面向全寿命过程的低速斜齿轮故障诊断方法研究

发布时间：2018-05-09 04:02

本文选题：低速斜齿轮 + 早期故障　；参考：《哈尔滨工业大学》2014年博士论文

【摘要】：低速重载斜齿轮是工业设备、风力发电机和舰船等大型机械装备普遍采用的一种机械传动形式。由于工作环境恶劣，长期在重载条件下运行，斜齿轮极易受到损害而出现故障，造成机械传动系统失效，进而影响整个机械装备的安全可靠运行。因此，本文对低速斜齿轮故障的产生、发展，直至失效的全寿命过程进行故障诊断方法研究，实现对不同失效阶段的斜齿轮故障特征提取与故障模式识别，为齿轮箱的视情维修和寿命预测提供决策依据。针对低速斜齿轮故障振动信号具有非平稳、非线性和多分量的特点，本文提出了基于局部时间尺度分解的时频分析方法，并分别对其端点效应的抑制措施、基线信号的平滑处理以及分解结果的瞬时幅值和瞬时频率求解方法进行研究。局部时间尺度分解方法能够依据信号自身的特征信息自适应地将一个复杂的非平稳信号分解为若干个瞬时频率具有物理意义的本征时间尺度分量之和，而且各分量能够如实地反映信号从高频到低频各个频段范围的组成成分。针对低速斜齿轮故障振动信号存在较强的噪声干扰，本文提出了基于噪声辅助分析的局部时间尺度分解降噪方法，该方法在定义了随机噪声分量的判定条件的基础上，通过对信号不断分解、筛选，可以将斜齿轮故障振动信号分解为若干个噪声分量与剩余信号之和的形式，进而可以将噪声分量从原始信号中分离，提高信号的信噪比。针对低速斜齿轮早期故障特征十分微弱、难以提取的问题，提出了基于振动信号局部能量放大的故障特征提取方法，该方法首先对斜齿轮故障振动信号进行分段，再对分段信号进行能量集中和放大，进而达到突出故障特征的目的。将基于噪声辅助分析的局部时间尺度分解降噪方法和振动信号局部能量放大方法相结合用于平稳转速条件下的低速斜齿轮早期故障特征提取。齿轮箱在工作过程中，由于载荷变化引起转速随机波动，会造成时域采样信号出现较强的非平稳性，直接进行时频分析会出现频率混叠的现象。本文采用光电编码器作为数据采集板的外触发器，对齿轮故障振动信号进行等角度采样，将时域内的非平稳振动信号转化为角度域内的平稳信号。将时域同步平均方法引入角度域内，提出了基于角度域同步平均降噪的方法，并将该方法与阶次分析相结合，实现对转速随机波动的低速斜齿轮早期故障特征提取。提出了基于角度域振动信号有效值分布的故障轮齿定位诊断方法，通过计算各个轮齿对应的降噪后的角度域故障振动信号的有效值分布情况，，确定故障轮齿的位置。斜齿轮故障数据样本难以获取，现有的支持向量机等小样本故障模式识别方法利用的故障特征信息少、分类精度低。本文建立了以时变函数为输入量的过程支持向量机分类模型，通过正交基函数展开的方法将离散的斜齿轮故障特征信号拟合为连续的时变函数，并将其作为过程支持向量机分类模型的输入特征量，充分利用有限的数据样本中包含的故障特征信息，提高了故障识别方法的准确性。将斜齿轮全寿命过程故障模式识别的多分类问题转化为若干个两分类问题组合的形式，提出了基于过程支持向量机的斜齿轮全寿命过程模式识别方法，实现了对斜齿轮全寿命过程中不同故障模式的准确识别。
[Abstract]:Low - speed heavy - duty helical gears are widely used in industrial equipment , wind - driven generators , ships and other large - scale mechanical equipment . Because of the poor working environment and long - term operation under heavy - load conditions , the bevel gears are extremely vulnerable to damage , which can lead to failure of mechanical transmission systems and thus affect the safe and reliable operation of the whole mechanical equipment . Therefore , this paper studies the fault feature extraction and fault pattern recognition of inclined gears in different failure stages , and provides the basis for decision - making for the case maintenance and life prediction of the gearbox .

In this paper , a time - frequency analysis method based on local time - scale decomposition is proposed for low - speed helical gear fault vibration signals , and the methods of suppressing the end effects , smoothing processing of baseline signals and instantaneous amplitude and instantaneous frequency solutions of the decomposition results are studied . The local time scale decomposition method can adaptively decompose a complex non - stationary signal into the sum of the intrinsic time scale components of physical meaning based on the characteristic information of the signal itself , and the components can faithfully reflect the components of the signal from high frequency to low frequency range .

On the basis of noise - assisted analysis , this paper proposes a method of noise - assisted analysis based on local time - scale decomposition and noise reduction , which can separate the signal from the original signal and improve the signal - to - noise ratio of the signal .

In the process of working , due to the random fluctuation of the rotating speed caused by the change of load , the phenomena of frequency aliasing can occur in the time domain sampling signal . In this paper , a photoelectric encoder is used as the outer trigger of the data acquisition board , and the unstable vibration signal in the time domain is converted into a stable signal in the angular domain .

This paper sets up a process support vector machine classification model with time - varying function as the input quantity , and makes full use of the fault feature information contained in the finite data sample as the input feature quantity of the process support vector machine classification model , and makes full use of the fault feature information contained in the limited data sample to improve the accuracy of the fault identification method .

【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TH132.41;TH165.3

【参考文献】