Vertical tail bending moment prediction method based on gradient boosting machine

Vertical tail is a flight safety critical component of aircraft structures, and its failure can lead to catastrophic accidents. Thus, accurately predicting the vertical tail load is crucial for ensuring structural safety and economic efficiency. Existing machine-learning-based load prediction methods often rely on preprocessed data collected under controlled conditions, which ensures accuracy but lacks generalizability due to distribution discrepancies between training and testing datasets. Moreover, most current load prediction approaches are based solely on maneuver actions, neglecting the influence of flight states on loads. To address these issues, this study proposes a Gradient Boosting Machine (GBM)-based maneuver data classification method that accounts for variations in flight states to enhance prediction accuracy across different maneuvers. Furthermore, a feature similarity point supplementation method is introduced to alleviate performance degradation caused by distribution inconsistencies. Visualization results show that the supplemented data exhibit a high degree of overlap with the test set distribution. The prediction results demonstrate that the correlation coefficients for climbing and turning maneuvers improve from 0.74 to 0.85 and from 0.77 to 0.88, respectively, indicating that the proposed method significantly enhances model generalization while maintaining relatively high prediction accuracy under complex practical data conditions.