基于KPCA和PNN的高加系统故障诊断  

Fault Diagnosis of High-pressure Heater System via Kernel Principal Component Analysis and Probabilistic Neural Network

作　　者：毕小龙[1] 王洪跃[1] 司风琪[1] 徐治皋[1] 

机构地区：[1]东南大学动力工程系,南京210096

出　　处：《汽轮机技术》2006年第5期386-388,392,共4页Turbine Technology

摘　　要：提出了一种新的高加系统故障诊断方法。首先使用核主元分析方法进行特征提取,降低数据维数,既简化了诊断过程,又提高了故障诊断的精度。然后使用概率神经网络进行故障模式识别。该神经网络训练速度快,容易添加新的训练样本。最后将该方法川于某汽轮机组高加系统故障诊断中,取得了较好的诊断效果,表明该方法具有一定的工程实用价值。A novel approach to diagnosing the faults in high - pressure heater system was presented. Firstly, the kernel principal component analysis was employed to extract main features from high dimension patterns by means of kernel trick. Not only was the diagnosing process simplified but also the diagnosing accuracy was ensured. Secondly, the probabilistic neural network （PNN） was utilized to identify the fault mode. PNN can be trained quickly. Moreover, the new trained samples can be added to PNN easily. Finally, the proposed scheme was applied to diagnose the faults in high - pressure system of a turbine unit. The diagnosis results proved that this method was practical in engineering.

关 键 词：高加系统 核主元分析 概率神经网络 故障诊断 

分 类 号：TM267[一般工业技术—材料科学与工程]