Multi-head attention-based variational autoencoders ensemble for remaining useful life prediction of aero-engines

Yuxiao Wang,Chao Suo,Yuyu Zhao
DOI: https://doi.org/10.1088/1361-6501/ad8b62
IF: 2.398
2024-10-26
Measurement Science and Technology
Abstract:Accurate remaining useful life (RUL) prediction of aero-engines through condition monitoring (CM) data is of great significance for flight reliability and safety. Although deep learning-based approaches have been widely considered, individual deep learning (DL) models suffer from significant stochasticity and limited generalizability when predicting the RUL. To solve this issue, a novel multi-head attention-based variational autoencoders (MHAT-VAEs) ensemble model is proposed. Two distinct MHAT-VAEs are designed, employing linear and convolutional operations to capture global and temporal compressed representations of the CM data. Additionally, a dual-level ensemble strategy is introduced to adaptively fuse the outputs of the two base learners. A hyperparameter optimization method is also implemented to further enhance the efficiency and performance of the base learners. The effectiveness of the proposed method is validated using the C-MAPSS and N-CMAPSS datasets, with experimental results showing that it outperforms state-of-the-art approaches.
engineering, multidisciplinary,instruments & instrumentation
What problem does this paper attempt to address?