Article ID: | iaor2009754 |
Country: | United Kingdom |
Volume: | 40 |
Issue: | 3 |
Start Page Number: | 253 |
End Page Number: | 269 |
Publication Date: | Mar 2008 |
Journal: | Engineering Optimization |
Authors: | Blasco X., Sanchis J., Martinez M. |
Keywords: | decision theory: multiple criteria, programming: multiple criteria, heuristics: genetic algorithms |
System design is a complex task when design parameters have to satisy a number of specifications and objectives which often conflict with those of others. This challenging problem is called multi-objective optimization (MOO). The most common approximation consists in optimizing a single cost index with a weighted sum of objectives. However, once weights are chosen the solution does not guarantee the best compromise among specifications, because there is an infinite number of solutions. A new approach can be stated, based on the designer's experience regarding the required specifications and the associated problems. This valuable information can be translated into preferences for design objectives, and will lead the search process to the best solution in terms of these preferences. This article presents a new method, which enumerates these