摘要
为解决齿轮疲劳退化过程中状态突变后剩余寿命难以准确预测问题,提出一种考虑退化突变点检测与剩余寿命预测相关联的齿轮疲劳实时剩余寿命预测新方法。针对齿轮磨损退化过程建立状态空间预测模型,利用接收到的齿轮实时监测振动信息实时更新模型参数,同时对退化过程中的突变状态点进行检测,并根据突变点所提供的寿命信息采用卡尔曼前向滤波及平滑算法结合期望最大化参数估计算法在滤波的同时不断对状态空间模型参数进行修正,改变退化突变后的滤波效果,进行实时状态预测与寿命估计。运用齿轮疲劳寿命试验台的实时监测数据对预测模型进行验证,结果表明利用突变点信息对预测模型进行修正后可以更快的对系统的动态变化进行跟踪,提高预测齿轮退化状态及实时剩余寿命的准确度。
In order to accurately predict the gear remaining useful life in the degradation process,a new method for the real-time prediction of gear contact fatigue remaining useful life was put forward,which is a method integrating the abrupt change detection and remaining life prediction. A state-space model for predicting degradation states of gear wear was established by using the real time monitoring vibration information to update the model parameters. The Kalman forward filtering and smoothing algorithm combined with the parameter estimation by the expectation-maximization algorithm was made in use and the prediction model was modified incessantly to change the filtering effect according to the life information from the abrupt change detection. The real-time monitoring data of the contact fatigue life collected on a gear test rig were used to verify the model proposed. The results show that a revised prediction model using the abrupt point information can achieve the prediction faster than a dynamic tracking system,and can improve the accuracy of gear degradation state and real-time remaining useful life prediction.
出处
《振动与冲击》
EI
CSCD
北大核心
2017年第21期173-184,共12页
Journal of Vibration and Shock
基金
国家自然科学基金(61573250)
山西省青年科技研究基金(201601D021065
201601D021082)
太原科技大学校博士启动基金(20152022)
关键词
剩余寿命预测
状态空间建模
卡尔曼滤波
突变状态检测
模型修正
remaining useful life prediction
state-space models
Kalman filtering
abrupt change detection
model correction