Remaining useful life prediction of aero-engines based on CR-GAN and ConvLSTM

Reliable forecasting of the remaining useful life (RUL) of aero-engines is essential for maintaining operational safety and for reducing overall maintenance expenses. However, this task relies heavily on the availability of sufficient run-to-failure data, which are often costly and difficult to obtain. To address these limitations, a Convolutional Recurrent Generative Adversarial Network (CR-GAN) is proposed. The CR-GAN framework is utilized to produce synthetic data of high fidelity. The artificial samples are integrated with real measurements to build an expanded training set, which is then applied for training a ConvLSTM driven RUL prediction model. Validation of the method is carried out on the CMAPSS dataset, where its performance is examined through comparative analyses and ablation experiments. The findings indicate that the proposed strategy attains superior prediction accuracy, reflected by notable enhancements in both the root mean square error and Score indicators.