论文部分内容阅读
根据一双跨转子实验台,模拟了转子与静子在轮盘处及轴颈处碰磨、轴系不对中及转子不平衡故障,通过一个信号自动处理装置记录下转子正常振动信号及发生各种故障时的信号,然后利用研制的人工神经网络系统对故障示例进行学习。通过在实际中诊断故障,证明这种是可行的。本文还针对人工神经网络(BP算法)存在的训练速度慢的问题,提出了一个加快网络训练速度的新方法(ARBP算法),较大提高了网络的训练速度。
According to a pair of cross-rotor experimental bench, the simulation of rotor and stator collision at the wheel hub and journal, shaft misalignment and rotor imbalance fault, through a signal processing device to record the normal rotor vibration signal and a variety of failures When the signal, and then use the developed artificial neural network system for fault examples to learn. It is possible to prove this by diagnosing faults in practice. Aiming at the problem of slow training speed of artificial neural network (BP algorithm), a new method to speed up network training (ARBP algorithm) is proposed in this paper, which greatly improves the training speed of network.