Nonlinear Parameter Estimation for Entry Capsule Dynamic Stability Analysis and Uncertainty Quantification

Identifying dynamic stability coefficients plays a pivotal role in missions that involve atmospheric entry. Although several methods exist to derive these coefficients from experimental and numerical data, there are deficiencies that need to be addressed. In this study, a modular nonlinear parameter estimation (NPE) framework is proposed, which employs a neural network and Markov chain Monte Carlo (MCMC) algorithm to infer the dynamic stability coefficients and identify the uncertainties in the predictions. Given the sensitivity of Markov chains to initial chain location, the neural network is used to generate an initial sample for the first chain that will be used in the MCMC method, where the static and dynamic stability coefficients are estimated along with the uncertainties. The time histories of the moment and angle of attack obtained from computational fluid dynamics simulations with US3D are fed into the NPE framework to estimate the static and dynamic stability coefficients. The results show that the trajectories generated from the estimated coefficients agree with those obtained from the US3D simulations.