基于信息熵理论和Parks聚类分析的水电机组振动故障诊断

Vibration Fault Diagnosis of Hydraulic Generating Units Based on Information Entropy Theory and Parks Clustering Analysis

针对水电机组振动故障与征兆之间复杂的非线性关系,将经过整理的水电机组典型故障分别作为标准故障类,每个标准故障类和它所对应的具有代表性的特征参数构成故障类特征向量,建立标准故障特征参数矩阵。采用信息熵理论和Parks聚类分析方法对待检样本进行聚类分析,从而辨识出待检样本最有可能属于哪个故障类,即最有可能是哪种故障。通过实例检验表明理论计算与现场检查结果相符,证明该方法能有效地确定故障类型和发生故障的部位,适合于故障诊断中自动模式识别,具有良好的实际应用前景,为水电机组状态监测及故障诊断提供了一种新途径。

The faults of hydraulic generating units are classified into different clusters by the characteristic parameters using statistical analysis methods. A characteristic parameters matrix of standard faults is established. Clustering analysis has been used to identify the new sample. This clustering technique is based on information entropy theory and Parks clustering analysis, which has been tested in a real application. Results demonstrated that the proposed method is a good candidate to be used as an online diagnosis tool for hydraulic generating units.