Leveraging the hybrid CNN-BiLSTM and multi-head self-attention mechanisms for Aero-engine RUL estimation

Predicting the Remaining Useful Life (RUL) is pivotal in Aero-engine Prognostics and Health Management (PHM). Deep Learning (DL) techniques have surfaced as a versatile choice for RUL prediction, excelling in handling nonlinear patterns due to their adaptable architectures. While mainstream DL models prioritize overall prediction accuracy, ensuring model reliability stands out as the crucial bottleneck hindering industrial implementation. To this end, this paper is intended to propose a novel CBiLSTM-MA model based on hybrid CNN-BiLSTM and multi-head self-attention mechanisms for reliable RUL prediction with quantified uncertainty. First, the optimal sensor signals are selected based on correlation metrics to be input into the CBiLSTM-MA model. Then, the RUL prediction model is constructed by hybrid CNN-BiLSTM which extracts virtual features and captures forward and backward dependencies from the optimal signals. Moreover, by employing a multi-head self-attention mechanism, the model enhances its pivotal information extraction capability regarding data dependencies. Finally, through comparative experiments, the proposed prediction model is evaluated based on the turbofan aero-engine operation failure prediction benchmark dataset (C-MAPSS). The experimental results show that by deleting some data with low correlation, the proposed model can achieve better prediction accuracy, which is superior to other mainstream models in the references.