotic multimodal range scanner

Collecting dense range measurements in uncontrolled environments is a challenging problem, as the quality of the measurements is highly dependent on the lighting conditions and the texture of the target surfaces. This dependence affects the registration and data-fusion processes and, consequently, degrades the accuracy of the surface or occupancy models that are computed from the range measurements. Typical approaches to address this issue have concentrated on improving a specific type of range sensor. On the other hand, the overall quality of the sensing can also be enhanced through the development of a mechanism that combines the various range-sensing technologies to form a multimodal range sensor. The resulting problem of the merging datasets can then be solved in two ways: system calibration of the multimodal sensor or data fitting of all the datasets into a single model, of which the latter is more widely implemented. The lack of multimodal-system calibration approaches is due to their complicated and lengthy nature, where individual calibration procedures must be applied to each subsystem and then applied between the subsystems of the multimodal range sensor. This paper proposes a technique to alleviate the problems encountered in a multimodal-system calibration. Straightforward and generic guidelines for the calibration are defined and applied to an in-house integrated multimodal system built from a laser-range-finder system, two structured-lighting systems, and a stereovision system. The system's intracalibration and intercalibration processes are detailed. Reconstructed renderings of the datasets collected with the calibrated multimodal range sensor, without the use of data fitting, are also presented. From these results, the potential benefits of multimodal calibration over the computationally intensive data-fitting methods and the advantages of merging the subsystem's strengths to complement other subsystem's weaknesses are put in evidence.     

Differential equation-based fault locator for unified power flow controller-based transmission line using synchronised phasor measurements

The fault location algorithm based on a differential equation-based approach for a transmission line employing a unified power flow controller (UPFC) using synchronised phasor measurements is presented. First, a detailed model of the UPFC and its control is proposed and then, it is integrated into the transmission system for accurately simulating fault transients. The method includes the identification of fault section for a transmission line with a UPFC using a wavelet-fuzzy discriminator. Features are extracted using a wavelet transform and the normalised features are fed to the fuzzy logic systems for the identification of fault section. After the identification of the fault section, the control shifts to the differential equation-based fault locator that estimates the fault location in terms of the line inductance up to the fault point from the relaying end. Shunt faults are simulated with wide variations in operating conditions and a pre-fault parameter setting. The instantaneous fault current and voltage samples at the sending and receiving ends are fed to the designed algorithm sample by sample, which results in the fault location in terms of the line inductance. The proposed method is tested for different fault situations with wide variations in operating conditions in the presence of a UPFC.     

基于微分经验模式分解和隐马尔科夫模型的滚动轴承故障诊断方法

提出基于微分经验模式分解（DEMD）和隐马尔科夫模型(HMM）的旋转机械故障诊断方法,并应用到滚动轴承故障诊断中。首先,对故障信号进行基于微分的经验模式分解,提取瞬时能量作为故障特征向量;然后将故障特征向量输入HMM分类器进行模式识别,输出各状态似然概率值;以最大似然概率所对应的故障状态作为诊断结果,最终实现滚动轴承故障诊断。滚动轴承点蚀故障的诊断实验证明了该方法的有效性。与基于EMD-HMM的故障诊断方法相比,基于DEMD-HMM的故障诊断方法更适用于滚动轴承故障诊断。     

一种负荷突变情况下异步电机转子断条故障在线检测新方法

异步电机发生转子断条故障时,其定子电流故障特征频率分量容易被电流基频淹没,加之实际工作中电机负荷突变的干扰,极大地增加了故障特征频率提取及检测的难度。为解决该问题,将解析小波和定子电流谱减法结合,提出一种有效的故障检测新方法。该方法首先利用解析小波变换来判断负荷突变点,然后通过谱减法来消除定子电流频谱中的基频分量,突出故障特征频率,进一步定义故障程度因子来量化转子断条故障程度。仿真和实验分析结果表明,该方法对于负荷突变情况下转子断条故障特征频率更加敏感,能够定量地描述转子故障程度。     

混沌驱动永磁同步电机系统的故障识别

研究了混沌驱动永磁同步电机系统的故障识别问题,设计了一种小波支持向量机故障识别器。首先对故障恢复信号进行经验模态分解,得到若干个平稳的本征模函数,将本征模函数的能量特征作为输入构建小波支持向量机故障识别器。训练完成后,冻结小波支持向量机结构与内部参数,以白噪声模拟实际运行中的未知扰动,并以加入扰动的故障信号作为测试输入,利用小波支持向量机故障识别器进行故障识别。结果表明,基于小波支持向量机的故障识别器能够较好地识别故障信号,拟合误差均在1%以内。