Full-scale fatigue testing plays a crucial role in verifying aircraft structural design and durability, but involves long test cycles, numerous sensors and complex operating conditions. Multi-channel data loss often occurs due to uncertainties in acquisition links and external interference, directly affecting data integrity and reliability of subsequent evaluations. To address this issue, the authors proposed a self-attention-based multi-channel imputation model for full-scale fatigue data. This model models long- and short-term dependencies through two-layer diagonal masked self-attention, focusing on intra-channel details and depicting inter-channel global correlations. It also explicitly integrates attention weights with missing masks via weighted combination to adapt to different missing mechanisms and noise conditions. In terms of learning strategy, the authors adopted dual-objective joint optimization of masked imputation and observation reconstruction, along with a hierarchical weighted loss function. These designs enable the model to balance local imputation accuracy and global temporal consistency, suppress information leakage and improve robustness. Based on real full-scale fatigue test data, comparative experiments and ablation experiments were conducted under scenarios with various missing patterns and missing rates. The results demonstrate that the proposed method is effective and stable, significantly enhances the integrity of fatigue test data, and provides a reliable data foundation for fatigue anomaly detection and life assessment.