Research on Confidence Evaluation Method of Flight Parameter Prediction in Aircraft Navigation System
To enhance the safety performance of aircraft,a confidence assessment process based on predictive modeling is proposed to quantify the uncertainty in flight parameter prediction.A convolutional neural network was employed to develop a prediction algorithm for flight parameter data in aircraft navigation systems.The selected flight parameter data was preprocessed and trained to achieve high-precision target parameter prediction.The modeling process incorporated both epistemic uncertainty in model construction and aleatoric uncertainty in the data by using ensemble methods to capture epistemic uncertainty and employing a dual-head network to capture aleatoric uncertainty.Based on this,a multi-source uncertainty model was constructed to evaluate the confidence of the flight parameter prediction model.Experimental testing,including the introduction of noise into normal flight parameter data to simulate real-world conditions,demonstrates that the proposed confidence assessment method effectively represents the accuracy of the flight parameter prediction results,and improves the safety and reliability of aircraft flight decision-making process.
国家科技期刊平台
NSTL
万方数据
