摘要
提出了一种基于在线学习神经网络的故障预报方法。该方法在网络设计过程中结合了“添加”准则和基于对网络输出贡献相对较小的“剪枝”准则。“添加”过程中利用隐层的最大输出判断神经元的活跃性;“剪枝”过程中加入了滑动窗口,避免了误“剪枝”。同时,调整过程只对输出响应比较大的神经元进行,大大减少了计算量,提高了实时性。仿真结果表明,利用该算法能够对一类带时变参数的非线性系统进行故障预报。
A fault prediction method is presented based on neural network on-line learning. The method combines the growth criterion with a pruning strategy based on the relative contribution of each hidden unit to the network output in the process of the network design. The growth process uses the maximum output to judge activation of the neurons. The sliding window is added into the pruning strategy to avoid improper pruning. Meanwhile, the adjusting course is carried out through only some neurons with the larger output, thus leading the reduction of calculating magnitude and the improvement of the real time. Simulation results indicate that the algorithm can predict faults in a class of nonlinear systems with time- varying parameter.
出处
《南京航空航天大学学报》
EI
CAS
CSCD
北大核心
2007年第2期249-252,共4页
Journal of Nanjing University of Aeronautics & Astronautics
基金
国家自然科学基金重点(60234010)资助项目
航空科学基金(05E52031)资助项目
关键词
故障预报
RBF神经网络
在线学习算法
时变参数
非线性系统
fault prediction
RBF neural network
on-line learning algorithm
time-varying parameter
nonlinear system
