基于压缩感知与改进的深度极限学习机的轴承故障诊断方法

大数据时代下的轴承故障监测存在海量数据处理实时性和故障特征选取的主观性问题,为了实时、智能的实现轴承故障诊断,提出了压缩感知(Compressed Sensing,CS)与改进的深度极限学习机(Multilayer Extreme Learning Machine,ML-ELM)相融合的轴承故障诊断新方法.该方法首先通过压缩感知理论从大量轴承监测数据中获取能够表达特征信息的少量数据,然后用轴承信号在压缩感知变换域中的测量值进行由PSO改进的深度极限学习机分类识别,实现故障智能诊断.此方法大幅减少了轴承诊断信号的数据量并省去了智能诊断时特征选取的步骤,充分利用了深度极限学习机从少量测量值中挖掘轴承信号的特征信息,实现了智能、准确的分类.实验分析表明:该方法对不同位置、不同损伤程度的故障能够准确的识别,为轴承状态监测和故障诊断提供了新方法.     

数据驱动的轴承早期故障诊断技术综述

轴承早期故障的实时诊断,是实际工程应用需求和基础科学问题研究的交汇点,是轴承故障诊断的发展方向之一。首先,阐述了轴承故障及演变过程;其次,根据轴承早期故障实时诊断的需求,总结了轴承早期故障诊断难点问题;之后,重点论述了轴承早期故障诊断3个关键环节所采用的各种技术：微小监测信号增强技术、监测数据的融合表示技术以及早期故障智能诊断技术;最后,总结展望了轴承早期故障诊断技术的发展趋势。     

基于多维度排列熵与支持向量机的轴承早期故障诊断方法

针对许多现有方法无法有效诊断滚动轴承早期故障的问题,引入排列熵的方法对轴承振动信号进行早期故障分析。通过研究嵌入维数和延迟时间对信号排列熵的影响,提出多维度排列熵的特征提取方法。利用多维度排列熵方法所提取的特征,建立了基于支持向量机的轴承早期故障智能诊断模型。对轴承不同类型、不同程度的故障数据进行分析,证明了多维度排列熵方法可以有效提取轴承不同状态的特征信息,与支持向量机结合的智能诊断模型可以精确地诊断轴承不同类型的早期故障,具有很强的通用性;该模型在贫样本的情况下,依然具有很高的诊断精度,适用于滚动轴承早期故障状态的在线监测。     

基于压缩采集与深度学习的轴承故障诊断方法

机械健康监测正进入大数据时代,针对传统轴承故障检测存在的采样数据量大、故障特征依赖主观选取等问题,研究了轴承故障信号变换域的压缩采集、自动特征提取及诊断方法。基于压缩感知和深度学习理论,研究用随机高斯矩阵实现轴承信号的变换域压缩采集,并将此信号输入深度神经网络实现故障的智能诊断。该方法克服了传统针对时域信号的特征提取计算复杂、受先验知识和主观经验影响较大等问题,直接利用含有大量故障信息的压缩感知域采集信号训练深度神经网络,充分利用深度学习挖掘少量压缩采集数据中隐藏的故障信息,从而实现智能、准确的分类。实验结果表明,该方法实现了对不同故障位置和缺损程度的故障特征自动提取与准确故障诊断。     

轴承故障智能诊断专家系统的研究CSCD

为满足RV减速器轴承故障分析的需要,实现对轴承故障的智能诊断,提出了一种基于专家系统的轴承故障智能诊断方法。针对RV减速器轴承的特点,通过模拟实际工况进行试验,提取不同故障类型的特征数据,结合轴承故障判别依据以及先验知识,采用Visual Studio软件建立RV减速器轴承故障智能诊断专家系统,提高轴承故障类型的识别率以及故障位置、程度的诊断精度。通过试验对故障诊断推理模块、知识更新模块、知识库管理模块等功能模块进行了验证,保证了专家系统的可靠性。     

轴承故障智能诊断专家系统的研究

为满足RV减速器轴承故障分析的需要,实现对轴承故障的智能诊断,提出了一种基于专家系统的轴承故障智能诊断方法。针对RV减速器轴承的特点,通过模拟实际工况进行试验,提取不同故障类型的特征数据,结合轴承故障判别依据以及先验知识,采用Visual Studio软件建立RV减速器轴承故障智能诊断专家系统,提高轴承故障类型的识别率以及故障位置、程度的诊断精度。通过试验对故障诊断推理模块、知识更新模块、知识库管理模块等功能模块进行了验证,保证了专家系统的可靠性。     

