基于自适应线调频基原子分解方法的机械故障诊断研究

发布时间：2018-01-19 00:07

本文关键词： 自适应线调频基原子分解时域同步平均循环频率神经网络齿轮滚动轴承故障诊断　出处：《湖南大学》2011年硕士论文　论文类型：学位论文

【摘要】：生产过程中发生的设备故障会导致设备停机或机器损坏，进而导致生产中断，因而在制造业尤其是流程制造业中，对机械设备的状态监测与故障诊断具有重要的理论意义和实用价值。从机械设备的故障振动信号中提取故障特征信息，是机械设备故障诊断的关键。在变转速工况下，机械设备的振动信号往往包含了更多的设备运转信息和故障信息，系统缺陷能更容易地被发现。然而以等采样频率采集的振动信号在转速波动工况下往往表现出强烈的非平稳性和低信噪比性，导致目前常用的信号处理技术无法从中准确提取故障特征信息。本文在国家高技术研究发展计划(863计划)项目“大型风力发电机组状态监控与故障诊断技术研究”(项目编号：2009AA04Z414)和国家自然科学基金项目“多尺度线调频基稀疏信号分解方法及其在机械故障诊断中的应用研究”(项目批准号：50875078)资助下，针对现有信号处理方法时频聚集性不够，抗噪性能不强等缺点，研究提出了一种新的信号处理方法—自适应线调频基原子分解(adaptive chirplet atomicdecomposition, ACAD)方法，并将其应用于转速变化的齿轮和滚动轴承的故障诊断中。本文的主要研究工作有： (1)针对多尺度线调频基稀疏信号分解方法算法效率低、分解分量幅值失真等问题，研究提出了ACAD方法，并证明了该方法具有良好的分解精度、较好的抗噪性能和较高的分解效率，非常适合于多分量非平稳信号的分析处理。 (2)针对变转速工况下低信噪比的故障齿轮振动信号调制边频带难以识别的问题，研究提出了基于ACAD的时域同步平均方法。ACAD方法可以有效地提取齿轮的啮合频率曲线，从而获得齿轮转速曲线，再对振动信号进行角域重采样，可满足时域同步平均对信号平稳性要求。仿真和实验分析证明了基于ACAD的时域同步平均方法能清晰获取齿轮的故障调制阶次，非常适合于转速剧烈波动情况下的齿轮故障诊断。 (3)提出了基于ACAD阶次包络和循环频率的变转速齿轮故障诊断方法。包络谱和循环频率分析方法是一种有效的齿轮幅值和相位调制频率提取方法，但在变转速工况下齿轮振动信号往往表现出剧烈的非平稳性，由于故障而产生的调制频率成分也会随着转速变化而变化，不满足FFT对信号的平稳性要求，包络谱和循环频率分析无法提取齿轮的故障信息。基于ACAD阶次包络和循环频率的变转速齿轮故障诊断方法先利用ACAD从齿轮振动信号中提取啮合频率，从而获得齿轮转速曲线，根据获得的转频曲线再对原始信号进行角域重采样。对重采样信号进行Hilbert变换分别提取其包络和相位。对包络信号进行FFT变换获取幅值调制频率，对相位信号进行循环频率获取相位调制频率，从而实现齿轮的故障诊断。 (4)将ACAD方法与神经网络结合应用于变转速工况下滚动轴承的故障识别。采用ACAD方法从滚动轴承振动信号的包络中提取故障特征频率及其倍频分量，再从这些特征故障分量中提取能量、方差等时域特征参数作为神经网络的输入参数来识别滚动轴承的故障模式。应用实例证明该方法可以准确有效地对滚动轴承的工作状态和故障类型进行分类。本文研究了适合处理多分量非平稳信号的ACAD方法，并在其基础上提出了基于ACAD的时域同步平均方法、基于ACAD的阶次包络和循环频率方法和基于ACAD的神经网络方法，，这些方法能有效应用于变转速工况下齿轮和滚动轴承的故障诊断。仿真算例和应用实例表明，ACAD方法在机械故障诊断中具有良好的应用前景。
[Abstract]:Equipment failure in the production process will lead to downtime or damage, leading to production disruptions, and especially in the manufacturing process in the manufacturing industry, and has important theoretical significance and practical value of state monitoring and fault diagnosis for mechanical equipment. The fault feature extraction from vibration signals of the mechanical equipment is the key fault diagnosis of mechanical equipment.
Under variable speed condition, the vibration signals of mechanical equipment often contains more equipment operation and fault information system defects can more easily be found. However, as the sampling frequency acquisition of vibration signal in the speed fluctuation conditions often show a strong non stationarity and low signal-to-noise ratio, leading to the current signal the commonly used processing technology is unable to accurately extract the fault feature information. Based on the national high technology research and development program (863 Program) project "large wind turbine condition monitoring and fault diagnosis technology research" (project number: 2009AA04Z414) and the application of the National Natural Science Fund Project "Multi-scale Chirplet and sparse signal decomposition method and in mechanical fault diagnosis" (Project No.: 50875078) supported by the existing signal processing method of time-frequency anti noise performance is not strong enough, etc. A new signal processing method, adaptive adaptive chirplet atomicdecomposition (ACAD), is proposed and applied to fault diagnosis of gear and rolling bearings with variable speed.
