故障诊断中的函数连接网络

FL neunal networks for fault diagnosis

函数连接网络通过产生一组线性独立的函数 ,将原输入模式进行一次非线性扩展后作为单层感知器的输入 ,从而克服了学习速度慢 ,易陷于局部极小点的问题 把这种网络用于故障模式分类并和BP网进行比较 。

FL network architecture is a supervised learning extension of the BP network method we have studied so far It allows specification of the category into which inputs will be classified In the designated category, the training set is extended into the input pattern of single layer perceptron with nonlinear change It is high efficient in the aspects of processing speed and avoiding local stability,and especially useful in fault diagnosis [