Surrogate Assisted Computation of the Parametric Safety Margin for a Flexible Launcher


A. Kamath (GE Global Research Centre)
P. P. Menon (College of Engineering, Mathematics and Physical Sciences, University of Exeter)
M. Ganet-Schoeller, G. Maurice (Ariane Group)
S. Bennani (Guidance, Navigation and Control Section, ESA/ESTEC)

PDF icon AL13-09_0.pdf1.38 MB

In order to assess the robustness of dynamical systems, an approach is to demarcate the uncertain parameter space as safe set and unsafe set. Unsafe set represents the region within which the system lacks the required level of performance, or even loses its stability. However, determining the minimum distance metric for the unsafe set from the nominal operating point, the so-called parametric safety margin, for a higher dimensional dynamical system is not trivial and is often computationally demanding. In this paper, the parametric safety margin for a closed loop industrial standard launch vehicle simulator during its thrust vector control phase is computed. Imposing certain basic topological restrictions for the multi-dimensional uncertain parameter space, the computation of the parametric safety margin can be posed as a constrained nonconvex global optimization problem, and is thus extremely challenging in the case of high-fidelity aerospace simulators. Various performance requirements become the constraints in the optimization problem. An approach exploiting the use of nonintrusive polynomial surrogate modeling is proposed for the efficient computation of the parametric safety margin for the industrial standard launch vehicle simulator.