ESA GNC Conference Papers Repository
An Overview of the RFCS V&V Framework: Optimization-Based and Linear Analysis Tools for Worst- Case Search
This article presents the application of nonlinear (simulation-based) and linear (structured singular value) worst-case tools to the VEGA launcher Verification and Validation (V&V) process, during the atmospheric ascent phase. The simulation-based worst-case evaluation is performed by minimizing certain cost functions that are intrinsically aligned with the performance of the launcher using an optimization tool and the high-fidelity nonlinear simulator of the launcher. A numerical sensitivity analysis of these functions is performed prior to the optimization campaigns, in order to select the variables with the largest impact on the criteria. In terms of optimization methods, differential evolution and hybrid differential evolution algorithms are adopted, due to their ability to cope with a large number of optimization variables. The linear analysis uses the structured singular value and a linear fractional transformation model, formed from a subset of the uncertainty and dispersion parameters defined for the nonlinear VEGA system, to identify physically feasible worst-case conditions. This analysis is complementary to traditional Monte-Carlo (MC) campaigns in that it provides an analytically guaranteed existence of worst-case conditions. The simulation-based and the linear analysis approaches adopted in this paper require, in general, only a fraction of the MC runs needed to detect worst-case conditions.