基于信息熵的涡旋压缩机的故障诊断研究

发布时间：2019-04-01 17:13

【摘要】：近年来基于涡旋压缩机的测试系统的平台搭建和对其振动噪声的分析，展示出了非常广阔的应用前景，但还是难以满足涡旋压缩机运行状态监测的要求。由于涡旋压缩机在国内的应用时间还不是很长,对其使用过程中的故障分析还不多，仍处于搭建测试平台采用传统单一的频谱分析描述其工作状态的初级阶段，其运转状况的性能数据还无法通过准确的数学模型计算获得。而且涡旋压缩机振动的激励源较多，，壳体表面信号表现为非平稳性和非线性，因此对其故障诊断较为复杂。本文在常规频谱分析的基础上，通过多角度的信息融合更准确的对非平稳信号进行了故障判别。本文从振动信号分析的思路出发,结合信息论中熵和灰关联度的理论，建立了一种基于时域的奇异谱熵、频域的功率谱熵、时-频域小波能量谱熵和小波空间特征谱熵新的的故障诊断方法并以此作为综合评价涡旋压缩机振动状态的定量特征指标。对其中几处难点作了重点分析，一是关于奇异谱熵的嵌入延迟理论的参数选取问题，这直接关乎于奇异分解对于信号的有效信息和噪声的区分效果。二是对传统概率熵加入了滑动时窗进行了改进，使其功率谱熵和小波能谱熵体现了信号局部特征的分布差异和变化。在MATLAB信号处理工具箱里利用信息熵的几种算法实现了对奇异谱熵、功率谱熵、小波能谱熵和小波特征谱熵这四种熵的数学模型建立。结合整理出的涡旋压缩机的不同转速下的熵值参考样本与待诊样本进行了灰关联对比分析，直接利用灰关联度的量化结果实现了对涡旋压缩机几种故障的很好识别,证明了该故障诊断方法的有效性。
[Abstract]:In recent years, the platform of the test system based on scroll compressor and the analysis of its vibration and noise have shown a very broad application prospect, but it is still difficult to meet the requirements of monitoring the running state of scroll compressor. Because the application time of scroll compressor in China is not very long, the fault analysis of scroll compressor in the process of its use is not much, and it is still in the initial stage of building test platform to describe its working state by traditional single spectrum analysis. The performance data of its operation condition can not be obtained by accurate mathematical model calculation. Moreover, the vibration source of scroll compressor is more, the signal of shell surface is non-stationary and non-linear, so the fault diagnosis of scroll compressor is more complicated. On the basis of conventional spectrum analysis, this paper uses multi-angle information fusion to distinguish the fault of non-stationary signal more accurately. Based on the idea of vibration signal analysis and the theory of entropy and grey correlation degree in information theory, a singular spectral entropy based on time domain and power spectral entropy in frequency domain is established in this paper. A new fault diagnosis method based on time-frequency domain wavelet energy spectrum entropy and wavelet spatial characteristic spectrum entropy is proposed and used as a quantitative characteristic index for comprehensive evaluation of the vibration state of scroll compressor. Several difficulties are analyzed in this paper. One is the parameter selection of the embedding delay theory of singular spectral entropy, which is directly related to the discrimination effect of singular decomposition on the effective information and noise of the signal. Secondly, the traditional probability entropy is improved when the sliding window is added, so that the power spectrum entropy and the wavelet spectrum entropy reflect the distribution difference and change of the local characteristics of the signal. The mathematical models of singular spectral entropy, power spectral entropy, wavelet energy spectrum entropy and wavelet characteristic spectral entropy are established by using several algorithms of information entropy in MATLAB signal processing toolbox. Combined with the entropy reference samples of the scroll compressor at different speeds and the samples to be diagnosed, the grey correlation analysis is carried out, and the quantitative results of the grey correlation degree are directly used to realize the good identification of several faults of the scroll compressor. The validity of this fault diagnosis method is proved.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH45;TN911.7;O236

【参考文献】