Lyapunov指数在异步电动机机械振动识别中的应用  

Application of Lyapunov exponent in mechanical vibration identification of asynchronous motor

作　　者：刘雁[1] 高宽 黄炎 张赫 肖军[2] LIU Yan;GAO Kuan;HUANG Yan;ZHANG He;XIAO Jun(School of Mechatronic Engineering,Northwestern Polytechnical University,Xi’an 710072,China;State Key Lab of Compressor Technology,Hefei General Machinery Research Institute Co.,Ltd.,Hefei 230031,China)

机构地区：[1]西北工业大学机电学院,西安710072 [2]合肥通用机械研究院压缩机技术国家重点实验室,合肥230031

出　　处：《振动与冲击》2022年第13期142-151,共10页Journal of Vibration and Shock

基　　金：国家自然科学基金(51775437);压缩机技术国家重点实验室开放基金项目(SKL-YSJ201902);国家重点研发计划(2019YFB1504601)。

摘　　要：基于非线性动力学理论,研究异步电动机振动信号的Lyapunov指数特征,并应用于故障诊断和识别。首先,搭建试验平台,并模拟异步电动机正常工作、转子不对中和底座安装不良的三种转动状态。分析电动机三种振动信号的波形,并进行去噪和预处理。然后,基于BBA算法计算振动信号在不同工作状态下的Lyapunov指数谱,选取最大Lyapunov指数作为特征用于识别异步电动机的机械振动;最后,为了对该分析方法的有效性及抗干扰性进行验证,引入了随机噪声,分析所提出算法在不同参数下的受噪声的影响水平。研究结果显示,异步电动机在正常运行时,其最大Lyapunov指数值在0.3~0.7;安装不良时其最大Lyapunov指数值在0~0.3,表明这两种工作状态下电动机振动信号序列出自于一个混沌过程;在电动机处于转子不对中状态时,其最大Lyapunov指数值近似为零,表明其振动序列中基本不存在混沌属性。在该研究结果的基础上,配合特征融合与机器学习分类算法,将有效提高异步电动机机械振动识别的准确率和效率。Here,based on the theory of nonlinear dynamics,Lyapunov exponent characteristics of an asynchronous motor vibration signals were studied and applied in fault diagnosis and identification.Firstly,a test platform was built to simulate 3 rotating states of the asynchronous motor including normal operation,rotor misalignment and poor base installation.Waveforms of the motor’s 3 kinds of vibration signals were analyzed,denoised and preprocessed.Then,Lyapunov exponent spectra of vibration signals under different working states were calculated using BBA algorithm,and the maximum Lyapunov exponent was selected as the feature to identify mechanical vibration of the motor.Finally,to verify the effectiveness and anti-interference of the analysis method,random noise was introduced to analyze influence levels of noise on BBA algorithm under different parameters.The study results showed that the maximum Lyapunov exponent of the motor is in the range of 0.3-0.7 during normal operation,the maximum Lyapunov exponent is in the range of 0-0.3 when the motor is not installed properly,so the motor vibration signal sequences under these two working conditions come from a chaotic process;when the motor is in the state of rotor misalignment,its maximum Lyapunov exponent is approximately zero,so there is basically no chaotic property in its vibration sequences;based on the study results,combined with feature fusion and machine learning classification algorithm,the accuracy and efficiency of mechanical vibration recognition of asynchronous motor can be effectively improved.

关 键 词：异步电动机 故障振动 LYAPUNOV指数 Brown Bryant Abarbanel(BBA)算法 

分 类 号：TM343[电气工程—电机] TK312[动力工程及工程热物理—热能工程]