CAJ | 학술논문

为了改进BP神经网络进行滚动轴承故障诊断时,网络存在收敛速度慢、易于陷入局部极小点的缺点.应用基于Levenberg?Marquardt法对BP网络进行改进,实现了改进后的BP神经网络结合小波包进行滚动轴承故障诊断的方法.首先,利用小波包多分辨率的特点对滚动轴承的振动信号进行分解和重构,计算各子频带能量并进行归一化,构造特征向量.然后,将所得到的特征向量作为两种BP神经网络的输入,即改进后的BP神经网络和常规的BP神经网络.最后,对两种网络进行训练并测试,结合实验数据验证改进方法的可行性.实验结果表明,改进后的BP神经网络不仅可行,同时提高了收敛速度和诊断的精确度.
위료개진BP신경망락진행곤동축승고장진단시,망락존재수렴속도만、역우함입국부겁소점적결점.응용기우Levenberg?Marquardt법대BP망락진행개진,실현료개진후적BP신경망락결합소파포진행곤동축승고장진단적방법.수선,이용소파포다분변솔적특점대곤동축승적진동신호진행분해화중구,계산각자빈대능량병진행귀일화,구조특정향량.연후,장소득도적특정향량작위량충BP신경망락적수입,즉개진후적BP신경망락화상규적BP신경망락.최후,대량충망락진행훈련병측시,결합실험수거험증개진방법적가행성.실험결과표명,개진후적BP신경망락불부가행,동시제고료수렴속도화진단적정학도.
In the process of rolling bearing fault diagnosis based on BP neural network,the network has the shortcoming of slow convergence and is easy to fall into local minima. In response for this shortcoming,BP network was improved based on the Levenberg?Marquardt method to realize fault diagnosis of rolling bearing. First,the rolling bearing vibration signal is decom?posed and reconstructed based on the characteristics of multi?resolution wavelet packet,each sub?band energy is calculated and normalized,and also eigenvectors are constructed. Then,the eigenvectors are regarded as input of two BP neural networks, namely improved BP neural network and the conventional BP ne ural network. Finally,the two networks are trained and tested to verify the feasibility of the improved approach in combination with the experimental data. The experimental results show that the improved BP neural network is feasible,and the convergence rate and the diagnosis accuracy are improved.