Multiobjective optimization of AMR systems

• We apply multiobjective optimization to AMR systems.
• We compare multiobjective optimization with single-objective optimization.
• We compare proposed methods in terms of supercriterion and Euclidian distance.
• Weighting methods are properly balanced yielding good compromise between objectives.
• The distance to a reference goal method allows reaching predefined goals.

In this paper, computational algorithms of different multiobjective optimization techniques and their application for AMR systems are presented. In order to illustrate the effectiveness of the proposed methods, they have been demonstrated on a layered AMR bed, where four objective functions are considered, namely: the temperature span, the cooling energy, the number of cycles to reach steady state and the masse of the MCM. The weighted sum, weighted product, Keeney-Raiffa, distance to a reference goal and ε-constraint methods are described along with a comparative study of the results. The results from the multiobjective optimization methods are evaluated in terms of the supercriterion S and the Euclidean distance N. It is concluded that the results obtained using some methods, like the weighting methods, are properly balanced yielding the best compromise in the presence of conflicting objectives while some other methods, like the distance to a reference goal method and the ε-constraint method, allow to reach predefined goals or to respect additional constraints. Moreover, methods like distance to a reference one or the ε-constraint one need some expertise and insights.