论文部分内容阅读
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。
Turbine generator failure manifests itself as anomalous frequency components in the vibration signal. The Hilbert-Huang transform adaptively extracts this frequency component and forms the time-frequency spectrum. However, during the transformation process, when the extreme points on both sides are not extreme points, the extreme envelope of the cubic spline fitting will deviate from the actual value and “pollute” inwardly as the decomposition continues. This paper proposes a Hilbert-Huang transform based on least-squares support vector machine regression. This method uses least-squares support vector machine regression to extend both ends of the original signal to obtain additional extreme points, and then uses cubic spline interpolation The method obtains the upper and lower envelopes, and realizes the accurate EMD decomposition. The improved Hilbert-Huang variation is applied to hydroelectric generating set fault diagnosis. The results show that the method effectively restrain the end effect and realize the accurate identification of the fault.