基于灰理论的变频调速系统故障诊断的研究

【作者】 李婧；

【导师】 王新；

【作者基本信息】 河南理工大学 ， 控制理论与控制工程， 2009， 硕士

【摘要】 最优经济运行导致设备系统日趋规模化、复杂化和高度自动化。新科技的飞速发展及其应用,又大大增加了设备结构的复杂性、运行环境的独特性和诱发故障的多源性,使得故障检测、诊断、预测愈发困难,故障造成的损失和维修费用也大幅上升。故障预测研究能够有效预防事故的发生,最大程度地降低故障损失,并能提供异常现象的事件数据及原因。利用新的科学理论,探索新的诊断预测方法是故障诊断预测研究领域的一个重要方面。灰色预测模型具有要求样本数据少、计算方便、预测精度较高并且可对结果进行校验等优点,因此被广泛应用。灰建模的累加生成运算在一定程度上强化了确定性成分,弱化了不确定性成分。由微分方程描述的灰模型能够较完整地描述被研究对象的运行行为,能够揭示系统内部事物连续发展变化的过程。但传统的GM (1,1)模型只适用于等间距序列的建模预测。实践中,经常会遇到不等间距的情况,于是一些不等距灰色建模方法被提出来,不过难免有些缺陷。其中一般累加生成差分UGM (1,1)模型中平滑值是一种等权生成,不符合数据的实际情况,因此本文在该模型的基础上提出一种平滑值权值寻优生成的RUGM (1,1)模型,使预测精度提高了10%以上,同时计算量并未明显增加。文章中通过实践指明UGM (1,1)模型应用中的一些条件及注意事项,并给出了一些建议。本文针对故障诊断中数据采集问题,采用虚拟仪器对变频器输出信号进行采集,然后把数据导入仿真软件Matlab中计算,并针对故障诊断中特征量难以提取的问题,运用小波分析及其它先进的信号分析方法对变频器的输出信号进行了分析和故障特征的提取,指出了输出电流的歪度的绝对值和三分频下输出电压的特定频带的能量可作为变频器的故障特征。更多还原

【Abstract】 The optimal economic system led to increasing size of equipment complicated and highly automated. The rapid development of new technology and its applications greatly increased the complexity of the structure of equipment, uniqueness of environment and multi-induced failure of endogenous, making fault detection, diagnosis, prediction more difficult, the losses caused by faults and maintenance costs have also increased significantly. Prediction of failure can effectively prevent accidents, minimize the failure losses and provide information and reasons of abnormal events. To use new scientific theories to explore new methods of diagnosis and forecast is an important aspect in field of prediction of failure diagnosis.Grey AGO computing model to a certain extent strengthens the element of uncertainty,weakens the element of uncertainty. Grey model described by differential equations can study the activities of objects more complete,and reveal the continuous changing process of internal system. However, traditional GM (1,1) model is only applicable to pitch sequence, such as the prediction model.Non-equidistance often happens in practice,so some of the non-equidistance method of grey model is put forward, but there are inevitably some shortcomings. Smooth value of general AGO difference UGM (1,1) model is a equal power generating, data do not meet the actual situation. Therefore, in this paper, a smooth weight optimization generated RUGM(1,1) model is put forward, so that the prediction accuracy increases by more than 10%, at the same time calculation does not increase significantly, some suggestions are given on conditions and notes about application of UGM(1,1) model.In this paper, virtual instruments are used to collect output signal of transducer, and then import the data in the simulation software Matlab for calculation,and aim at the problem which features for diagnosis is difficult to extract, using wavelet analysis and other advanced signal analysis methods on output signals for analyzing and extraction of fault features, pointing out that the absolute value of the degree of distortion of the output current and a particular band of energy of the output voltage under one-third of converter can be used as the characteristics of the fault.更多还原