基于模糊SOM神经网络的汽轮机通流部分故障诊断研究

Study on Fault Diagnosis of Steam Turbine Flow Passage based on Fuzzy SOM Neural Network

汽轮机实际运行时通流部分经常出现故障,对其进行监测与诊断很有必要。以某600MW机组为对象,提出一种基于模糊理论与SOM神经网络相结合的故障诊断方法。该方法通过对故障样本进行训练,建立故障诊断模型,根据输出神经元在输出层上的位置对故障进行诊断,分析其所属故障模式。研究结果表明:基于SOM神经网络的汽轮机通流部分故障诊断方法是准确和可靠的。该方法克服了故障样本选取的困难,通过聚类功能,对故障模式分类实现了具体数字化和图形可视化,诊断结果简单和直观。

It is necessary to monitor and diagnose the faults of steam turbine flow passage that frequently occur during the operating time. The fault diagnosis method for 600MW unit based on fuzzy theory and SOM neural network was proposed. In the method, the fault samples are trained to establish a fault diagnosis model which diagnoses the failures based on the position of output neurons in the output layer and the mode of fault is analyzed. The result shows that fault diagnosis method of steam turbine flow passage based on SOM neural network is accurate and reliable. The difficulty of the selection of fault samples is overcome, and specific digitization and graphic visualization for fault mode classification are achieved by clustering function, that make the diagnosis result simple and intuitive.