基于高斯混合模型的汽轮机振动故障诊断

采用高斯混合模型(GMM)与小波包分析相结合的方法,对汽轮机振动故障进行了诊断研究.首先对振动故障信号进行小波包分解,去除干扰信号,提取包含故障特征信息的频段作为故障特征矢量.以此特征矢量建立GMM,并用建立的模型识别各种故障.利用在Bently实验台上测得的实验数据进行建模及故障识别.计算结果中,当模数M=12时,GMM识别故障的正确率约80%～90%,表明GMM结合小波包分析进行汽轮机振动故障诊断的方法能取得较好的效果.

Turbine Faults Diagnosis Based on Gaussian Mixture Models

The Gaussian Mixture Models and the wavelet packet analysis are used to the turbine vibration faults diagnosis. De-compound firstly the vibration faults signal and delete the disturbed component. Then, take the frequency segments which contain the fault characteristics as the fault characteristics vector. To set up the Gaussian Mixture Models with the vectors, and identify the different faults with the built model. The experiment data measured in Benlty experiment platform is adopted to set up the model and identify the faults. In the calculation results, when the modulus equal to twelve, the precision for the faults diagnosis by the Gaussian Mixture Models is approximately 80% ～ 90%. It indicates that the turbine vibration fault can be diagnosed effectively by the Gaussian Mixture Models and the wavelet packet analysis.