443 / 2018-12-30 12:34:46

基于谱聚-KNN模型的汽轮机故障诊断方法研究 Research on the Fault Diagnosis Method of Steam Turbine Based on Spectral-Clustering-KNN Model

steam turbine; large data; spectral clustering; KNN; deep learning

全体主题 > 涡轮机

传统的诊断方法虽然广泛应用于汽轮机振动的故障诊断，但由于数据量大，往往出现收敛时间长、诊断可靠性低的问题。为了提高大型数据环境下汽轮机故障诊断的准确性，提出了一种基于谱聚类和KNN（SC-KNN）的故障诊断模型。根据火电厂实际运行参数，选取常见的汽轮机故障和汽轮机参数组成特征库，根据实际运行要求设置故障样本比例，对SC-KNN进行训练。训练结果表明，该模型不仅提高了汽轮机故障诊断的准确性，而且在故障间的误判方面也有很好的效果。误差率为标准KNN模型的33%，SVM模型的2.8%。
Though traditional methods are widely applied to fault diagnosis of steam turbine vibration, problems such as long convergence time and low diagnostic reliability often appear because of the large data. In order to improve the accuracy of steam turbine fault diagnosis in large data environment, a fault diagnosis model based on spectral clustering and KNN (SC-KNN) is proposed in this paper. Based on the actual operation parameters of thermal power plants, the common steam turbine faults and steam turbine parameters are selected to form a feature library, and then the proportion of fault samples is set according to the actual operation requirements, on which the SC-KNN is trained. The training results show that the proposed model can not only improve the accuracy of steam turbine fault diagnosis, but also perform very well in misjudgment between faults. The error rate is 33% of the standard KNN model and 2.8% of the SVM model.