Interval Prediction of Aeroengine Remaining Useful Life Based on BCT-Bayesian Bidirectional LSTM Network

Considering the uncertainty due to the inconsistency of data disturbance and the variation of operation condition, the accurate estimation of remaining useful life (RUL) is challenging. This paper proposes an interval prediction method based on box-cox transformation-bayesian bidirectional long and short time memory (BCT-bayesian-biLSTM) network. The proposed method has two highlights to improve the prediction accuracy. Firstly, BCT is presented to make the different data distribution consistent with the required hypothesis distribution which improves the model uncertainty. Then, a probabilistic model of bidirectional long and short time memory network based on Bayesian deep learning framework is established to extract the temporal correlation features hidden in contextual data. The parameters of BCT-bayesian-biLSTM are expressed as random variables to realize the RUL point prediction and its interval estimation simultaneously. Comparison experiment results with the state-of-art methods show that the proposed method has the smallest prediction error and a narrower prediction interval. It exhibits the advantage of the prediction accuracy and the model generalization ability for unknown distribution data.