Surrogate Model-based Optimization of Spacecraft Design

Conceptual design optimization of spacecraft systems is complex and multidisciplinary process. In this case evaluation of the objective functions relies heavily on running iterative simulation models and analysis codes between various subsystems (such as structures, payload, electrical power supply, attitude determination and control, communication, command and data handling).The conventional sequential optimization approaches to such a complex design problem is time consuming, which it does not guarantee to achieve the best compromise among the various competing coupled subsystems and may even lead to non-optimal design. In addition, the design search space can be multi-modal, non-convex with multiple local minima and hence it is time consuming or difficult to rapidly evaluate trade-offs between various subsystems (disciplines). To address these issues, in this paper an efficient surrogate (response surface) model-based multidisciplinary spacecraft systems design optimization technique with discrete and continuous design variables is presented. The methodology is based on the utilization of genetic algorithms (GA) for both system level and discipline level as an optimizer. Surrogate-modeling as an efficient tool is also used to decrease computational cost, in discipline (subsystem) level within a collaborative optimization (CO) framework. Results obtained in this study, show that the method introduced in this paper provides an effective way of improving computational efficiency of a complex space system design such as conceptual design optimization of a spacecraft.