RUL prediction for aeroengines with dispersedly clumped gas path parameters based on callback model

The prediction of remaining useful life (RUL) for aeroengines is a critical task in condition-based maintenance. The measurement values of gas path parameters from aeroengines of the same type in service are dispersed within a certain range due to assembly tolerances, operational conditions, and maintenance. This dispersion makes it difficult to predict the RUL of all these engines. To address this issue, a RUL prediction solution for aeroengines with dispersedly clumped parameters based on the callback network is proposed. Initially, a convolutional long short-term memory network with multi-window output is proposed to predict RUL, which can extract information from multiple time windows simultaneously to enhance the capture of important features. Next, a callback model is proposed that achieves feature reuse through a callback mechanism and selects the result from the optimal window among all windows as the final prediction. Additionally, a novel algorithm is developed for selecting the best time window to optimize the prediction performance. To assess the effectiveness of the proposed method, numerical simulations are conducted. The simulation results demonstrate that the proposed method outperforms the other methods in terms of RUL prediction.