汽轮机组非平稳运行过程振动数据的时间序列分析

VIBRATION ANALYSIS OF TURBINE UNIT′S NONSTATIONARY PROCESS WITH TIME SERIES ANALYSIS METHOD

时间序列的分析方法已经在许多工程领域得到广泛而成功的应用。文中通过对实际电厂运行汽轮机组非平稳过程中轴系振动数据的相关分析 ,对汽轮机组振动数据非平稳时间序列的MA(q)模型的识别和建模进行初步的探讨。为验证MA(q)模型对于汽轮机组非平稳过程中的振动数据分析的有效性 ,分别采用MA(2 )、AR(2 0 )和 3阶多项式回归分析的方法 ,利用现场实测数据进行对比预测实验。实验结果表明利用MA(2 )模型的预测效果明显优于采用AR(p)模型预测方法和传统的多项式回归预测方法的预测结果 ,从而验证采用MA(q)模型对于汽轮机组非平稳运行过程中振动信号分析研究的有效性。

The time series analysis method has been widely used and succeeded in a lot of fields of engineering. Due to the complexity of systems and the lack of real data source, the time series analysis method is not used widely in the field of large and complex machinery system in China. Through the correlation analysis of vibration data of turbine unit's nonstationary process, the autocorrelation and partial autocorrelation functions have been compared, the identification and modeling of the turbine's nonstationary process with MA(q) model have been discussed. On the other hand, compared with the result of correlation analysis of stationary process, statistical difference has been showed between these two processes. In order to verify the validity of this kind of time series model, three trend prediction experiments, which respectively uses MA(2), AR(20) and regression analysis method, have been designed according to the fieldwork data. The results of the experiments show that the trend prediction result of MA(2) is much better than those of AR(20) model and regression analysis method. A conclusion can be made that MA(q) model is effective way to analyze the vibration data of turbine unit's nonstationary process. A point of view has been showed to explain the inside connection between the real physical process and MA(q) model. Development of the application of the time series method in this field has been discussed.