【摘要】 分析了前向型神经网络动力系统模型 ,根据该模型的特点提出了能够克服传统 BP算法学习速度慢、容易陷入局部极小的新算法。改进后的算法用于滚动轴承故障诊断 ,试验结果表明 ,该算法可以有效缩短网络在训练过程中滞留于局部极小区域的时间 ,大大提高网络的学习速度更多还原

【Abstract】 BP neural networks have such disadvantages as too many learning times and easy getting into local minimum. In order to overcome these shortcomings, in this paper,a dynamical model of feed-forward neural networks is analyzed,and is a new algorithm put forward,which can overcome the shortcoming of traditional BP algorithm such as the occurrence of temporary minimum and total training time is too long. The improved algorithm has been used in the diagnosis for rolling bearings. Results show that the new algorithm can minimize the time of network trapping in a temporary minimum and improve the learning speed greatly.更多还原