Multi-fidelity modeling and adaptive co-kriging based optimization for all-electric Geo satellite systems

All-electric GEO satellite systems design is a challenging multidisciplinary design optimization (MDO) problem, which is computation-intensive due to the employment of expensive simulations. In this paper, the all-electric GEO satellite MDO problem with multi-fidelity models is investigated. The MDO problem involving six inter-coupled disciplines is formulated to minimize the total mass of the satellite system subject to a number of engineering constraints. To reduce the computational cost of the multidisciplinary analysis (MDA) process, multi-fidelity transfer dynamics models and finite element analysis (FEA) models are developed for the geosynchronous transfer orbit (GTO) and structure disciplines respectively. To effectively solve the all-electric GEO satellite MDO problem using multi-fidelity models, an adaptive Co-Kriging based optimization framework is proposed. In this framework, the samples from a high-fidelity MDA process are integrated with those from a low-fidelity MDA process to create a Co-Kriging metamodel with moderate computational cost for optimization. Besides, for refining the Co-Kriging metamodels, a multi-objective adaptive infill sampling approach is developed to produce the infill sample points in terms of expected improvement (EI) and probability of feasibility (PF) functions.