该文在函数不一定下半连续,集合不一定是闭集的条件下,利用函数次微分性质,引进新的约束规范条件,等价刻画了鲁棒复合优化问题的最优性条件以及原问题与其松弛型Fenchel-Lagrange对偶问题之间的全对偶。In the case when the functions ...该文在函数不一定下半连续,集合不一定是闭集的条件下,利用函数次微分性质,引进新的约束规范条件,等价刻画了鲁棒复合优化问题的最优性条件以及原问题与其松弛型Fenchel-Lagrange对偶问题之间的全对偶。In the case when the functions are not necessarily lower semicontinuous and the sets are not necessarily closed, by using the properties of subdifferential of functions, we introduce some new weaker constraint qualifications. Under those constraint qualifications, the total duality and optimality condition between the robust composite convex optimization problem and its relaxed Fenchel-Lagrange dual problem are established.展开更多
A new strategy is presented to solve robust multi-physics multi-objective optimization problem known as improved multi-objective collaborative optimization （IMOCO） and its extension improved multi-objective robust c...A new strategy is presented to solve robust multi-physics multi-objective optimization problem known as improved multi-objective collaborative optimization （IMOCO） and its extension improved multi-objective robust collaborative （IMORCO）. In this work, the proposed IMORCO approach combined the IMOCO method, the worst possible point （WPP） constraint cuts and the Genetic algorithm NSGA-II type as an optimizer in order to solve the robust optimization problem of multi-physics of microstructures with uncertainties. The optimization problem is hierarchically decomposed into two levels： a microstructure level, and a disciplines levels, For validation purposes, two examples were selected： a numerical example, and an engineering example of capacitive micro machined ultrasonic transducers （CMUT） type. The obtained results are compared with those obtained from robust non-distributed and distributed optimization approach, non-distributed multi-objective robust optimization （NDMORO） and multi-objective collaborative robust optimization （McRO）, respectively. Results obtained from the application of the IMOCO approach to an optimization problem of a CMUT cell have reduced the CPU time by 44% ensuring a Pareto front close to the reference non-distributed multi-objective optimization （NDMO） approach （mahalanobis distance, D2M =0.9503 and overall spread, So=0.2309）. In addition, the consideration of robustness in IMORCO approach applied to a CMUT cell of optimization problem under interval uncertainty has reduced the CPU time by 23% keeping a robust Pareto front overlaps with that obtained by the robust NDMORO approach （D2M =10.3869 and So=0.0537）.展开更多
文摘该文在函数不一定下半连续,集合不一定是闭集的条件下,利用函数次微分性质,引进新的约束规范条件,等价刻画了鲁棒复合优化问题的最优性条件以及原问题与其松弛型Fenchel-Lagrange对偶问题之间的全对偶。In the case when the functions are not necessarily lower semicontinuous and the sets are not necessarily closed, by using the properties of subdifferential of functions, we introduce some new weaker constraint qualifications. Under those constraint qualifications, the total duality and optimality condition between the robust composite convex optimization problem and its relaxed Fenchel-Lagrange dual problem are established.
文摘A new strategy is presented to solve robust multi-physics multi-objective optimization problem known as improved multi-objective collaborative optimization （IMOCO） and its extension improved multi-objective robust collaborative （IMORCO）. In this work, the proposed IMORCO approach combined the IMOCO method, the worst possible point （WPP） constraint cuts and the Genetic algorithm NSGA-II type as an optimizer in order to solve the robust optimization problem of multi-physics of microstructures with uncertainties. The optimization problem is hierarchically decomposed into two levels： a microstructure level, and a disciplines levels, For validation purposes, two examples were selected： a numerical example, and an engineering example of capacitive micro machined ultrasonic transducers （CMUT） type. The obtained results are compared with those obtained from robust non-distributed and distributed optimization approach, non-distributed multi-objective robust optimization （NDMORO） and multi-objective collaborative robust optimization （McRO）, respectively. Results obtained from the application of the IMOCO approach to an optimization problem of a CMUT cell have reduced the CPU time by 44% ensuring a Pareto front close to the reference non-distributed multi-objective optimization （NDMO） approach （mahalanobis distance, D2M =0.9503 and overall spread, So=0.2309）. In addition, the consideration of robustness in IMORCO approach applied to a CMUT cell of optimization problem under interval uncertainty has reduced the CPU time by 23% keeping a robust Pareto front overlaps with that obtained by the robust NDMORO approach （D2M =10.3869 and So=0.0537）.