基于模糊支持矢量数据描述的早期故障智能监测诊断

为了解决机电设备早期故障难以正确识别及故障发展状态不易准确监测的问题,提出了一种基于模糊支持矢量数据描述(FSVDD)的早期故障智能监测诊断新方法。该方法只需要一类目标样本作为学习样本就可以建立起单值分类器,同时在核函数中引入非目标样本的模糊隶属度,从而把非目标样本与目标样本分等级地区分开来。将这种方法应用在机电设备状态监测和故障诊断中,只需要将正常运行时的数据信号作为目标样本,就可以实现对设备早期故障的准确识别,同时判断故障的严重程度。在轴承运行状态监测中的测试结果表明,该方法不仅能快速识别轴承的早期故障,而且可以对故障的严重程度做出准确的判断。     

基于双时域微弱故障特征增强的轴承早期故障智能识别*

针对轴承早期微弱故障难以准确识别的问题,提出一种基于双时域微弱故障特征增强的轴承早期故障智能识别方法。利用广义S变换和Fourier逆变换推导出一种双时域变换,将轴承振动信号变换为双时域二维时间序列。根据双时域变换的能量分布特点,提取二维时间序列的主对角元素以构建故障特征增强的时域振动信号。仿真信号和轴承故障信号分析验证了双时域微弱故障特征增强的可行性和有效性。采用脉冲耦合神经网络和支持向量机对增强后的轴承信号进行时频特征参数提取和智能识别,平均识别精度达到了95.4%。试验结果表明所提方法能有效提高轴承早期故障的智能识别精度。     

振动筛故障智能诊断系统研究

研究分析一种针对选煤厂振动筛在使用过程中可能出现的振动筛异常振动、激振器轴承故障以及支撑弹簧故障等问题而设计的振动筛故障智能监测系统.该系统通过布置的压力、温度以及加速度传感器等监测振动筛运行状态,并通过软件系统中的自动学习、控制以及预警等功能模块实现振动筛故障智能诊断,可在一定程度上提升振动筛运行保障能力.     

利用轴承磨合机增加段修故障轴承诊断的探讨

在轴承磨合机上加装轴承智能诊断功能,实现信息采集、频谱分析、故障程度、故障部位等诊断处理,完善了轴承磨合机诊断功能,强化了段修轴承抽检手段,提高了段修轴承故障判断准确率。     

智能轴承关键技术及发展趋势

智能轴承是具有自感知、自决策及自调控功能的轴承单元,是高端轴承发展的主要方向。首先,分析了智能轴承的特点,总结了智能轴承的国内外研究进展和发展动态,认为目前智能轴承技术尚处在初级阶段,自决策和自调控功能尚不成熟;其次,探讨了智能轴承关键技术,包括传感器技术、自供电/无线供电技术、信号无线传输技术、轴承状态智能评估及智能诊断技术以及轴承状态智能调控技术;最后,讨论了智能轴承系统化、监测信息多样化及智能化的发展趋势。     

基于微传感器的智能轴承技术

介绍了近年来新发展的基于微传感器的智能轴承技术。根据智能轴承系统的组成，该技术主要涉及四个方面的问题，即智能轴承机械结构的设计与分析、微传感器的设计与开发、信号采集与传输技术、故障信号的处理与分析技术。分别对智能轴承这四个方面的技术与发展进行了阐述。     

MEMs hydrodynamic bearings: Applications to and implications for machine failure prevention

Microdynamic systems have been studied for a number of years. Only limited work, however, has been completed on integrating microdynamic components into systems that satisfy mechanical tasks on macroscopic scales. In this paper, we describe microdynamic components needed to produce a surface which is actively deformable on local scales. In particular, we consider the design and demonstration of smart journal and thrust bearings capable of using embedded sensors and actuators to change dynamically the surface geometry. The ability actively to deform bearing surfaces allows for the design of bearings which are less prone to failure, the design of bearings with greater load-carrying abilities, and a fundamental study of the effect of surface geometries and fluid conditions on bearing performance, such as start-up and shut-down conditions. Some results of our new bearing designs are presented, focusing on numerical bearing models, sensor and actuator design and fabrication, and physical experimentation.     

A method of temperature monitoring in fluid film bearings

In the present paper a method to improve temperature monitoring of fluid film bearings is described. The method is tested in an industrial tilting pad thrust bearing. Test results for steady state and transient operating conditions are reported. Temperatures monitored by a thermocouple through the utilisation of the proposed method are compared to those measured by thermocouples in the pad backing and in the collar. The method is also tested on a PTFE-faced bearing. Test results show that the proposed method improves sensitivity to thermal transients in conventional babbitted bearings and provides adequate means of temperature monitoring in the PTFE-faced bearings.     

