Prediction of Remaining Useful Life for Aero-Engines

The information on remaining useful life of any component or equipment is always helpful in being prepared for replacements or maintenance, if any. Engine is considered as the heart of an aircraft and the prediction of its remaining useful life has become a topic of utmost importance. Lot of research is in progress towards this area to safeguard the aircraft and passengers from catastrophic events. Model based approach is unfeasible and time consuming for proposed application and hence data driven approach is employed here. In this paper, an attempt is made to predict remaining useful life (RUL) of aero-engine using Long Short-Term Memory (LSTM) with and without Convolution Neural Network (CNN). In order to analyze and assess the performance of proposed models, benchmark NASA CMAPSS dataset comprising of four different sub datasets is employed. It is observed that the LSTM model without CNN performed better over LSTM model with CNN and the results reported are on par with the results reported in literature on using various other algorithms for RUL prediction using CMAPSS dataset.