Design by adaptive infill sampling with multi-objective optimization for exploitation and exploration

A surrogate-based design optimization with an adaptive sampling technique based on the innovative infill criteria (ISC) is developed, where key issues are mathematically formulated in space and how many sample points infilled to guarantee desirable accuracy at minimal computational cost. The ISC developed study involve multi-objective (MOO) determine infilling separately for exploitation exploration of space, which represent global local surrogate model, respectively. samples found by MOO Pareto front terms variance estimation uncertainty a predicted function value. To dynamically control location count per iteration infilling, two balancing dynamic switching approach developed. selects equally from far ends as well middle it. uses cut-off switch exclusively exploitation, or vice versa adaptively model. Solution optimality computation efficiency present method EGO, compared analytic functions those EGO conventional, multi-point Expected Improvement (q-EI) Latin Hypercube Sampling (LHS) method. gradient-based without using model was also carried out independently comparison purpose solution efficiency. proposed shows greatest efficiency, requiring smallest number training set becomes even compatible For practical problem, high-life multi-element airfoil chosen maximize lift coefficient non-increasing drag constraints. showed about 18% increase force.