面向时变传递路径的行星齿轮箱耦合振动特征提取及故障仿真建模

Planetary gearboxes are widely used in complex mechanical systems. However, planetary gearbox has some unique characteristics. For example, a planetary gearbox is composed of many rotating parts and components, many vibration sources are located in a planetary gearbox and the revolution of planetary gears makes vibration transmission paths from the vibration sources to the sensor time-varying, etc., which lead the observed vibration is complex and difficult for analysis. Up to now, the methods for the characteristics extraction and faults simulation modeling are still not studied enough. To address this issue, the characteristic extraction methods and faults simulation modeling of planetary gearboxes will be well studied in this project. The main contents included in this project are listed as follows. 1) Feature extraction under time-varying transmission path. By combined the windowed synchronous averaging in the angle domain with some modern signal processing methods, the characteristic extraction methods suit for the time-varying transmission paths of planetary gearboxes will be studied. 2) Weak characteristics detection method. For enhancing the weak signal and removing the transmission path effect, weak characteristics detection methods for the incipient faults of gears or planetary bearings will be studied based on the adaptive envelope extraction method and the minimum entropy deconvolution (MED) algorithm. 3) Multi-source coupling signal separation method. Based on blind sources separation and dimensions reduction theories, multi-source coupling signal separation methods under limited number of sensors and time-varying vibration sources mixing system will be studied. 4) Faults simulation modeling. Based on the effect on the gear meshing stiffness by gear crack etc., the sudden change on the phase of the signal caused by the fault is added. In this way, the simulation model can be established more completely to reveal the characteristics. As a result, the characteristic extraction methods for planetary gearboxes and the faults simulation modeling theory can be achieved .

行星齿轮箱广泛用于复杂机械装备，其零部件繁杂、振源多，尤其因行星轮公转引起的时变振动传递路径使振动响应复杂，对应的特征提取方法和故障仿真模型亟待完善。课题针对其时变传递路径下的故障检测难题，开展特征提取和故障仿真建模研究，内容包括：1）时变传递路径下的特征提取。基于角域加窗同步平均等信号处理方法，研究适合时变传递路径及变速工况的行星齿轮箱特征提取方法；2）弱特征信号检测。结合最小熵解卷积弱冲击特征增强和传递路径抑制理论，以基于谱峭度的自适应包络提取法研究行星轴承等早期故障弱特征信号检测方法；3）多源耦合振动分离。基于盲源分离、维数缩减等理论，研究有限传感器数量和时变振源混合系统条件下的多源耦合振动分离方法；4）故障仿真建模。基于齿根裂纹等对齿轮啮合刚度影响，在现有模型基础上融入故障对应信号相位突变特征，建立较为完备的故障振动仿真模型。最终获得适合行星齿轮箱的振动特征提取方法及故障建模理论。

行星齿轮箱作为重要的动力传递单元广泛用于风力机组、直升机等复杂机械装备中。由于行星轮公转引起的齿轮啮合振源与固定安装传感器间的振动传递路径存在时变性和多个行星轮同时与齿圈及太阳轮啮合，导致传感器获得的振动响应复杂，传统定轴齿轮箱故障检测方法不能直接应用，对应的特征提取方法和故障仿真模型亟待发展完善。为此，项目开展了针对行星齿轮箱的故障特征提取方法和仿真建模研究，内容包括：1）时变传递路径下的特征提取方法；2）弱特征信号检测；3）多源耦合振动分离；4）故障仿真建模。. 项目已取得以下成果：1) 改进加窗振动分离法。先按原方法将拾取的振动信号在对应离传感器最近位置通过加窗截取得到多齿宽对应数据段，在此基础上创造性提出按齿宽对应数据长度将数据段分为各齿对应数据块，按齿序对相同啮合位置数据块进行拼接，准确构造消除传递路径影响的多个人工合成齿轮振动并由同步平均消除非同步干扰成分，最后由最大峰峰值选取与真正啮合位置对应的人工合成齿轮振动。2）针对弱特征检测，提出基于MOMEDA的行星轮齿根裂纹故障特征增强提取方法；提出基于SANC、倒谱预白化、Autogram等行星轮轴承故障弱特征提取方法。3）针对多源耦合振动分离，将提出的改进加窗振动分离法与同步平均、窄带解调等方法结合，提出了适合太阳轮、行星轮故障检测的多种方法；提出了基于卷积神经网络、长短时记忆神经网络的故障分类方法。4）针对故障仿真建模，提出了一种适用于加窗振动分离的行星轮局部故障振动仿真模型；目前正在撰写行星架裂纹故障仿真模型论文。. 本项目研究已完成原预定目标，提出了有效解决时变传递路径的改进加窗振动分离方法、提出了行星轮、太阳轮、行星架等局部故障的检测新方法，发展了行星轮、行星架等故障的振动仿真模型，研究中产生了行星齿轮箱故障检测新知识，发展了特征提取、检测和仿真等新方法，取得研究成果对后续开展该领域的相关研究和工程应用有重要价值。