基于全矢—灰色模型的频谱预测研究

发布时间：2017-12-28 11:41

本文关键词：基于全矢—灰色模型的频谱预测研究　出处：《郑州大学》2017年硕士论文　论文类型：学位论文

【摘要】：设备故障预测就是要通过技术手段,预测在未来的一定时间内机组的运行状态,预知将要发生的设备故障,以便指导生产调度和设备维修。当前,国内外学者对于故障预测的研究主要集中于故障强度、剩余寿命和有无故障方面,多以事先确定的某种特定的单一故障展开,而对于故障性质和类型的预测则相对较少。要对故障类型和性质进行预测,就需要对设备振动信号的频谱结构的发展趋势进行预测,并根据预测所得的频谱结构判断机组的故障性质和类型。从不同方向采集到的信号进行频谱分析时,发现频谱结构存在差异性,因此,仅仅依靠单源信号进行预测,很难反映出机组的运行状态。为提高旋转机械故障预测的可靠性,将基于同源信息融合的全矢谱技术引入预测模型,可有效的避免单源信号故障特征提取不完整的弊端,完整的表述转子在整个截面内的振动状态。本课题的研究内容和主要成果如下:(1)研究了全矢-GM(1,1)模型频谱预测方法和建模过程。给出了趋势预测方法灰色GM(1,1)模型结合全矢谱技术的全矢-GM(1,1)模型的具体建模过程,试验表明,全矢-GM(1,1)模型具有较高的预测精度,并在机械振动强度的预测中体现出了较好的预测效果。(2)研究了基于全矢-MGM(1,m)模型的频谱结构预测流程及应用。将灰色MGM(1,m)模型预测方法结合全矢谱技术应用到机械振动频谱预测研究中,并给出了频谱预测的全过程。结果表明,全矢-MGM(1,m)模型也具有较好的预测效果,并在中长期预测中,展现出更高的预测精度。(3)研究了基于全矢-BPGM(1,1)模型的频谱预测方法。运用EMD的方法分离特征值中的趋势量与随机量,将趋势量送入GM(1,1)模型中进行预测,随机量送入BP神经网络预测,将预测结果进行整合,得到最终预测值。实验表明,全矢-BPGM(1,1)模型拥有更高的预测精度,可以为设备预知维修提供技术支持。
[Abstract]:Equipment fault prediction is to predict the running state of the unit in the future time by means of technology and predict the equipment failure that will happen, so as to direct production scheduling and equipment maintenance. At present, the research on fault prediction mainly focused on fault intensity, residual life and whether there is fault or not at home and abroad. Most of the research on fault prediction is carried out in advance with a specific single fault, while the prediction for the nature and type of fault is relatively few. To predict the types and properties of faults, we need to predict the development trend of the spectrum structure of equipment vibration signals, and judge the nature and types of faults based on the spectrum structure predicted. When spectrum analysis is carried out from different directions, spectrum structure is different. Therefore, it is difficult to predict the running state of the unit by predicting single source signal. In order to improve the reliability of rotating machinery fault prediction, the full vector spectrum technology based on homologous information fusion is introduced into the prediction model, which can effectively avoid the drawback of incomplete fault feature extraction of single source signal, and completely express the vibration state of the rotor in the whole section. The main contents and main achievements of this research are as follows: (1) the spectrum prediction method and modeling process of the full vector -GM (1,1) model are studied. The concrete modeling process of the full vector -GM (1,1) model with the trend prediction method, the grey GM (1,1) model and the full vector spectrum technique is given. The experiment shows that the full vector -GM (1,1) model has high prediction accuracy and shows a good prediction effect in the prediction of the mechanical vibration intensity. (2) the spectrum structure prediction process based on the full vector -MGM (1, m) model and its application are studied. The grey MGM (1, m) model prediction method and full vector spectrum technique are applied to the study of mechanical vibration spectrum prediction, and the whole process of spectrum prediction is given. The results show that the full vector -MGM (1, m) model also has a better prediction effect, and shows a higher prediction accuracy in the medium and long term prediction. (3) the spectrum prediction method based on the full vector -BPGM (1,1) model is studied. The trend and random variables in the eigenvalues are separated by EMD, and the trend is sent to the GM (1,1) model to predict. The random amount is sent to BP neural network prediction, and the prediction results are integrated to get the final prediction value. The experiment shows that the full vector -BPGM (1,1) model has higher prediction precision, and it can provide technical support for equipment maintenance.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17

【相似文献】