基于混合模糊聚类分析的汽轮发电机组振动故障诊断

Vibration Fault Diagnosis of Turbo-generator Set Based on Hybrid Fuzzy Clustering Analysis

针对模糊C-均值(FCM)算法在汽轮发电机组振动故障诊断中的不足,提出了一种加权混沌优化FCM(WCOFCM)算法。WCOFCM算法首先将混沌优化策略与传统FCM算法相结合,用混沌变量搜索对模糊聚类目标函数进行全局寻优,并结合梯度算子,使方法有效收敛到极值点。然后依样本相似度原理对样本特征进行加权,对不同的特征赋予权重,突出敏感特征对聚类结果的主导作用,提高了聚类性能。最后依据聚类有效性函数指标自动确定聚类数,实现自适应分类。用该方法对国际标准测试数据进行了聚类分析实验,并将该方法应用于某发电厂汽轮发电机组振动故障诊断,其结果表明该方法有效降低了误分类率,能对汽轮发电机组振动故障进行有效诊断。

To overcome the disadvantages of fuzzy C-means （FCM） method in the turbo-generator vibration fault diagnosis, a novel weighted chaos option fuzzy C-means （WCOFCM） algorithm is proposed. The WCOFCM algorithm combines chaos optimization strategy with the conventional FCM method, and optimizes fuzzy clustering objective function via chaos variable searching and gradient operator to achieve effective convergences. The similarity in samples is considered in assigning weights for different characteristics to obtain an adaptive classification. Moreover, the WCOFCM algorithm can automatically get the number of clusters according to the validity function. The algorithm is successfully applied in analyzing an international standard testing data set, as well as diagnosing vibration faults of a turbo-generator set in a power plant. The results show that it reduces the misclassification rate and therefore effectively diagnoses the turbo-generator vibration fault.