A Kriging-based method for the efficient computation of debris impact zones

To prevent or assess launch risk, evaluation of launchers impact zones is a key element. Several methods are currently used to predict impact zones at the French space agency (CNES), but the highest-fidelity method uses a series of computationally costly Monte Carlo simulations. This process can be very time consuming and the computation time can become prohibitive. A machine learning method called Kriging or Gaussian Process Regression is studied as a potential avenue to speed up the impact zones evaluation. This Kriging-based method, is tested in this paper in different flight phases and its potential for estimating debris impact zones is evaluated in terms of processing time, accuracy and genericity.