基于RBF和优化Wiener模型的轴承剩余寿命预测

针对滚动轴承寿命准确预测缺乏表征其健康状态的可靠退化指标的问题,提出径向基(RBF)神经网络及带有漂移参数的维纳(Wiener)模型进行剩余寿命预测。首先,使用小波包奇异谱熵提取轴承振动信号初始特征;其次,利用早期无故障样本特征和失效样本特征训练RBF神经网络模型,将已提取特征全寿命数据输入到RBF神经网络模型,计算隶属度,作为轴承退化指标;最后,根据滚动轴承的退化轨迹,选择不同Wiener模型进行退化建模,根据AIC信息准则和对数似然值选择合适的模型,利用极大化轮廓似然函数在线更新模型参数,预测轴承寿命。结果表明,所提出的轴承退化指标能够表征健康状态,基于该退化指标的Wiener模型能够准确预测轴承的剩余寿命。

Remaining Life Prediction of Bearings Based on RBF and Optimized Wiener Model

Based on the lack of reliable degradation indicators for the health states of rolling bearings to accurately predict their life.A radial basis function(RBF)neural network and a Wiener model with drift parameters are proposed for remaining life prediction.First,wavelet packet singular spectrum entropy is used to extract the initial characteristics of bearing vibration signal.Then,the early fault-free sample characteristics and failure sample characteristics are used to train the RBF neural network model,and the extracted feature life data are input into the RBF neural network model to calculate the membership degree,which is used as the degradation index of bearing.Finally,according to the degradation trajectory of the rolling bearing,different Wiener models are selected for degradation modeling.Appropriate models are selected according to AIC information criterion and logarithmic likelihood value.The maximum contour likelihood function is used to update model parameters and predict bearing life.The results show that the bearing degradation index proposed by the method can represent the health state,and the Wiener model based on the degradation index can accurately predict the remaining life of the bearing.