Estimation of RUL for Aircraft Engine Using LSTM

A data driven RUL estimation based on LSTM approach was developed to improve the operational cycles of aircraft engines. In light of the significant quantity of high dimensional time series data given by sensors and the adequate use of this information in the network model. The model could choose the key characteristics from the time-series data, mine the internal connections using the LSTM layer, and then use two completely connected layers to obtain the RUL projected outputs. This method's accuracy beats deep learning techniques like convolutional neural networks (CNN), Random Forest, Linear and Logistic regression methods based on support vector regression (SVM), which offers strong support for judgements about aviation engine operation and maintenance. The NASA provided dataset for aircraft engines was used for verification and analysis in comparison to other algorithms.