Flight parameter prediction for high-dynamic Hypersonic vehicle system based on pre-training machine learning model

Given the harsh operating circumstances, hypersonic vehicles operating at high Mach number demand accurate advanced information of the flight and health state. Flight parameter prediction is a crucial foundation for achieving this requirement. This work addressed the trade-off between prediction accuracy and efficiency by proposing a flight parameter prediction model with the model pre-training and online parameter updating. To create training data, a mechanism model is established. Then, we construct and evaluate three distinct prediction models to increase prediction accuracy. Finally, we conducted comparative validation experiments to compare the prediction performance of the three models. The findings demonstrate that the suggested model greatly raises prediction accuracy without raising model complexity, better balancing prediction accuracy and efficiency. The prediction accuracy of the suggested model has increased by 81.9% when compared to the traditional model.