基于LSTM-CAPF框架的岸桥起升减速箱轴承寿命预测方法  

RUL Prediction Method for Quay Crane Hoisting Gearbox Bearing Based on LSTM-CAPF Framework

作　　者：孙志伟 胡雄[1] 董凯 孙德建[1] 刘洋 SUN Zhiwei;HU Xiong;DONG Kai;SUN Dejian;LIU Yang(Logistics Engineering College,Shanghai Maritime University,Shanghai 201306,China;Shanghai Marine Diesel Engine Research Institute,China State Shipbuilding Co.,Ltd.,Shanghai 201108,China)

机构地区：[1]上海海事大学物流工程学院,上海201306 [2]中国船舶集团有限公司第七一一研究所,上海201108

出　　处：《上海交通大学学报》2024年第3期352-360,共9页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金(62073213)资助项目。

摘　　要：岸桥起升减速箱轴承的健康状况对港口生产安全具有重要意义.针对岸桥变工况的工作条件,提出一种起升减速箱轴承的剩余使用寿命(RUL)预测框架.首先,对工作载荷进行离散化,并确定工况边界.然后,利用长短时记忆(LSTM)网络模型预测载荷和相应的运行工况.其次,以维纳过程为基础,建立了考虑不同工况下退化率和跳变系数的状态退化函数.最后,利用工况激活粒子滤波(CAPF)方法预测轴承退化状态和RUL.采用NetCMAS系统采集的上海某港口起升减速箱轴承全寿命数据验证了所提出的预测框架.与其他3种预测模式比较表明,所提出的框架能够在变工况条件下获得更准确的退化状态和RUL预测.The health condition of hoisting gearbox bearings of quay cranes is of great importance for the safety of port production.A remaining useful life(RUL)predicting framework for lifting gearbox bearings of quay crane under time-varying operating conditions is proposed.First,the working load is discretized and the condition boundaries are determined.Then,the long short-term memory(LSTM)network model is adopted to predict the load and the corresponding operating conditions.Afterwards,considering the degradation rates and jump coefficients under different operating conditions,the state degradation function is established based on the Wiener process.Finally,the condition-activated particle filter(CAPF)is used to predict the degradation state and RUL of bearings.The proposed prediction framework is verified by the full-life data of the hoisting gearbox bearings in a port in Shanghai collected by the NetCMAS system.A comparison with the other three prediction methods shows that the proposed framework is able to obtain more accurate degradation states and RUL predictions under time-varying operating conditions.

关 键 词：岸桥轴承 剩余寿命预测 长短时记忆网络 工况激活粒子滤波 时变工况 

分 类 号：TH215[机械工程—机械制造及自动化]