Technology selection and architecture optimization of in-situ resource utilization systems

Abstract

This paper discusses an approach to exploring the conceptual design space of large-scale, complex electromechanical systems that are technologically immature. A modeling framework that addresses the fluctuating architectural landscape (an inherent feature of developing technology systems) is applied to the design of a lunar in-situ resource utilization (ISRU) oxygen plant. Four optimization methods using genetic algorithms are compared on both a quadratic-based test function and the ISRU plant design with the goal of balancing the resources spent on exploiting individual architectures and exploring a broad selection of architectures. These include two dual-level approaches that address the discrete architecture design space differently from the continuous sizing design space and two combinatorial approaches that address both the discrete and continuous simultaneously. It was found that the single-level, combinatorial approaches worked better on the real-world ISRU case study, providing a balance between computation time spent on optimizing sizing and performance of each architecture and time spent searching a large number of architectures. For the ISRU architecture search, the single-level approaches on average covered ~300 architectures with ~5000 function evaluations. A heuristic-based dual-level approach covered ~266 architectures with ~5,500 function evaluations.

(cont.) A nested dual-level approach with gradient-based optimization of internal continuous variables nested within a heuristic search of discrete architecture variables would have required on the order of 300,000 function evaluations. The ISRU plant architecture search found that a 300 kg mass ISRU oxygen plant can produce around 1500 kg O₂/year, which is about the amount needed to sustain a crew of four for one year on the lunar surface. These preliminary results also indicate that ISRU plants exhibit an economy of scale of .78, implying that fewer, larger plants would be less costly than many smaller plants in building up a high production capacity.