TY - JOUR
T1 - Multi-surrogates and multi-points infill strategy-based global optimization method
AU - Ye, Pengcheng
AU - Pan, Guang
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature.
PY - 2023/4
Y1 - 2023/4
N2 - Surrogate-based global optimization (SBGO) methods are widely used to deal with the computationally expensive black-box optimization problems. In order to reduce the computational source, multiple popular individual surrogates containing polynomial response surface (PRS), radial basis functions (RBF), kriging (KRG) and multiple derived ensemble models are constructed to replace the computationally expensive black-box functions. Moreover, a new multi-points infill strategy is presented to accelerate the optimization. New promising points are located by alternately using a hybrid and adaptive promising sampling (HAPS) method and a multi-start sequential quadratic programming (MSSQP) method. The proposed multi-surrogates and multi-points infill strategy-based global optimization (MSMPIGO) method is examined using eighteen unconstrained optimization problems, six nonlinear constrained engineering problems, and one airfoil design optimization problem. Three basic surrogate PRS, RBF, KRG-based global optimization methods using the similar multi-points infill strategy, PRSMPIGO, RBFMPIGO and KRGMPIGO are both considered as the comparative methods. In comparison with PRSMPIGO, RBFMPIGO, KRGMPIGO and three recently introduced SBGO methods, MSMPIGO shows superior search efficiency and strong robustness in locating the global optima.
AB - Surrogate-based global optimization (SBGO) methods are widely used to deal with the computationally expensive black-box optimization problems. In order to reduce the computational source, multiple popular individual surrogates containing polynomial response surface (PRS), radial basis functions (RBF), kriging (KRG) and multiple derived ensemble models are constructed to replace the computationally expensive black-box functions. Moreover, a new multi-points infill strategy is presented to accelerate the optimization. New promising points are located by alternately using a hybrid and adaptive promising sampling (HAPS) method and a multi-start sequential quadratic programming (MSSQP) method. The proposed multi-surrogates and multi-points infill strategy-based global optimization (MSMPIGO) method is examined using eighteen unconstrained optimization problems, six nonlinear constrained engineering problems, and one airfoil design optimization problem. Three basic surrogate PRS, RBF, KRG-based global optimization methods using the similar multi-points infill strategy, PRSMPIGO, RBFMPIGO and KRGMPIGO are both considered as the comparative methods. In comparison with PRSMPIGO, RBFMPIGO, KRGMPIGO and three recently introduced SBGO methods, MSMPIGO shows superior search efficiency and strong robustness in locating the global optima.
KW - Computationally expensive
KW - Global optimization
KW - Multi-points infill strategy
KW - Multi-surrogates
KW - Search efficiency
UR - http://www.scopus.com/inward/record.url?scp=85122244396&partnerID=8YFLogxK
U2 - 10.1007/s00366-021-01557-7
DO - 10.1007/s00366-021-01557-7
M3 - 文章
AN - SCOPUS:85122244396
SN - 0177-0667
VL - 39
SP - 1617
EP - 1636
JO - Engineering with Computers
JF - Engineering with Computers
IS - 2
ER -