往复泵液力端故障诊断研究
发布时间：2018-08-28 15:58
【摘要】：往复泵作为工业领域重要的往复设备,广泛应用于石油钻井、油气开采等生产领域。液力端作为往复泵的薄弱环节在使用当中经常发生故障,由于液力端的结构复杂、激励源多,导致液力端的故障诊断比较困难。开展对往复泵液力端的故障诊断研究工作对于提高油田往复泵的使用效率,保证安全生产具有十分重要的意义。柱塞磨损故障和泵阀组件故障是本文诊断研究的重点,本文利用时域统计指标诊断方法对柱塞磨损故障进行了诊断；同时,利用小波包分析方法提取了泵阀组件的故障特征,在MATLAB中创建了BP神经网络,并利用创建的人工神经网络实现了液力端故障的智能识别。这些诊断方法可较好地用于往复泵液力端的故障诊断当中。 本文的研究工作主要体现在以下几个方面： 首先,在广泛调研国内外文献以及现场资料的基础上,分析了往复泵液力端的工作原理、常见故障及故障原因。建立了简化后阀盖振动的力学模型,论述了利用振动信号进行液力端故障诊断的可行性。指出了液力端振动信号的非平稳性和故障是促使振动信号发生变化的主要原因。设计了液力端的故障实验方案,在3DS-1/12.5型往复泵上进行了故障实验； 其次,针对液力端振动信号的非平稳性特点,对柱塞磨损故障采用了时域统计指标诊断方法,计算出了各指标的重要度系数,建立了柱塞磨损程度的诊断模型,并通过故障程度值较为直观地反映了柱塞的磨损程度；针对泵阀振动信号具有非平稳特性、故障特征提取难度大的特点,在对几种时频分析方法进行对比分析的基础上,本文引入了小波和小波包变换方法,对振动信号进行小波包分析；提出了能量最大差别法来确定小波包分解方案,通过建立“频率—能量—故障”的诊断模型,有效提取出了泵阀组件发生不同类型、不同程度时的故障特征。 然后,详细介绍了BP神经网络的算法原理,在MATLAB中利用BP网络的改进措施实现了BP网络的创建与训练,通过所创建的神经网络实现了液力端故障的智能识别,并且诊断准确率均达到了90%以上。 最后,针对实验中存在的问题和不足,设计了往复泵液力端故障的振动诊断方案与整体性能诊断方案,并对往复泵液力端测试硬件系统的方案进行了设计,提高了诊断测试方法在油田现场的可实施性。
[Abstract]:As an important reciprocating equipment in industry, reciprocating pump is widely used in oil drilling, oil and gas production. As the weak link of reciprocating pump, the hydraulic end often breaks down in use. Because of the complex structure of the hydraulic end and many exciting sources, it is difficult to diagnose the fault of the hydraulic end. It is very important to develop the fault diagnosis of reciprocating pump hydraulic end for improving the efficiency of oil field reciprocating pump and ensuring safe production. The fault of plunger wear and the fault of pump valve assembly are the focal points of this paper. The fault diagnosis of plunger wear is carried out by using time domain statistical index diagnosis method, and the fault characteristics of pump valve assembly are extracted by wavelet packet analysis method. The BP neural network is established in MATLAB, and the intelligent identification of hydraulic end faults is realized by using the created artificial neural network. These methods can be used in fault diagnosis of hydraulic end of reciprocating pump. The research work of this paper is mainly reflected in the following aspects: firstly, on the basis of extensive investigation of domestic and foreign literature and field data, the working principle, common faults and fault causes of hydraulic end of reciprocating pump are analyzed. A simplified mechanical model of valve cover vibration is established, and the feasibility of hydraulic end fault diagnosis using vibration signal is discussed. It is pointed out that the non-stationarity and fault of the vibration signal at the hydraulic end are the main reasons for the change of the vibration signal. The fault test scheme of hydraulic end is designed, and the fault experiment is carried out on 3DS-1/12.5 reciprocating pump. Secondly, according to the non-stationary characteristic of vibration signal at hydraulic end, the time-domain statistical index diagnosis method is used to diagnose the plunger wear fault. The importance coefficient of each index is calculated, the diagnostic model of piston wear degree is established, and the wear degree of plunger is intuitively reflected by the value of fault degree, and the vibration signal of pump valve has non-stationary characteristic. On the basis of comparing and analyzing several time-frequency analysis methods, wavelet transform and wavelet packet transform are introduced to analyze vibration signal. In this paper, the maximum difference method of energy is proposed to determine the wavelet packet decomposition scheme. By establishing a diagnosis model of "frequency-energy-fault", the fault characteristics of different types and different degrees of pump valve components are extracted effectively. Then, the algorithm principle of BP neural network is introduced in detail, and the establishment and training of BP network are realized by using the improvement measure of BP network in MATLAB, and the intelligent identification of hydraulic end fault is realized by the created neural network. And the diagnostic accuracy is over 90%. Finally, aiming at the problems and shortcomings in the experiment, the vibration diagnosis scheme and the overall performance diagnosis scheme of hydraulic end fault of reciprocating pump are designed, and the scheme of testing hardware system of hydraulic end of reciprocating pump is designed. The practicability of diagnosis and test method in oil field is improved.
