TY - GEN
T1 - A CMA-ES enhanced MOEA/D applied to multi-objective aerodynamic optimization design
AU - Zhu, Xinqi
AU - Gao, Zhenghong
N1 - Publisher Copyright:
© 2016, American Institute of Aeronautics and Astronautics. All right reserved.
PY - 2016
Y1 - 2016
N2 - Multi-objective aerodynamic optimization design requires a multi-objective optimization algorithm that has rapid convergence speed and can provide the decision maker with a limited number of non-dominated Pareto solutions according to his or her requirements. To obtain an algorithm meeting such requirements, the CMA-ES is introduced into MOEA/D-DE obtaining MOEA/D-DE+CMA. MOEA/D can approximate different part of Pareto front according to the requirements of the decision maker through altering the weight vector, get evenly spread Pareto solutions without extra management in the algorithm, and perform well even with small population. The improved algorithm adjusts the proportion of CMA-ES and DE used in the next generation: (1) linearly within the evolution process of the algorithm and (2) according to the degree of the scalar improvement of subproblems. The improved algorithm has more rapid convergence speed, and obtains better distributed pareto front, especially when approximating part of the pareto front with respect to user preference. Multiobjective test functions and aerodynamic optimization design of RAE2822 airfoil is tested. The results show that the MOEA/D-DE+CMA outperforms MOEA/D-DE with respect to convergence speed, and gets better Pareto solutions in airfoil optimization case.
AB - Multi-objective aerodynamic optimization design requires a multi-objective optimization algorithm that has rapid convergence speed and can provide the decision maker with a limited number of non-dominated Pareto solutions according to his or her requirements. To obtain an algorithm meeting such requirements, the CMA-ES is introduced into MOEA/D-DE obtaining MOEA/D-DE+CMA. MOEA/D can approximate different part of Pareto front according to the requirements of the decision maker through altering the weight vector, get evenly spread Pareto solutions without extra management in the algorithm, and perform well even with small population. The improved algorithm adjusts the proportion of CMA-ES and DE used in the next generation: (1) linearly within the evolution process of the algorithm and (2) according to the degree of the scalar improvement of subproblems. The improved algorithm has more rapid convergence speed, and obtains better distributed pareto front, especially when approximating part of the pareto front with respect to user preference. Multiobjective test functions and aerodynamic optimization design of RAE2822 airfoil is tested. The results show that the MOEA/D-DE+CMA outperforms MOEA/D-DE with respect to convergence speed, and gets better Pareto solutions in airfoil optimization case.
UR - https://www.scopus.com/pages/publications/85088206463
U2 - 10.2514/6.2016-3517
DO - 10.2514/6.2016-3517
M3 - 会议稿件
AN - SCOPUS:85088206463
SN - 9781624104398
T3 - 17th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference
BT - 17th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference
PB - American Institute of Aeronautics and Astronautics Inc, AIAA
T2 - 17th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, 2016
Y2 - 13 June 2016 through 17 June 2016
ER -