基于非线性故障重构的旋转机械故障预测方法

Nonlinear Fault Reconstruction Based Fault Prognosis for Rotating Machinery

对旋转机械的状态进行在线监测和故障预测是一个:!WriE要应用价伉的工程M题.釆⑴驻于核fe元分析的非线性故障巫构技术研究了多变量相关条件下旋转机械的故障估计及预测问题.首先利用核主元分析对旋转机械系统进行离线非线性建模,并进行异常检测.通过对故障程度进行定量描述,用最优化方法求解故障重构意义F的故障估计;然后用多层递阶的方法对估计出的故障幅值的发展趋势进行预测.最后,以中国石化北京燕山分公司的烟气轮机作为实际应用对象,验证了该方法的有效性.

It is an engineering problem of great application value to monitor the rotating machinery online and perform fault prognosis. A fault prognosis method is proposed for rotating machinery based on kernel principal component analysis and nonlinear fault reconstruction. First, kernel principal component analysis is used to build a model of rotating machinery and perform process monitoring. By describing the fault magnitude quantificationally, an optimal fault estimation is obtained with an optimization method. At last, a multi-layer hierarchical forecasting method is adopted to predict the fault magnitude evolution. The flue gas turbine of Beijing Yanshan Petrochemical Corporation is used to demonstrate the effectiveness of the proposed method.