【学位授予单位】：西南石油大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3
本文编号：2209888
[Abstract]:As an important reciprocating equipment in industry, reciprocating pump is widely used in oil drilling, oil and gas production. As the weak link of reciprocating pump, the hydraulic end often breaks down in use. Because of the complex structure of the hydraulic end and many exciting sources, it is difficult to diagnose the fault of the hydraulic end. It is very important to develop the fault diagnosis of reciprocating pump hydraulic end for improving the efficiency of oil field reciprocating pump and ensuring safe production. The fault of plunger wear and the fault of pump valve assembly are the focal points of this paper. The fault diagnosis of plunger wear is carried out by using time domain statistical index diagnosis method, and the fault characteristics of pump valve assembly are extracted by wavelet packet analysis method. The BP neural network is established in MATLAB, and the intelligent identification of hydraulic end faults is realized by using the created artificial neural network. These methods can be used in fault diagnosis of hydraulic end of reciprocating pump. The research work of this paper is mainly reflected in the following aspects: firstly, on the basis of extensive investigation of domestic and foreign literature and field data, the working principle, common faults and fault causes of hydraulic end of reciprocating pump are analyzed. A simplified mechanical model of valve cover vibration is established, and the feasibility of hydraulic end fault diagnosis using vibration signal is discussed. It is pointed out that the non-stationarity and fault of the vibration signal at the hydraulic end are the main reasons for the change of the vibration signal. The fault test scheme of hydraulic end is designed, and the fault experiment is carried out on 3DS-1/12.5 reciprocating pump. Secondly, according to the non-stationary characteristic of vibration signal at hydraulic end, the time-domain statistical index diagnosis method is used to diagnose the plunger wear fault. The importance coefficient of each index is calculated, the diagnostic model of piston wear degree is established, and the wear degree of plunger is intuitively reflected by the value of fault degree, and the vibration signal of pump valve has non-stationary characteristic. On the basis of comparing and analyzing several time-frequency analysis methods, wavelet transform and wavelet packet transform are introduced to analyze vibration signal. In this paper, the maximum difference method of energy is proposed to determine the wavelet packet decomposition scheme. By establishing a diagnosis model of "frequency-energy-fault", the fault characteristics of different types and different degrees of pump valve components are extracted effectively. Then, the algorithm principle of BP neural network is introduced in detail, and the establishment and training of BP network are realized by using the improvement measure of BP network in MATLAB, and the intelligent identification of hydraulic end fault is realized by the created neural network. And the diagnostic accuracy is over 90%. Finally, aiming at the problems and shortcomings in the experiment, the vibration diagnosis scheme and the overall performance diagnosis scheme of hydraulic end fault of reciprocating pump are designed, and the scheme of testing hardware system of hydraulic end of reciprocating pump is designed. The practicability of diagnosis and test method in oil field is improved.
【学位授予单位】：西南石油大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3
【相似文献】
相关期刊论文 前10条
1 刘光恒,曾艳丽,孙震,秦臻,韩乃波,刘庆云;高压水射流清洗机对称角度布置液力端设计[J];清洗世界;2005年11期
2 张庆华;李宗义;;浅析往复式泥浆泵液力端结构优化[J];中小企业管理与科技(下旬刊);2014年07期
3 方长传;;钻井泵液力端的维护保养与焊接修理技术[J];长江大学学报(自科版);2013年13期
4 蒋金泉;;浅谈复合式高低压缸液力端在四缸试压泵中的应用[J];价值工程;2014年06期
5 薛苏玲;李洪波;康亮;周天明;张福;冯利杰;曾兴昌;;F-1600泥浆泵液力端阀座的有限元分析[J];机械工程师;2009年08期
6 曾文;别锋锋;;往复柱塞泵液力端故障诊断方法研究[J];中国石油和化工标准与质量;2013年14期
7 叶晓琰;杨孙圣;许建强;胡敬宁;;基于CFD技术的高压泵液力端优化设计[J];排灌机械;2008年06期
8 沈科君;裴峻峰;贺超;黄显茹;;基于盲源分离的往复泵液力端故障机理[J];化工进展;2014年03期
9 丁锡锋;;电动往复泵液力端常见故障原因及排除[J];石油和化工设备;2011年08期
10 ;往复泵工况综合诊断仪[J];国外油田工程;2005年12期
相关会议论文 前1条
1 范春春;;油田常用特车泵液力端润滑系统的发展趋势[A];第五届设备管理第八届设备润滑与液压学术会议论文集——《设备管理设备润滑与液压技术》[C];2004年
相关硕士学位论文 前1条
1 吴聚;往复泵液力端故障诊断研究[D];西南石油大学;2012年
，本文编号：2209888
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2209888.html
