摘要
针对BP算法容易陷入局部极小值、收敛速度慢及容易振荡等缺点,采用小波BP网络且对小波网络采用基于梯度符号变化的局部学习率自适应算法和引入动量项的改进。将改进后的算法对多变量非线性的田纳西-伊斯曼过程进行了仿真研究,结果表明改进算法提高了故障分类的辨识精度。
BP algorithm trends to fall into the local minimum value, slow convergence speed and frequent oscillation. The wavelet BP network was used, and self-adaptive learning rate algorithm based on the sign change of gradient and momentum item were added in it. The improved algorithm was applied in Tennessee-eastman process of a multiple-variable and nonlinear system. The results show that the algorithm can improve the recognition accuracy of fault classification.
出处
《辽宁石油化工大学学报》
CAS
2007年第4期64-67,共4页
Journal of Liaoning Petrochemical University
关键词
小波网络
故障诊断
TE过程
Wavelet network
Fault diagnosis
Tennessee - eastman process
