Remaining useful life prediction in prognostics using multi-scale sequence and Long Short-Term Memory network⋆

Abstract

The development of sensors and artificial intelligence technology provides practical tools for aircraft Prognosis and Health Management (PHM). The remaining useful life (RUL) prediction is the critical process of PHM. A novel data-driven framework is proposed to estimate the RUL of complex systems in this paper. The framework evaluates the system's RUL based on multi-scale sequences and Long Short-Term Memory (LSTM) networks. First, the sliding time window method is used to prepare training samples, and the degradation features are directly mapped to RUL predictions. In addition, the model parameters are adjusted through the input multi-scale sequence to obtain the best prediction performance. This method integrates the application of time window, multi-scale sequence, and LSTM structure to improve prediction accuracy. The proposed method is validated using the NASA C-MAPSS data set, and the results demonstrate the superiority of the proposed framework.