Remaining Useful Life Prediction for Turbojet Engines Using Deep CNN on Raw Sensor Data

Accurate Remaining Useful Life (RUL) prediction is critical for optimizing maintenance schedules and reducing operational costs in industries reliant on complex machinery, such as turbofan engines. Traditional RUL estimation methods often require manual feature extraction, which is time-consuming and susceptible to errors. This paper provides a system that utilizes deep learning techniques using Convolutional Neural Networks (CNNs) to autonomously forecast Remaining Useful Life (RUL) from raw sensor time-series data, therefore obviating the necessity for human feature engineering. Trained on the FD001 dataset from the dataset for simulating degradation in turbofan engines, our CNN model significantly reduces prediction error by approximately 25% lower than traditional methods while achieving an average Root Mean Squared Error (RMSE) of 17.0 & standard deviation of 7.01. The model's capacity to manage varying sequence lengths and make accurate end-of-life predictions enhances its reliability and efficiency in real-world predictive maintenance applications. This approach can potentially reduce unplanned downtime, optimize maintenance processes, and improve overall operational efficiency across various industries.