基于LMD时频分析的旋转机械故障特征提取方法研究

发布时间：2018-04-06 06:17

本文选题：LMD分解　切入点：故障特征提取　出处：《燕山大学》2013年硕士论文

【摘要】：故障诊断是研究设备运行状态信息的变化，进而识别其运行状态的科学，随着现代化工业生产的不断发展，机械设备故障诊断技术得到了广泛的关注。在故障诊断中，，机械振动信号的故障特征提取是其研究的瓶颈，信号的处理和分析是特征提取最常用的方法。在实际应用中，机械振动信号都是非平稳的，对这类信号进行准确地分析和处理，是特征提取成功与否的关键所在。近年来，信号处理的时频分析方法在机械故障诊断领域得到了迅速的发展。传统以Fourier变换为基础的信号分析方法在非平稳信号处理方面不能达到理想效果。本文重点研究基于局部均值分解(LMD)的时频分析方法及其在旋转机械故障特征提取方面的应用。首先，研究瞬时频率、单分量和多分量信号、调频调幅信号的基本概念；分析LMD的基本原理及算法，并与经验模态分解(EMD)进行对比；阐述瞬时频率的求取方法：“直接法”、基于LMD的Hilbert变换法、能量算子解调法，分析各方法不足，在此基础上提出三点对称差分能量算子解调法，仿真验证该方法抑制LMD端点效应的有效性。其次，分析噪声对LMD的影响，针对模态混叠，研究小波半软阈值去噪和LMD相关度降噪技术，将两者结合提出一种故障特征提取的方法，采用理论分析和机械故障信号相结合验证上述方法在改善时频分析性能方面的有效性。最后，研究时频熵概念，提出局域时频熵理论，将其与LMD结合提取旋转机械故障特征，并进行了仿真实验的验证。
[Abstract]:Fault diagnosis is a science to study the change of equipment operation state information and then identify its running state. With the development of modern industrial production, the technology of mechanical equipment fault diagnosis has been paid more and more attention.In fault diagnosis, fault feature extraction of mechanical vibration signal is the bottleneck of its research, and signal processing and analysis is the most commonly used method of feature extraction.In practical applications, mechanical vibration signals are non-stationary, and accurate analysis and processing of these signals is the key to the success of feature extraction.In recent years, time-frequency analysis of signal processing has been rapidly developed in the field of mechanical fault diagnosis.The traditional signal analysis method based on Fourier transform can not achieve ideal results in non-stationary signal processing.This paper focuses on the time-frequency analysis method based on local mean decomposition (LMD) and its application in fault feature extraction of rotating machinery.First of all, the basic concepts of instantaneous frequency, single component and multi-component signals, frequency modulation signal are studied, the basic principle and algorithm of LMD are analyzed, and compared with empirical mode decomposition (EMD), the method of obtaining instantaneous frequency is described as "direct method".Based on LMD's Hilbert transform method and energy operator demodulation method, the deficiency of each method is analyzed. Based on this, a three-point symmetric differential energy operator demodulation method is proposed. The simulation results show that the proposed method is effective in suppressing the LMD endpoint effect.Secondly, the effect of noise on LMD is analyzed. Aiming at modal aliasing, wavelet semi-soft threshold de-noising and LMD correlation de-noising are studied, and a fault feature extraction method is proposed.Theoretical analysis and mechanical fault signal are used to verify the effectiveness of the proposed method in improving the performance of time-frequency analysis.Finally, the concept of time-frequency entropy is studied, and the local time-frequency entropy theory is proposed, which is combined with LMD to extract the fault features of rotating machinery.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TH165.3

【参考文献】