基于RBF神经网络的凝汽器故障诊断研究

Study on Fault Diagnosis of Condenser Based on RBF Neural Network

针对BP网络的训练收敛速度慢,网络初值对学习性能影响比较大的缺陷,提出了一种基于RBF神经网络的故障诊断方法。介绍了RBF的网络结构和训练方法,并应用于凝汽器故障诊断中。通过对现有凝汽器运行中常见的典型故障及征兆集的分析,建立了完备的学习样本。通过实例证明,RBF网络训练速度快,分类性能良好,在故障诊断领域具有很好的实用性。

Aiming at the insufficiency of BP networks, such as the low learning convergence speed and instability learning performance caused by initial value, this paper proposed a new diagnosis method based on RBF neural networks. The RBF networks structure and training algorithm applied to the condenser fault diagnosis was introduced. The typical learning sample was established. Utilizing MATLAB neural network toolbox monitors and diagnoses the condenser status. Example verification indicated that RBF networks have very high learning convergence speed and better classified performance and RBF networks have good practicality in the field of fault diagnosis.