A hybrid approach of adaptive surrogate model and sampling method for reliability assessment in multidisciplinary design optimization

Abstract

Uncertainty is an inherent element of multidisciplinary design optimization (MDO), often leading to undesirable performance and potentially infeasible designs. Reliability-Based Multidisciplinary Design Optimization (RBMDO) aims to deliver solutions that achieve desirable performance metrics while remaining resilient to uncertainties. However, the RBMDO process is computationally intensive and can be impractical for launch vehicle (LV) design optimization. This paper presents an innovative hybrid approach that integrates Adaptive Response Surface Methodology (ARSM) with Directional Sampling (DS) to enhance the efficiency of reliability analysis. The ARSM-DS method yields faster and more effective results compared to traditional Monte Carlo Simulation (MCS) techniques. The study specifically focuses on the reliability assessment of a two-stage launch vehicle in its conceptual design phase. The methodology encompasses several critical steps: defining the reliability problem, identifying potential failure modes, establishing target reliability at the system level, modeling reliability, allocating reliability to subsystems, formulating the RBMDO problem, analyzing subsystem reliability using the ARSM-DS method, conducting multidisciplinary optimization based on reliability criteria, predicting overall system reliability, and evaluating computed reliability against the established target. This approach not only enhances the reliability analysis process but also significantly increases the feasibility of design optimization efforts in aerospace applications.