Multistage and multivariable optimization for hybrid spatial structures

Abstract

The solution to the multistage and multivariable optimization of hybrid spatial structures by uniform design lacks precision. We introduce iterative operation to approach the real response surface. Several test points are needed in the first experiment design. At each updating step, a few new test points are arranged according to maximum error distribution. New regression response surfaces are fitted by considering the weights of historical data. Compared with traditional experimental design, it needs fewer test points for the same precision. Design verifications demonstrate the reliability and efficacy.