最小二乘支持向量机回归的HHT在水轮发电机组故障诊断中的应用

The Hilbet-Huang Transform Based on Least Squares Support Regression Machine and Its Application in the Fault Diagnosis of Hydroelectric Generation Unit

水轮发电机组的故障表现为振动信号中出现异常频率成分,Hilbert-Huang变换可自适应地将这种频率成分提取出来并形成时频谱。但变换过程中,当两侧端点不为极值点时,会造成三次样条拟合的极值包络线偏离实际值,并且随着分解的不断进行向内“污染”。提出基于最小二乘支持向量机回归的Hilbert-Huang变换,该方法采用最小二乘支持向量机回归的方法对原信号两端进行拓延,得到附加的极值点,再利用三次样条插值的方法得到上、下包络线,实现了准确的EMD分解。将改进后的Hilbert-Huang变化应用于水轮发电机组故障诊断中,结果表明,该方法有效抑制了端点效应,实现了故障的准确识别。

Faults of hydroelectric generation unit appear when abnormal frequencies are found in the signal of vibration. Hilbert-Huang transform can distill these frequencies automatically and time-frequency spectrum can be obtained. In the process of Hilbert-Huang transform, if ends are not the extremism, end effects occur due to the spline fitting at the data ends and the effect will be expended to inner data set along with the decomposition. An improved Hilbert-Huang transform based on least squares support vector regression machine is derived. Additional extremum can be obtained firstly by data extending based on least squares support vector machine, then the envelop can be found and the empirical mode decomposition can be process exactly. This method is used in the fault diagnosis of hydroelectric generation unit, the result demonstrates that end effects can be controled effectively and faults can be recognized exactly.