基于不同学习算法的RBF神经网络在故障诊断中的应用

基于对RBF神经网络常用的3种学习算法的研究,通过对凝汽器典型故障类型与故障征兆分析,提出了基于不同学习算法的RBF神经网络凝汽器故障诊断,并对诊断结果进行比较.诊断结果表明,基于3种常见学习算法的RBF神经网络都可以准确诊断出凝汽器的各种故障,但聚类方法和OLS算法学习速度要快得多,梯度训练方法速度较慢.研究还表明,RBF神经网络在故障诊断领域具有很好的实用性.

Application of Fault Diagnosis Based on RBF Neural Network with Different Learning Algorithms

RBF neural network, widely used function approximation and classification problems, is simple in structure, simple in training and fast in convergence and can approximate any nonlinear function.In the field of fault diagnosis, it processes good practicality. Through the research of three kinds of learning algorithms for RBF neural network and analysis of typical faults and symptom sets in the operation of the existing condenser , establish a complete learning sample, RBF neural network based on different learning algorithms for the condenser fault diagnosis is presented in this paper, finally, compare the result of the diagnosis. Diagnosis results show RBF neural network based on hree kinds of learning algorithms can accurately diagnose the various fault diagnosis of the condenser. But the study speed of the clustering method and the OLS algorithm is faster, to the contrary,the gradient training method is slower.The research also shows that the RBF neural network has good practicability in the field of fault diagnosis.