摘要
齿轮箱的剩余寿命预测为维修人员做出维修更新决策提供重要信息。为解决在缺少历史数据和非线性非平稳运行状态下的齿轮箱剩余寿命预测难题,提出一种基于数据驱动的齿轮箱的剩余寿命方法。该方法首先根据齿轮箱振动信号特征值,通过状态空间模型(State Space Model,SSM)建立齿轮箱退化状态与特征值之间的关系,来描述齿轮箱的非线性动态变化。其次,当获取到新的信号时,通过扩展卡尔曼滤波(Extend Kalman Filter,EKF)估计准确的模型状态,EM算法(Experience Maximization,EM)估计状态空间模型的参数,根据更新的状态和模型递推预测未来特征值到达故障阈值的时间,从而估计出齿轮箱的剩余寿命。最后,运用齿轮箱全寿命试验数据对预测模型进行检验,实验结果表明该方法能利用实时监测的状态数据准确的预测齿轮箱的剩余寿命,具有较强的工程使用价值和通用性。
An efficient remaining useful life (RUL) can provide key information to form a maintenance and replacement strategies of gearbox. To solve remaining useful life prediction problems of nonlinear and non-stationary process of components, a data-driven approach is presented. The approach constructs a state space model (SSM) to describe degradation evolution process ; uses extend kalman filter to estimate state distribution in SSM and take the Expectation-Maximization (EM) algorithm to update parameters. Relaying on measured data, the time to reach the critical value is determined by estimating the distribution of the remaining useful life by using the estimated nonlinear model. A practical case study of run-to-failure gearbox is present in the last, the results show the approach accurately estimating remaining useful life of gearbox.
出处
《机械强度》
CAS
CSCD
北大核心
2014年第4期614-619,共6页
Journal of Mechanical Strength
基金
"十二五"武器预先研究项目(51327020101)~~
关键词
状态空间模型
EM算法
扩展卡尔曼滤波
剩余寿命
State space model
EM algorithm
Extend kalman filtering
Remaining useful life prediction
作者简介
林国语,男,1988年11月生,福建漳州人,汉族,硕士研究生,研究方向为基于状态的维修,装备维修工程学。