分形与几何特征融合的转子故障特征提取方法

转子系统故障信号是典型的非线性、非平稳信号，分形几何为描述转子系统故障信号的特性提供了一个分析工具，但仅仅依靠分形维数无法有效的提取转子系统的故障特征。本文引入紧密度和丰度两个量，与基于的分形维数一起，对转子系统故障信号进行分析；最后采用神经网络技术对转子系统的正常、不对中、不平衡、碰磨、松动五种不同的运行状态进行分类识别。实验结果表明，通过对分形维数和紧密度、丰度的联合可较好地评定和区分转子系统的运行状态。

Fault Feature Extraction Method of Rotor System Based on Combination of Fractal and Geometrical Characteristics

Fault signal of rotor systems is a typical nonlinear and non-stationary signal. Fractal geometrical theory is an efficient tool to characterize the feature of the signal. However, relying on fractal dimension solely cannot effectively extract the fault feature of the rotor system. Therefore, compactness and abundance are proposed and used along with the mathematical-morphology-based fractal dimension to analyze the fault signal of the rotor system. Finally, signals acquired from 5 different types of the rotor system conditions, such as normal, misalignment, imbalance, contact and loose, are identified and classified by using neural network technique. The operation status of the rotor system can be assessed by the combination of the mathematical-morphology-based fractal dimension, compactness and abundance. Experimental results reveal that this method has a good capability for identification and classification of the fault signals of rotor systems.