基于信噪比的多目标决策方法及其工程应用

求解多目标优化问题得到非劣解集后,为了确定出最终的设计方案,研究简单而有效的多目标决策方法十分必要。为此,提出了一种基于信噪比的多目标决策方法。该决策方法在多目标优化设计中引入稳健设计的思想,运用信噪比的概念在非劣解集中选择最稳健的设计作为最终方案。基于信噪比的多目标决策方法,不仅可以方便地对非劣解进行评价和排序,而且考虑到工程实际中误差的客观存在,具有较大的工程应用价值。文中采用某轻型飞机齿轮箱多目标优化设计作为算例,验证了采用该决策方法的必要性和有效性。

An Effective but Simple Multi-Objective Decision-Making Method Using SN Concept

Aim.The existing methods of multi-objective decision-making for selecting the best out of a very large number of possible solutions are,in our opinion,not sufficiently effective.We now present what we believe to be an effective but simple multi-objective decision-making method.Our method draws on the SN (signal noise) concept first proposed by G.Taguchi.We explain our method first theoretically and then through the example of the speed reduction gearbox of a certain light-weight aircraft.In the full paper,we explain our method in detail; in this abstract,we just add some pertinent remarks to listing the two topics of theoretical explanation:(1) G.Taguchi's SN concept and the brief introduction to Chen's method of calculation~(3]);its subtopics are the calculation method for a single evaluation index(subtopic(1.1),) the SN concept for multiple evaluation indices(subtopic 1.2),and the calculation method for multiple evaluation indices(subtopic 1.3);(2) multi-objective decision-making method using SN concept;in topic 2,we discuss the three advantages of this method,which are that the decision-maker can easily establish preference,that it takes into consideration practically unavoidable deviations,and that it is easily adapted to any multi-objective decision-making problem.Finally,we applied our method to selecting the best solution out of a very large number of Pareto-set(possible) solutions for the speed reduction gearbox of a certain light-weight aircraft.We found that,out of 222 Pareto-set solutions,four violated constraining relationships due to practically unavoidable deviations and if they had not been eliminated from consideration,they would have led to possibly false and dangerous optimization result.