基于PCA故障重构的烟气轮机故障预测

Fault Prediction of Turbine Machine Based PCA Fault Reconstruction Method

主元分析作为一种统计分析方法,被广泛的用于过程监控中。文中假设故障可完全重构,利用基于平方预测误差的故障重构方法对故障幅值进行估计,并采用多层递阶方法对估计出的故障幅值进行预测。以北京燕山石化公司炼油厂的烟气轮机组作为研究对象,验证了该方法的有效性。

As a statistical analysis method, Principal Component Analysis（PCA） is widely used for process monitoring. Assuming fault can be completely reconstructed in this paper. The fault reconstructed method is explored by using Square Predicted Error （SPE） index, and the fault magnitude is estimated by fault reconstruction and predicted by multi-level recursive prediction method. We take the huge stack gas turbine set as the research object. The test shows efficiency of the proposed approach.