Predicting the Remaining Useful Life of a Turbofan Engine using LSTM and Manhattan Distances

The ability to determine Remaining Useful Life (RUL) of a turbofan engine is a foundational ability with regard to predictive maintenance. In this article, Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) have been studied in order to estimate the RUL.The initial RUL is shown to impact RUL predictions. To find the optimal initial RUL, a better method is needed. The method suggested in this article involves creating geometric centers with the data and finding the distance between these centers using the Manhattan norm. The infliction point of the distances between the centers is then used to determine the initial RUL.Furthermore, the LSTM multilayering and the optimal sequence length for the predictions are explored in this paper. Through experimental data evaluation on the Mean Squared Error (MSE), Mean Absolute Error (MAE), and the given scoring function, the Manhattan norm outperforms other methods for determining the initial RUL by more than 100% with the scoring function value of 6274.