动车组转向架轴承局部损伤振动特性分析

针对动车组转向架轴承,根据轴承故障产生机理建立了轴承故障动力学工程模型。模型充分考虑了车轴弯曲刚度、轴承间隙及滚动体和滚道间的非线性接触力等因素,并包含内圈、外圈以及滚动体故障轴承动力学模型,使用龙格库塔数值积分方法进行了动力学仿真分析。针对实际轴承搭建实验台,对不同故障类型及不同程度故障进行了实际测试。仿真分析与实验结果吻合度较高,最大误差不超过5%,证明了该动力学模型的有效性。     

变工况下滚动轴承静电多传感器融合监测方法研究

滚动轴承是轨道车辆中最关键的部件之一,其质量好坏直接影响轨道车辆的运行.针对滚动轴承的故障监测,基于多传感器信息融合技术,采用多个静电传感器对滚动轴承进行同时监测,通过时域算法和复杂度度量算法对特征参数进行提取,并提出MWLOF检测算法.根据滚动轴承静电综合监测平台在不同工况下对滚动轴承进行监测试验,结果表明:考虑工况...     

Diagnostic Monitoring of Rolling-Element Bearings by High-Frequency Resonance Technique

In the present work, investigations by high-frequency resonance technique for diagnosis of defect frequencies of rolling-element bearings are reported. Raw vibration signature of the bearings at different speeds of operation has been demodulated. Envelope detected spectrum is analyzed to evaluate various defect frequencies and their energy levels. Experimentally evaluated frequencies are compared with theoretically determined defect frequencies. These frequency values and their energy levels are used to monitor intrinsic condition of bearings as well as to establish severity of existing/developed defects in the bearings. Relative comparison of bearings of the same type are made at various operating speeds under identical conditions of operation on the basis of identified defect frequencies and severity of defects.

The paper gives a realistic approach to monitor intrinsic condition of a bearing. Investigations given in the paper may have a potential for performance evaluation and may act as a reliable tool to establish safe limit for bearing operation. Investigations may serve as a quality control instrument for the earliest detection of even the smallest nature of defect existing in a bearing and may be used for “on-line” bearing condition monitoring as well as routine “quality control” instrument on the test bed for the diagnostic monitoring of rolling-element bearings.     

A comparison of some condition monitoring techniques for the detection of defect in induction motor ball bearings

Rolling element bearings are critical components in induction motors and monitoring their condition is important to avoid failures. Several condition monitoring techniques for the bearings are available. Out of these stator current monitoring is a relatively new technique. Vibration, stator current, acoustic emission and shock pulse methods (SPMs) for the detection of a defect in the outer race of induction motor ball bearing have been compared. The measurements were performed at different loads. The defect in the bearing could be detected by all the methods. Acoustic emission monitoring proved to be the best method followed by SPM when the increase in the levels of the measured parameters were compared with respect to those of healthy bearings.     

Smart machines with flexible rotors

The concept of smart machinery is of current interest. Several technologies are relevant in this quest including magnetic bearings, shape memory alloys (SMA) and piezo-electric activation. Recently, a smart bearing pedestal was proposed based on SMAs and elastomeric O-rings. However, such a device is clearly relevant only for the control of rigid rotors, for flexible rotors there is a need for some modification on the rotor itself. In this paper, rotor actuation by piezo-electric patches on the rotor is studied. A methodology is presented for the calculation of rotor behaviour and appropriate control strategies are discussed.     

滚动轴承内圈温度无线监测技术

温度是反映滚动轴承运行状态的重要指标之一,高速运行中轴承旋转套圈的温度远高于静止套圈温度,实现轴承旋转套圈（通常为内圈）的实时温升监测对保证运行安全尤为重要。基于电气和机械设计技术的协同,提出一种滚动轴承内圈温度无线监测方案,将测试系统嵌入轴承锁紧螺母中,利用锁紧螺母与轴承接触配合实现对轴承内圈温度采集。设计研究集成温度测试、无线数据传输及无线电能供给电路。重点实现电能无线供给以满足轴承旋转套圈温度持续监测的电能需求。最后通过对32008圆锥滚子轴承温度采集,验证该方案的可行性。该方法在保障轴承完整性的条件下能实现轴承旋转套圈温度的监测,为轴承状态监测和热特性试验研究提供了新的技术手段,解决了实际工况下轴承旋转套圈测温难题。