Lifetime Prediction Method for Space Agencies by Combining Digital Twin and TimeGAN

Space agency lifetime prediction is of great significance for ensuring its safe operation. Aiming to address the issue of insufficient space agency degradation samples, a digital twin and TimeGAN-enhanced data generation method are proposed. The degradation data obtained from macro and micro experiments of key components are utilized and loaded into the space agency's twin model for virtual operation under full working conditions, enabling the augmentation of performance degradation data. On this basis, TimeGAN is used for data augmentation to acquire the complete life cycle data of the space agency. The data is applied to the BILSTM-Attention life prediction method for particle swarm optimization (PSO), which achieves high-precision prediction for the space agency. In comparison to traditional life prediction methods, this approach tackles the challenge of predicting with small sample data.