摘要
利用BP神经网络对转子故障进行建模分析。发挥神经网络的自学能力和联想能力,对非训练样本,做出控制决策,表现非常灵活。可根据实验数据进行网络训练,用新数据进行模型验证。还与概率神经网络(PNN)进行对比验证。实验表明只要选择合适的节点数,BP神经网络比概率神经网络对转子故障有较强的学习和辨识能力,收敛较快,性能稳定,预测结果显著。
Modeling and analyzing the motor rotor fault are carried out by BP-NN （Back Propagation Neural Network）. Playing the abilities of self-learning and image, it makes control decision for non-trained samples. Its performance is very flexible. It is able to carry network training for the experimental data, and the model is verified by the fresh data. Compared with the PNN （Probabilistic Neural Network）, the test results show that the BPNN has more strong learning and identification abilities to motor rotor fault, faster convergence, stable performance and notable prediction.
出处
《煤矿机电》
2014年第4期74-76,80,共4页
Colliery Mechanical & Electrical Technology
关键词
神经网络
转子故障
预测
系统辨识
neural network
motor rotor fault
prediction
system identification
