中文摘要：

提出一种用多目标技术求解约束优化问题的算法.该算法有3个特征：1）将约束优化问题转化为等价的动态约束多目标优化问题,然后用动态约束多目标演化算法求解动态约束多目标优化问题;2）演化初始阶段,拓宽约束边界以使整个种群可行;演化过程中,约束边界微弱的收缩以确保动态约束多目标演化算法中种群的大多数个体仍是可行的,这使动态约束多目标演化算法如同多目标演化算法求解无约束问题一样有效;3）采用基于学习的机制自适应调整演化算法的参数,以提高算法效率.实验结果表明,与4个当前较为先进的约束处理算法相比,本文算法效果更优.

英文摘要：

A novel multi-objective technique is proposed for solving constrained optimization problems COPs. The method highlights three different perspectives： 1） The COP is converted into an equivalent dynamic constrained muhiobjective optimization problem （DCMOP）, and the DCMOP is solved by a dynamic constrained multi-objective evolutionary algorithm （DCMOEA） ; 2） The initial constrained boundary is set large enough so as to obtain a feasible initial population, and the boundary is slightly reduced during the evolving in order to keep most solutions in the population feasible, which guarantees that the DCMOEA performs as effective as that of a multi-objective evolutionary algorithm （MOEA） in solving an unconstrained multi-objective optimization problem; 3） The scale parameter in the mutation is controlled adaptively by learning technique which improves the algorithm performance. Compared with four state-of- the-art methods on benchmark problems, the method proposed in this paper outperforms those.