Preference Ranking on the Basis of Ideal-Average Distance Method for Multi-Criteria Decision-Making

Over the past 2 decades, numerous papers have been published on multi-objective optimization (MOO) in chemical engineering. For conflicting objectives, MOO produces a set of non-dominated optimal solutions that are equally good. Only one of these optimal solutions has to be selected for implementation. Ranking of non-dominated solutions and this selection have been discussed, mostly in the area of multi-criteria decision-making or -analysis (MCDM/MCDA). There is limited research in this area for chemical engineering applications. This paper proposes a novel MCDM method, the PROBID (preference ranking on the basis of ideal-average distance) method, and its simpler variant, the sPROBID. The PROBID method comprehensively considers a spectrum of ideal solutions and the average solution to determine the performance score of each optimal solution. The sPROBID method takes into consideration only the first and last quarters of ideal solutions. Comparative sensitivity analysis is performed, and both the methods are seen to be sensitive to changes in the weights of objectives, which is desirable. Then, the ranking consistency of the PROBID and sPROBID is tested and compared with 7 popular/recent MCDM methods, for 12 scenarios on 3 different data sets from the literature; these scenarios include the linear transformation of measurement unit, equivalent objective formulation, and addition/deletion of optimal solutions in the data set. The results of these tests show that the PROBID method is more consistent and robust compared to the other MCDM methods tested. Although sPROBID is less robust than PROBID, it outperforms four of the seven MCDM methods in terms of ranking consistency.

Automated summary

This paper introduces a novel MCDM method PROBID and its simplified variant sPROBID for multi-objective optimization in chemical engineering. Results show that PROBID is more consistent and robust in ranking compared to other MCDM methods tested, while sPROBID outperforms four out of seven methods in terms of ranking consistency.