齿轮箱状态识别与剩余寿命预测的MoG-BBN法  

Gearbox's Fault Diagnosis and Residual Life Prediction Based on MoG-BBN Method

作　　者：张星辉[1] 李凤学[1] 赵劲松[1,2] 曹端超[1] 滕红智[3] 

机构地区：[1]军械工程学院 [2]军事交通学院 [3]68129部队

出　　处：《噪声与振动控制》2014年第2期158-163,共6页Noise and Vibration Control

摘　　要：基于混合高斯输出贝叶斯信念网络模型的齿轮箱退化状态识别与剩余寿命预测新方法,应用聚类评价指标对全寿命过程退化状态数进行优化,通过计算待识别故障特征向量的概率值来确定齿轮箱退化状态,在退化状态识别的基础上,提出了剩余寿命计算方法。再利用齿轮箱全寿命实验数据对此进行验证。结果表明,该方法可以有效地识别齿轮箱故障状态并实现剩余寿命预测,平均预测正确率为96.47%,为齿轮箱的健康管理提供参考。A new approach for gearbox’s fault diagnosis and residual life prediction based on Mixture of Gaussian-Bayesian Belief Network (MoG-BBN) is presented. State number optimization method is established based on cluster validity measures. The gearbox’s degradation state can be recognized by calculating the probability of the fault characteristic vectors. On this basis, the method for residual life prediction is presented. Finally, results of this prediction method are verified by the data of gearbox’s full life cycle test. The results show that the mean accuracy can reach 96.47%.

关 键 词：振动与波 混合高斯输出贝叶斯信念网络 状态识别 剩余寿命预测 

分 类 号：TH165.3[机械工程—机械制造及自动化]