Comparing Two Multidisciplinary Optimization Formulations of Trimmed Aircraft Subject to Industry-relevant Loads and Constraints

Abstract

Industry-relevant multidisciplinary design and optimization (MDO) processes engage numerous computationally intensive, disciplinary analysis and design, tasks. This motivates not only the use of high-performance computing resources for tackling the design and optimization problems, but also the search for efficient MDO algorithms and formulations. On the algorithmic side, the MDO chains developed in this work are gradient- based. Such algorithms are known for their efficiency in finding optima when compared to gradient-free algorithms. Concerning the formulations, several studies in the literature showed the effect of MDO formulation on the convergence efficiency. However, these studies used mostly model problems rather than realistic aircraft design problems. The target of our study is to validate a promising data-driven scalable methodology for testing MDO formulations by comparing the obtained results to the results of realistic MDO tasks that are provided and defined by AIRBUS. Two MDO formulations are highlighted and investigated. The first one is driven by one optimizer that controls all design parameters. The second is driven by two optimizers; one responsible for aeroelastic performance analysis and the other responsible for the structure sizing process. The work is performed on the research aircraft configuration; AIRBUS’ XRF-1. The first results of the realistic optimizations showed that engaging two optimizers is more efficient, for the given objective function. The final paper will validate these results carefully and conclude on the reliability of the data-driven scalable methodology.