基于流形学习方法的汽轮机组振动特征提取

为了提高汽轮机振动故障信号的可分性和诊断正确率，应用流行学习方法对汽轮机振动信号进行故障特征提取。研究结果表明，应用流行学习方法可以有效地提取汽轮机振动故障的特征信息，将不同故障类型的特征信息有效地区分开来。运用流行学习方法进行故障特征提取后的诊断结果与小波包分析方法相比，诊断正确率明显提高。

Feature Extraction of Vibration of Turbine Unit Based on Manifold Learning Method

In order to improve the classification ability and diagnostic accuracy of turbine vibration signals, a new feature extraction method from fault signals of turbine vibration based on the manifold learning method (MLM) is proposed. The results show that the MLM effectively extracts fault feature information of turbine vibration and separates different types of fault feature information. The diagnostic accuracy of features extracted by the MLM is significantly higher than that of the wavelet packet analysis method.