Dynamic approximate modeling and deviation analysis methods for sounding rocket flight performance

The development of low-cost sounding rockets is considered highly significant for the accurate acquisition of in-situ atmospheric data in near-space environments and the enhancement of the strike accuracy of medium- to long-range missile weapons. However, sounding rockets are often designed as uncontrolled, self-stabilizing rockets, which may be affected by atmospheric wind fields, design flaws, and manufacturing errors during flight missions. This can result in rocket instability or flight altitudes that do not meet design specifications. Therefore, a quantitative uncertainty analysis of the flight performance parameters of sounding rockets is conducted. The impact of different flight environments and rocket parameter deviations on the actual flight performance of sounding rockets is calculated. An evaluation is conducted to determine whether the flight performance parameters, such as the maximum trajectory height, maximum angle of attack, sideslip angle, and maximum dynamic pressure, meet the design requirements under the influence of uncertainties during actual flight. This analysis is used to provide reliable references for actual flight missions.

In this study, the coupling relationships among various rocket disciplines were analyzed, and a multi-disciplinary integrated simulation process for low-cost sounding rockets was established, enabling efficient computation of flight performance. The uncertainty propagation analysis problem of flight performance was addressed by conducting research based on dynamic augmented sampling and surrogate modeling. A bounded sequential augmented Latin hypercube experimental design scheme was proposed, facilitating efficient acquisition of training data for different random variables. An improved augmented radial basis hybrid approximate model was developed and applied to uncertainty propagation analysis. Through dynamic augmented sampling, an uncertainty bias model was established, and the improved hybrid approximate model was used to predict the characteristic parameters of rocket flight performance. Finally, the precision of the predictions was compared with the results obtained from the traditional MCS method to validate the effectiveness of the proposed approach.

In this study, an augmented radial basis function approximate model, based on 200 sample points, was used to predict the flight performance of sounding rockets. The statistical values of the predictions were found to be close to those obtained from 5,000 flight simulations using the MCS method. From the experimental data comparison, it was observed that the prediction accuracy of the statistical mean of the flight performance parameters could reach levels of 1% or even 0.1‰ and the prediction accuracy of the standard deviation could be controlled within 10%. This indicates that an approximate model built using 200 flight simulation samples can achieve high-precision predictions of the statistical values of the flight performance parameters of sounding rockets, with accuracy comparable to that obtained from 5,000 Monte Carlo simulations. Additionally, as the number of samples increases, the ARBF approximate model can be dynamically updated using the BRELHD method, facilitating the assessment of the feasibility of the prediction results.

In this study, the problem of uncertainty propagation analysis for the flight performance of sounding rockets was addressed by investigating a method based on dynamic augmented sampling and surrogate modeling. A bounded sequential augmented Latin hypercube experimental design scheme was proposed for uniform sampling of different distribution random variables, enabling efficient acquisition of training data for different random variables. To enhance the generalization performance of the approximate model, the advantages of both PCE and RBF models were leveraged to establish an improved augmented radial basis hybrid approximate model, which was applied to the uncertainty propagation analysis of sounding rockets. Through dynamic augmented sampling, an uncertainty bias model was established, and the improved hybrid approximate model was used to predict the characteristic parameters of the rocket's flight performance. Finally, the algorithm's accuracy was compared with the results obtained from the traditional MCS method, demonstrating the effectiveness of the uncertainty propagation method for the flight performance of sounding rockets developed in this research.