Research And Application Of Many-objective Evolutionary Algorithm Based On Bipolar Preferences Dominance

Multi-Objective evolutionary algorithm (MOEA) is a burgeoning heuristic-based methodfor searching optimal solutions; the main idea is to simulate the phenomenon of survival of thefittest and random information exchange in natural evolution. It has caused wide concern for thecharacteristic of stochastic parallel which suit for solving multi-objective optimization problem(MOP). And one of the current hot issues and difficulties is many-objective optimization.The main contribution of this thesis is to integrate decision maker’s preferences into MOEAto reduce the proportion of non-dominated solutions in high-dimension space, and improve theperformance of classical MOEA in solving many-objective optimization problem. First, a newbipolar preferences dominance relation is proposed by taking the method of TOPSIS and idea ofdividing search space. Then a new algorithm2p-NSGA-II is formed by integrating bipolarpreferences dominance relation into NSGA-II. Its effectiveness is validated by simulationexperiment. Finally, the proposed algorithm2p-NSGA-II is applied into practical engineeringproblem two bar truss design.The main work and innovation of this thesis are summarized as follows:1. In order to articulate decision maker’s preferences more precisely, conception of bipolarpreferences is adopted. And then a new bipolar preferences dominance relation is proposed formany-objective optimization by taking the method of TOPSIS and idea of dividing search space.The proposed dominance relation considers the decision maker’s positive preference andnegative preference simultaneously and creates a strict dominance relation amongnon-dominated solutions, which has ability to reduce the proportion of non-dominated solutionsin population and lead the race to the Pareto optimal area that close to the positive preference andfar away from negative preference.2. A new algorithm for many-objective optimization problems is proposed after integratingbipolar preferences dominance relation into NSGA-II, in which non-dominated sorting anddiversity strategy are redesigned. The performance of2p-NSGA-II is tested on benchmarks oftwo to fifteen-objective test problems. The results of comparative simulation experiments showed that2p-NSGA-II is superior to g-NSGA-II and r-NSGA-II on a whole, no matter theaccuracy of obtained solutions or the efficiency of algorithm.3.2p-NSGA-II is improved by designing constraint handling strategy. The improvedalgorithm is able to solve MOP with constraints and more suitable to practical problem. Whenapplying the algorithm into practical engineering problem two bar truss design, two and threeobjective bar truss design models are solved successfully. And comparison results showed that2p-NSGA-II is superior to r-NSGA-II and R-NSGA-II in convergence speed, obtained solutions’quality and algorithm’s efficiency as a whole.