论文部分内容阅读
提出一种故障诊断的方法,利用模糊C均值聚类,对汽轮机故障信号中提取的特征样本进行聚类分析,利用故障样本隶属度矩阵寻找其中一些故障特征不太明显甚至错误的样本,以此加强不同故障样本的特征,最后对优化的样本进行基于支持向量机训练,以此训练模型进行故障诊断。通过实验可以看出,经过优化后的样本,训练出的故障诊断模型精度得到了提高。
A method of fault diagnosis is proposed. By using fuzzy C-means clustering, cluster analysis is made on the characteristic samples extracted from the fault signals of turbines, and some of the fault samples with less obvious or even fault characteristics are found by using the membership matrix of fault samples. Strengthen the characteristics of different fault samples, and finally optimize the samples based on support vector machine training, training model for fault diagnosis. It can be seen through experiments that the accuracy of the fault diagnosis model trained has been improved after the optimized samples.