FARIMA模型在复杂机械系统的故障诊断中的应用
发布时间：2019-01-05 10:54
【摘要】：长期以来对复杂的机械系统，人们希望能够及时、准确地发现故障、判断故障的损伤程度，并且做出评估与预测，因此故障诊断技术也随之越来越受到重视，并且在工业领域及信号检测领域是很有价值的课题。时间序列分析是一种经典的分析方法，在故障诊断中有独特的优势，，大多数情况是对振动信号建立ARMA模型并进行分析，但是这种方法有一定的局限性。注意到很多情形的振动信号的具有长记忆特性，本文尝试用FARIMA模型对故障诊断问题进行分析。 本文的目的是引入体现长记忆特征的模型，通过实例证明FARIMA模型比传统的ARMA模型对模拟故障诊断的振动数据更为精确。本文对FARIMA模型的长记忆特性和分数差分的两个特性进行了分析，从理论上说明了用FARIMA建模的条件和优势。本文详细叙述了平稳时间序列的建模步骤、FARIMA模型的建模步骤，并且总结了参数估计的方法，说明了FARIMA模型虽然是ARMA模型的推广但是它们之间有很大的不同。通过对Bently实验台得到的汽轮机转子的振动信号进行分析，结合MATLAB、SAS软件实现的模拟结果和对参数的估计结果，本文显示了这种建模方法比传统的建模方法更能有效的模拟并且进行振动信号的分析。在建立FARIMA模型的时候，考虑了非高斯噪音扰动的S α S-FARIMA模型和随时间变化参数会发生改变的T-V-FARIMA模型，这两种特殊的情况反映了FARIMA模型对某些实际数据进行建模的灵活性和有效性，同时给出了修正这个模型的方向。
[Abstract]:For a long time, people hope to find the fault in time and accurately, judge the damage degree of the fault, and make the evaluation and prediction for the complex mechanical system. Therefore, the fault diagnosis technology has been paid more and more attention. And in the field of industry and signal detection is a very valuable subject. Time series analysis is a classical analysis method, which has unique advantages in fault diagnosis. In most cases, the ARMA model of vibration signal is established and analyzed, but this method has some limitations. Noting that many vibration signals have long memory characteristics, this paper attempts to use FARIMA model to analyze the problem of fault diagnosis. The purpose of this paper is to introduce a long memory model. It is proved that the FARIMA model is more accurate than the traditional ARMA model in simulating the vibration data of fault diagnosis. In this paper, the long memory characteristics of FARIMA model and the two characteristics of fractional difference are analyzed, and the conditions and advantages of FARIMA modeling are explained theoretically. In this paper, the modeling steps of stationary time series and FARIMA model are described in detail, and the methods of parameter estimation are summarized. It is shown that although FARIMA model is a generalization of ARMA model, there are great differences between them. By analyzing the vibration signals of the turbine rotor obtained from the Bently test bench, combining the simulation results of the MATLAB,SAS software and the estimation of the parameters, the vibration signals of the turbine rotor are analyzed. This paper shows that this modeling method is more effective than the traditional modeling method in simulating and analyzing vibration signals. In establishing the FARIMA model, the S 伪 S-FARIMA model with non-Gao Si noise disturbance and the T-V-FARIMA model with time-varying parameters are considered. These two special cases reflect the flexibility and validity of the FARIMA model for modeling some real data, and the direction of modifying the model is given.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3;O211.61
本文编号：2401683
[Abstract]:For a long time, people hope to find the fault in time and accurately, judge the damage degree of the fault, and make the evaluation and prediction for the complex mechanical system. Therefore, the fault diagnosis technology has been paid more and more attention. And in the field of industry and signal detection is a very valuable subject. Time series analysis is a classical analysis method, which has unique advantages in fault diagnosis. In most cases, the ARMA model of vibration signal is established and analyzed, but this method has some limitations. Noting that many vibration signals have long memory characteristics, this paper attempts to use FARIMA model to analyze the problem of fault diagnosis. The purpose of this paper is to introduce a long memory model. It is proved that the FARIMA model is more accurate than the traditional ARMA model in simulating the vibration data of fault diagnosis. In this paper, the long memory characteristics of FARIMA model and the two characteristics of fractional difference are analyzed, and the conditions and advantages of FARIMA modeling are explained theoretically. In this paper, the modeling steps of stationary time series and FARIMA model are described in detail, and the methods of parameter estimation are summarized. It is shown that although FARIMA model is a generalization of ARMA model, there are great differences between them. By analyzing the vibration signals of the turbine rotor obtained from the Bently test bench, combining the simulation results of the MATLAB,SAS software and the estimation of the parameters, the vibration signals of the turbine rotor are analyzed. This paper shows that this modeling method is more effective than the traditional modeling method in simulating and analyzing vibration signals. In establishing the FARIMA model, the S 伪 S-FARIMA model with non-Gao Si noise disturbance and the T-V-FARIMA model with time-varying parameters are considered. These two special cases reflect the flexibility and validity of the FARIMA model for modeling some real data, and the direction of modifying the model is given.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3;O211.61
【参考文献】
相关期刊论文 前3条
1 吴庚申,梁平,龙新峰;基于ARMA的汽轮机转子振动故障序列的预测[J];华南理工大学学报(自然科学版);2005年07期
2 于开平;庞世伟;赵婕;;时变线性/非线性结构参数识别及系统辨识方法研究进展[J];科学通报;2009年20期
3 夏松波,张新江,刘占生,徐世昌;旋转机械不对中故障研究综述[J];振动.测试与诊断;1998年03期
本文编号：2401683
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2401683.html
