对分式凝汽器典型故障模式及其神经网络诊断

Investigation on Typical Fault Pattern of the Halving-style Condenser and Its Fault Diagnosis with BP Neural Network

针对大型火电机组广泛采用的对分、表面式凝汽器，结合其结构、性能特点和现场实际情况，确立 了便于工程应用的对分式凝汽器典型故障知识库。在此基础上采用ＢＰ神经网络方法实现对分式凝汽器故障诊断。提出一种恒误差修正率控制的网络学习率自适应调整方法，大大缩短了网络训练的收敛时间。

The typical fault patterns of the halving-style condenser of the 300 MW thermal power unit are established in this paper. The artificial neural network (ANN) method are used to realize fault diagnosis. A new network training method, based on self-adaptive adjustment of study coefficient controlled by constant error correction rate, is put forward to shorten convergence time.