(1) based on Multi-scale Chirplet and sparse signal decomposition algorithm low efficiency problem decomposition component amplitude distortion, the research put forward the ACAD method, and proves that this method has good accuracy of decomposition, the decomposition efficiency and better anti noise performance, very suitable for the analysis on the multi-component non-stationary signals.
(2) aiming at the fault of gear vibration signal modulation variable speed under the condition of low SNR sidebands are difficult to identify problems, put forward the research method of ACAD.ACAD synchronous averaging method based on time domain can effectively extract the meshing frequency curve of gear, gear and speed curve is obtained, and then the vibration signal of angle domain resampling, can to meet the time synchronous average signal stationarity. Simulation analysis and experimental results show that the ACAD based on the time-domain synchronous average method can clearly obtain the gear fault modulation order, gear fault diagnosis is very suitable for the circumstance of drastic speed fluctuation.
(3) the variable speed gear fault diagnosis method of ACAD order envelope and cycle frequency. Based on the envelope spectrum and frequency analysis method is an effective method for detection of circular gear amplitude and phase modulation frequency, but in the condition of variable speed gear vibration signals often exhibit non-stationary intensity, modulation frequency components the fault will also change with the speed change, does not meet the FFT stationary signal, envelope spectrum and cycle frequency analysis to extract fault information of the gear. The ACAD order envelope and cycle frequency variable speed gear fault diagnosis method based on the first use of ACAD extracted from the vibration signal of the gear meshing frequency, thus the gear speed curve is obtained, according to the rotation frequency of the obtained curve and the original signal is resampled in angle domain. The sampling signal Hilbert transform were extracted from the envelope and phase of the envelope. The amplitude modulation frequency is obtained by the FFT transform, and the phase modulation frequency is obtained by the cycle frequency of the phase signal, thus the fault diagnosis of the gear is realized.
(4) the ACAD fault identification method and neural network combined with the application of rolling bearing in variable speed conditions. Extracting fault characteristic frequency and frequency component from the envelope of the rolling bearing vibration signal using ACAD method to extract energy from these characteristics of fault component, variance domain parameters as input parameters of neural network to identify fault pattern of rolling bearing. The application example proves that the method can effectively and accurately classify the working condition of roller bearings and fault types.
In this paper the ACAD method for processing the multi-component non-stationary signals, and proposed a ACAD based on the time-domain synchronous average method on the basis of it, based on the order envelope and cycle frequency and method of ACAD based on ACAD neural network method, these methods can be effectively applied to variable speed under the condition of gear and rolling bearing fault the diagnosis. The simulation and application examples demonstrate that the ACAD method has a good application prospect in mechanical fault diagnosis.

【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH165.3

【相似文献】