导航

ACTA AERONAUTICAET ASTRONAUTICA SINICA ›› 2023, Vol. 44 ›› Issue (5): 226918.doi: 10.7527/S1000-6893.2022.26918

• Solid Mechanics and Vehicle Conceptual Design • Previous Articles     Next Articles

Remaining useful life prediction based on multi-scale adaptive attention network

Bin LIU(), Jing XU, Meiling HUO, Xueying CUI, Xiufeng XIE, Donghui YANG, Jia WANG   

  1. School of Applied Science,Taiyuan University of Science and Technology,Taiyun 030024,China
  • Received:2022-01-10 Revised:2022-03-11 Accepted:2022-03-29 Online:2023-03-15 Published:2022-04-06
  • Contact: Bin LIU E-mail:liubin@tyust.edu.cn
  • Supported by:
    National Natural Science Foundation of China(11701406);Fundamental Research Program of Shanxi Province(202103021224274);Research Project Supported by Shanxi Scholarship Council of China(2022-163);Social and Economic Statistical Research Project in Shanxi Province(KY[2022]73)

Abstract:

Remaining useful life prediction is the key to and premise of intelligent maintenance decision-making of complex equipment systems. The potential relationship among different index data brings challenges to prediction accuracy, while parameter selection also increases model prediction difficulty. We use the multi-scale adaptive attention network method to fuse the feature relationship among data from vertical and horizontal dimensions, respectively, and give the piecewise nonlinear target function to improve the prediction accuracy and reduce the root mean square error. The adaptive mechanism is employed to automatically select the size of convolution kernel, improving the efficiency of network training. Empirical analysis of Company-Modular aero-propulsion system simulation data sets prove the effectiveness of this method in remaining useful life prediction of complex systems.

Key words: remaining useful life, multi-scale learning, adaptive attention, nonlinear target function, complex systems

CLC Number: