Truss structure satellite bus geometry-structure optimization involving mixed variables and expensive models using metamodel-based optimization strategy

In order to efficiently resolve the truss structure satellite bus geometry and structure optimization (TSSB-GSO) problems, this paper presents a general framework that integrates commercial CAD/CAE softwares and adaptive metamodel-based optimization strategy. Since TSSB-GSO using expensive FEM analysis models involves continuous-discrete mixed size variables, continuous geometry variables and Boolean topology variables, it is rather computationally intensive and ineffienct to employ traditional global optimization methods to treat such TSSB-GSO problems. Thus this work adopts an adaptive metamodel-based optimization strategy using sequential radial basis function (SRBF-MDC) to improve the computational efficiency for TSSB-GSO problems. This methodology features a sequential sampling method to iteratively update RBF approximation models for both objective and expensive constraints. Through application to practical industrial case, the proposed framework shows the capability of automatically updating geometry and structure models for TSSB-GSO. Moreover, in terms of comparison with other well-known strategy, the SRBF-MDC demonstrates excellent efficiency and remarkable performance of searching global optimum for TSSB-GSO.