B2GRU: Boosted Bidirectional Gated Recurrent Unit for Remaining Useful Life Prediction of Mechanical Bearings
作者单位：Jiangnan Institute of Mechanical and Electrical Design and Research Jiuquan Satellite Launch Center School of Mathematical Sciences Zhejiang University
会议名称：《第35届中国过程控制会议》
会议日期：2024年
学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 080203[工学-机械设计及理论] 0835[工学-软件工程] 0802[工学-机械工程] 0811[工学-控制科学与工程] 080201[工学-机械制造及其自动化] 0812[工学-计算机科学与技术（可授工学、理学学位）]
关 键 词：Remaining Useful Life Prediction Mechanical Bearings Stacked AutoEncoders Gated Recurrent Unit Skip Connection Multiple Scale
摘 要：Modern industrial and production realms heavily rely on rotating machinery, in which mechanical bearings play a crucial role in ensuring equipment reliability and efficiency. Accurate prognostics of these bearings Remaining Useful Life(RUL)can significantly impact the operational effectiveness of the entire machinery. Utilizing comprehensive vibration data, we developed a novel predictive framework by integrating Stacked AutoEncoders(SAE) for advanced feature extraction with a Boosted Bidirectional Gated Recurrent Unit(BiGRU) including Skip Connections for multi-scale joint prediction. This innovative approach, B2GRU, provides precise RUL predictions, significantly enhancing machinery reliability and efficiency while reducing operational downtime and costs. Our experimental results on the PRONOSTIA dataset validate the superior performance of BGRU over contemporary methods, achieving a substantial 42.9% decline in Root Mean Square Error(RMSE) compared to basic Artificial Neural Networks(ANN). This study highlights the robust capabilities of our model in feature extraction and model boosting mechanisms, demonstrating its effectiveness and suitability for RUL prediction of mechanical bearings.