Remaining Useful Life prediction of Aircraft Engines Using DCNN-BiLSTM with K-means Feature Selection

Predicting the remaining useful life of aircraft components has emerged as a crucial research focus in the field of aviation maintenance. Traditional methods typically rely on empirical rules and statistical models. However, these methods often face challenges in accurately forecasting the intricate lifespan of components. In recent years, with the rapid advancements in deep learning technology, there has been a growing inclination to apply these techniques to predict the remaining useful life of aircraft components. In this paper, we propose a novel approach to predict the Remaining Useful Life of aircraft engines. Our method integrates the K-means algorithm to reveal meaningful data categories, enhancing data preprocessing efficiency. The processed data is then input to the DCNN model to extract spatial features. Furthermore, we harness the bidirectional time series encoding capability of BiLSTM to capture both long-term and short-term temporal dependencies in the data. Ultimately, regression is utilized to accurately predict the components’ remaining useful life. Experimental results using the well-known aircraft remaining useful life C-MAPSS dataset reveal that our approach outperforms several baseline models in terms of predictive performance (the most significant improvement over a single CNN is recorded at 111%). The Root Mean Squared Error (RMSE) values for the four subsets of C-MAPSS are recorded as 13.54, 15.94, 11.56, and 13.77, respectively.