基于基追踪、波形匹配和支持向量机的往复式压缩机气阀故障诊断研究
发布时间：2018-08-07 12:08
【摘要】：往复压缩机广泛应用于石油和化工行业,在生产线上往往属于关键设备,其工作状态的好坏直接影响着企业的经济效益。其结构复杂且很多重要零部件工作在高温高压的恶劣环境下,并承受着往复载荷的作用,因而故障容易发生。与往复压缩机中其它零部件相比,气阀最脆弱,故障发生更频繁。从故障诊断、事故预防、维修决策和节约成本的角度来说,研究有效准确的往复压缩机气阀故障诊断方法是非常重要和有意义的。 往复压缩机的振动信号含有大量的周期成分和瞬态冲击成分,有明显的非平稳特征。其各零部件,如活塞、连杆、气阀等运动周期相同,频率特征在频谱上是重叠的,因此很难分辨。本文首先从往复压缩机的运动机理出发建立了气阀振动信号模型。然后,综合信号处理和模式识别技术,提出了一套新的往复压缩机气阀故障诊断方法。该方法结合了基追踪,波形匹配和支持向量机三种算法,从时域上对气阀状态进行识别和故障诊断。基追踪用来提取振动信号的主要成分和抑制背景噪声。波形匹配是本文提出的一种新的特征提取算法。传统的特征提取算法大多数通过统计特征和熵等指标来提取特征。与此不同的是,波形匹配通过把振动波形和参数化波形做匹配来提取特征。匹配过程由差分进化算法来优化。波形匹配提取的特征的维数小且各特征含有明确的物理意义。支持向量机适合处理小样本问题,在本文中用来实现故障识别和分类。在理论介绍后,本文将提出的方法应用于实验数据和现场数据。实验数据处理结果表明该诊断方法可以准确可靠地识别往复压缩机气阀的三种状态(正常,弹簧恶化和阀片变形)。现场数据结果表明该诊断方法可以有效地识别气阀故障程度。最后,本文对提出的新方法的优点和局限性进行了讨论,并且对在该方法基础上的进一步研究进行了展望。
[Abstract]:Reciprocating compressors are widely used in petroleum and chemical industry, and often belong to the key equipment in the production line. The working state of reciprocating compressors directly affects the economic benefits of enterprises. Its structure is complex, and many important parts work in the harsh environment of high temperature and high pressure, and bear the function of reciprocating load, so the faults are easy to occur. Compared with other parts of reciprocating compressor, the valve is the most vulnerable and faults occur more frequently. From the point of view of fault diagnosis, accident prevention, maintenance decision and cost saving, it is very important and meaningful to study an effective and accurate fault diagnosis method for reciprocating compressor valve. The vibration signal of reciprocating compressor contains a large number of periodic components and transient shock components, and has obvious non-stationary characteristics. Its components, such as piston, connecting rod, valve and so on, have the same motion cycle, and the frequency characteristics overlap in frequency spectrum, so it is difficult to distinguish. In this paper, the vibration signal model of air valve is established based on the motion mechanism of reciprocating compressor. Then, a new fault diagnosis method for reciprocating compressor valve is proposed by integrating signal processing and pattern recognition techniques. This method combines three algorithms: base tracking, waveform matching and support vector machine to identify and diagnose the valve state in time domain. Base tracking is used to extract the main components of vibration signal and suppress background noise. Waveform matching is a new feature extraction algorithm proposed in this paper. Most of the traditional feature extraction algorithms use statistical features and entropy to extract features. In contrast, waveform matching extracts the feature by matching the vibration waveform with the parameterized waveform. The matching process is optimized by differential evolution algorithm. The dimension of the feature extracted by waveform matching is small and each feature has definite physical meaning. Support vector machine (SVM) is suitable for small sample problem and is used to realize fault identification and classification in this paper. After the introduction of the theory, the proposed method is applied to experimental data and field data. The experimental data show that the diagnostic method can accurately and reliably identify the three states of the reciprocating compressor valve (normal, spring deterioration and valve plate deformation). The field data show that the diagnosis method can effectively identify the fault degree of the valve. Finally, the advantages and limitations of the proposed new method are discussed, and the further research based on this method is prospected.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH45;TH165.3
本文编号：2169977
[Abstract]:Reciprocating compressors are widely used in petroleum and chemical industry, and often belong to the key equipment in the production line. The working state of reciprocating compressors directly affects the economic benefits of enterprises. Its structure is complex, and many important parts work in the harsh environment of high temperature and high pressure, and bear the function of reciprocating load, so the faults are easy to occur. Compared with other parts of reciprocating compressor, the valve is the most vulnerable and faults occur more frequently. From the point of view of fault diagnosis, accident prevention, maintenance decision and cost saving, it is very important and meaningful to study an effective and accurate fault diagnosis method for reciprocating compressor valve. The vibration signal of reciprocating compressor contains a large number of periodic components and transient shock components, and has obvious non-stationary characteristics. Its components, such as piston, connecting rod, valve and so on, have the same motion cycle, and the frequency characteristics overlap in frequency spectrum, so it is difficult to distinguish. In this paper, the vibration signal model of air valve is established based on the motion mechanism of reciprocating compressor. Then, a new fault diagnosis method for reciprocating compressor valve is proposed by integrating signal processing and pattern recognition techniques. This method combines three algorithms: base tracking, waveform matching and support vector machine to identify and diagnose the valve state in time domain. Base tracking is used to extract the main components of vibration signal and suppress background noise. Waveform matching is a new feature extraction algorithm proposed in this paper. Most of the traditional feature extraction algorithms use statistical features and entropy to extract features. In contrast, waveform matching extracts the feature by matching the vibration waveform with the parameterized waveform. The matching process is optimized by differential evolution algorithm. The dimension of the feature extracted by waveform matching is small and each feature has definite physical meaning. Support vector machine (SVM) is suitable for small sample problem and is used to realize fault identification and classification in this paper. After the introduction of the theory, the proposed method is applied to experimental data and field data. The experimental data show that the diagnostic method can accurately and reliably identify the three states of the reciprocating compressor valve (normal, spring deterioration and valve plate deformation). The field data show that the diagnosis method can effectively identify the fault degree of the valve. Finally, the advantages and limitations of the proposed new method are discussed, and the further research based on this method is prospected.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH45;TH165.3
【引证文献】
相关期刊论文 前1条
1 董玉琼;刘锦南;张藻平;;往复压缩机气阀故障的预知性维修[J];中国设备工程;2012年10期
，本文编号：2169977
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2169977.html
