多重分形信号特征提取及在往复压缩机故障诊断中的应用

发布时间：2018-03-27 17:28

本文选题：MF-DFA　切入点：IKNNModel　出处：《东北石油大学》2015年硕士论文

【摘要】：往复压缩机广泛应用于石油化工等工业生产中,并成为工业装置中的核心设备。目前,往复压缩机状态监测与故障诊断技术,已成为国内外故障诊断领域的研究热点问题,而特征提取又是其故障诊断技术中的难点之一。针对具有非线性非平稳特性的往复压缩机振动信号,本文提出了运用多重分形的方法,对其进行局部奇异特性量化描述,并采用了新的故障识别方法进行故障分类识别,结果表明,此方法能够有效地对往复压缩机常见故障进行诊断,并为往复压缩机故障诊断研究提供了新的思路。首先,通过查阅大量文献资料,对目前往复压缩机的故障诊断技术研究现状进行了综述,提出了本文的研究构想。在对往复压缩机的基本结构、工作原理进行简要概述的基础上,着重对主要零部件的常见故障形式及其故障机理进行了深入分析。其次,针对往复压缩机振动信号的非线性特性,本文提出采用多重分形的方法对其进行故障特征提取,主要介绍了分形和多重分形理论及基本算法,在此基础上,提出多重分形去趋势波动分析(MF-DFA)理论,并对往复压缩机气阀故障振动信号进行分析,验证了MF-DFA方法在往复压缩机振动信号特征提取方面的有效性。而后,对于采用MF-DFA方法,提取出的往复压缩机故障特征向量参数而言,选用一般常见的故障识别方法,存在故障识别率低模式状态相互混淆难以区分等问题,引用了一种新的故障识别方法——增量学习KNNModel算法(IKNNModel)。通过与其他故障识别方法比较,证明了此方法在往复压缩机故障分类识别方面具有明显的优越性。最后,将多重分形去趋势波动分析理论和增量学习KNNModel算法结合起来,从往复压缩机故障特征提取与识别的角度出发,详细阐明了往复压缩机振动信号数据预处理故障特征提取故障特征参数优选和故障特征识别等关键问题,提出了一套完整的基于MF-DFA与IKNNModel算法的往复压缩机故障诊断方法。对2D12型往复压缩机常见故障实测数据的分析表明,该方法可准确提取故障特征,且可准确识别故障类型,为往复压缩机故障诊断提供了新的研究途径。
[Abstract]:Reciprocating compressors are widely used in petroleum? At present, the state monitoring and fault diagnosis technology of reciprocating compressor has become a hot issue in the field of fault diagnosis at home and abroad. Feature extraction is one of the difficulties in fault diagnosis. The vibration signals of reciprocating compressors with non-stationary characteristics are described quantitatively by multifractal method, and a new fault identification method is used to classify and identify the faults. This method can effectively diagnose the common faults of reciprocating compressors and provide a new idea for the research of fault diagnosis of reciprocating compressors. The present situation of fault diagnosis technology of reciprocating compressor is summarized, the research conception of this paper is put forward, and the basic structure and working principle of reciprocating compressor are briefly summarized. The common fault forms and fault mechanism of main parts are analyzed in detail. Secondly, in view of the nonlinear characteristics of vibration signals of reciprocating compressors, a multifractal method is proposed to extract the fault features of reciprocating compressors. This paper mainly introduces the fractal and multifractal theory and its basic algorithm. On this basis, the multifractal detrend fluctuation analysis theory and MF-DFAtheory are put forward, and the vibration signal of reciprocating compressor valve fault is analyzed. The validity of MF-DFA method in vibration signal feature extraction of reciprocating compressor is verified. Then, for the parameters of fault eigenvector extracted from reciprocating compressor by MF-DFA method, the common fault identification method is used. Is there a low rate of fault identification? In this paper, a new method of fault identification, incremental learning KNNModel algorithm, is proposed, which is difficult to distinguish the pattern states from each other. By comparing with other fault identification methods, an incremental learning KNNModel algorithm is proposed. It is proved that this method has obvious advantages in fault classification and identification of reciprocating compressors. Finally, the theory of multifractal de-trend fluctuation analysis and incremental learning KNNModel algorithm are combined. From the point of view of fault feature extraction and identification of reciprocating compressor, the data preprocessing of vibration signal of reciprocating compressor is expounded in detail. Fault feature extraction? In this paper, a complete method of fault diagnosis for reciprocating compressor based on MF-DFA and IKNNModel algorithm is proposed. The analysis of common fault data of 2D12 reciprocating compressor shows that, This method can accurately extract fault features and identify fault types, which provides a new approach for fault diagnosis of reciprocating compressors.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TH45

【相似文献】