Control Surface Geometry Surrogate-Based Optimization for Spin-Stabilized Projectile Course Correction

Abstract

In the present paper, a surrogate-based methodology is developed and applied to the design of a control surface placed on an artillery projectile. An artificial neural network is used to estimate the aerodynamic contribution of the control device from numerical simulations. These estimations are coupled to a flight mechanics code to evaluate the trajectories of controlled projectiles. From these six-/seven-degree-of-freedom computations, the trajectories modifications are modeled using kriging. A correction device should minimize the distance to a ellipse (with as the projectile standard deviation of the projectile dispersion) around the mean impact point while ensuring problem-specific constraint fulfillment. The kriging databases are sequentially enriched with the impact point position of the projectile configurations maximizing the expected improvement criterion until a convergence state, assessed by leave-one-out cross validation, is reached. Computational-fluid-dynamics-based evaluations of the optimum configuration coefficients provide a refinement of the neural network databases in the relevant area of the design space. A course correction corresponding to a decrease of two standard deviations in the lateral direction and an increase of two standard deviations in range is achieved, respectively, by reducing the pitching moment of the projectile and increasing its lift-to-drag ratio.