摘要
利用现场的运行数据,将基于输入训练神经网络的非线性主元分析(PCA)方法应用到水轮机调节系统传感器故障诊断中,讨论了基于输入训练神经网络的非线性主元分析实现方法,建立了输入训练神经网络和反向传播网络,实现了对实测数据的重构,讨论了利用平方预测误差(SPE)进行故障检测和识别的方法,并用现场实测数据对该方法进行了仿真。仿真结果表明,该方法有效且实用。
A nonlinear principal component analysis methodology based on input-training neural network is proposed and applied to sensors diagnosis of hydro turbine regulating process,which is completed by establishing an input-training neural network and a backpropagation network to reconstruct sensors value.The scheme of fault detection and fault identification is discussed via the application of the squared prediction error（SPE）.Simulating results prove that this method is practically feasible with high fault recognizing rate and application value.
出处
《西安理工大学学报》
CAS
北大核心
2012年第2期204-209,共6页
Journal of Xi'an University of Technology
基金
教育部博士点专项基金资助项目(20096118110012)
