Abstract: In view of the defects of design of experiment (DOE) methods based on probability theory, a boundary and best neighbor based intelligent sampling (BBNS) method is proposed. This sampling scheme also can be combined with moving least square (MLS) response surface (RS) metamodel method to optimize nonlinear problems. In order to verify the validity of this intelligent sampling method, this method is adopted to carry out extreme solving of nonlinear test function. Compared with current main DOE methods, this method shows certain improvement in solution precision and in response surface fitting precision. The intelligent sampling scheme coupled with MLS RSM is applied for optimization of draw bead in sheet forming system, and expected results are obtained. It is used successfully in the forming of actual products.

Key words: Intelligent sampling, Metamodel, Moving least square, Optimization, Response surface