Multi-objective Optimization Method Library Development And Applied Research

Most problems in design and engineering need to be optimized have more than only one objective, and they are called multi-objective optimization (MOO) problems. All optimization software at present has too few MOO algorithms and is difficult to manipulate, so they can not solve MOO problems efficiently. Thus it 's of great importance to develop a MOO software which has sufficient MOO algorithms and is easy to use. In this paper, the development of Multi-Objective Optimization Algorithm Library using Objective Oriented Method (OOM) was researched, the whole process of designing and developing the software as well as the using of it in both numerical and practical examples was introduced.Due to the compete among objectives, the solution of MOO problems is not one single "optimal solution" that would optimize all the objectives simultaneously, instead, it's a noninferior set formed by many noninferior solutions. MOO tries to get one noninferior solution that satisfies the decision maker or the whole noninferior set.Multi-Objective Optimization Algorithm Library was developed in Microsoft Visual C++ .NET using C++ programming language. The software was analyzed and designed using OOM, and was divided into some modules such as optimization model and optimization algorithms. The optimization algorithms module contains 11 MOO algorithms that fall into two classes: preference-based methods and generating methods. Preference-based methods, such as weight sum method and max-min method, solve MOO problems by translating it into single objective optimization problems using preference information and getting one final solution. Generating methods need no preference information, they tries to get the whole noninferior set or the approximation of it. Generating methods include all kinds of multi-objective genetic algorithms, multi-objective particle swarm optimization and multi-objective simulated annealing.The library has a powerful Graphical User Interface (GUI) which is simple and effective. Three modes were provided to import MOO models of different characteristics into the library. The GUI reduced the workload in solving MOO problems and makes the optimal design more effective.In the last part of this paper, some numeral test functions were calculated by the software to test the efficiency of the algorithms included, then a reentry vehicle with cruciform flaps were taken as examples to testify the utility of Multi-Objective Optimization Algorithm Library.