转子系统三维轴心轨迹和流形学习的故障诊断方法
Fault Diagnosis Method of 3D Axis Orbit and Manifold Learning for Rotor System

提出一种基于三维轴心轨迹和流形学习的故障诊断方法。提取转子系统水平、竖直和轴向的位移信号，采用EEMD分解对原始信号进行降噪，将降噪后的信号合成三维轴心轨迹，采用LTSA流形学习算法对三维轴心轨迹进行降维得到其二维流形图。相较于三维轴心轨迹，降维后的二维流形图更方便分析与识别，并且保留了三维轴心轨迹各数据点的空间拓扑关系。应用该方法进行试验，获取转子系统的正常、不对中、油膜涡动、油膜振荡的三维轴心轨迹及其降维后的二维流形图。利用LTSA算法得到的二维流形图相比于三维轴心轨迹具有简单直观的特征区分。
A fault diagnosis method based on 3D axis orbit and manifold learning is proposed. The horizontal, vertical and axial displacement signals of rotor system are extracted and the ensemble empirical mode decomposition (EEMD) is used to denoise the original signal. The denoised signals are combined into 3D axis orbit. Locally tangent space arrangement (LTSA) manifold learning algorithm is used to reduce the dimension of the 3D axis orbit, thus to obtain its 2D manifold diagram. Compared with the 3D axis orbit, the 2D manifold diagram is more convenient for analysis and recognition, and the spatial topological relation of the data points of the 3D axis orbit is retained. The method is applied to the test, 3D axis orbit and 2D manifold diagram of the normal, misalignment, oil whirl and oil whip in the rotor system are obtained. Compared with the 3D axis orbit, the 2D manifold diagram has simple and intuitive feature differentiation