Recently a computer simulation model via Finite Elements Analysis(FEA) becomes significantly complicate in the field of computer-aided engineering. However, running 3-dimension finite elements model is time-intensive work, hence kriging model as a metamodel has been used to conduct design optimization.
Global metamodel of constraints in a constraint optimization problem require good accuracy around neighborhood of optimal point because optimum result can frequently violate constraints according to inaccuracy of metamodel. To satisfy this requirement, more sample points must be located around the boundary and inside of feasible region (constraint zone). A new sampling strategy capable of identifying feasible domain region should be applied to select sample points for metamodels of the constraints.
In this research, we conduct optimization using a kriging model as a metamodel, based on maximum entropy sampling. For constrained optimization problem, more sampling points should be assigned around the boundary or inside of the feasible region to achieve good accuracy of a metamodel around the feasible region.
To validate the excellence of constraint based maximum entropy sampling, the proposed method is compared with other sampling methods for real optimization problems.