An efficient variance-based approach in RMDO framework of a space capsule under material and manufacturing uncertainties

Abstract

This paper presents an efficient variance-based Robust Multi-disciplinary Design Optimization (RMDO) framework aimed at minimizing the total mass of a re-entry space capsule under uncertainties, including manufacturing and assembly dimensional tolerances, as well as structural material properties (density, Young's modulus, Poisson's ratio). The method integrates an All-At-Once (AAO) approach within the MDO framework, utilizing a genetic algorithm optimizer. After finding an optimal design point, its robustness is evaluated. If the design is not robust, design constraints are adjusted to move away from the optimal boundary, and the MDO process is repeated until robustness criteria are satisfied. The robustness assessment begins by analyzing the correlation between objective functions and constraints through Design Of Experiment (DOE) using Latin hypercube sampling (LHS) to model input uncertainties. Surrogate Method (Kriging) is used to generate the objective function and constraints. The final step in the RMDO framework evaluates how input uncertainties affect the optimal design and constraints. Results show the RMDO-optimized capsule is 10.7 % lighter than the native version while maintaining safety and stability. Consequently, according to the variance-based RMDO evaluation, the critical constraint functions have at least a 2-sigma uncertainty-based margin of safety in the problem, which confirms the design's robustness.