Research And Application Of Constrained Multi-objective Evolutionary Algorithm Based On Dual-Population Mechanism

Optimization problems exist widely in real-world life.As an important tool for solving optimization problems,evolutionary algorithms(EAs)have gradually attracted more and more attention due to their advantages in various aspects.In optimization problems,constrained multi-objective optimization problems(CMOPs)are a class of problems that exist widely in various engineering.To solve CMOPs using EAs,many constraint handling techniques(CHTs)have been developed.The dual-population mechanism has been rapidly developed due to its excellent effect.However,in most of the proposed algorithms that used the dual-opulation mechanism,auxiliary population usually cannot contribute positively to the output population,and even in some CMOPs,the auxiliary population will waste computing resources.Taking this as the starting point,this thesis takes the dual-population constrained multi-objective evolutionary algorithm(CMOEAs)as the research object,through the research on the attributes and update methods of the auxiliary population,the algorithm design,algorithm effect verification and practical application have been completed.The main contents of this thesis are as follows:1.By introducing the research background and significance of this thesis,the necessity of this research is brought out.Then,the advantages of the dual population mechanism are illustrated by analyzing the shortcomings of each CHT.Then,the CMOEAs based on the dual-population mechanism that have been proposed are introduced.Finally,the defects existing in the above-mentioned dual-population CMOEAs are summarized,and the research idea of this thesis is explained.2.Aiming at the problem that auxiliary population often waste computing resources in dual-population CMOEAs,a dual-population evolutionary algorithm based on dynamic population size is designed.In this algorithm,in order to avoid the situation that the auxiliary population wastes too much computing resources in the later stage of evolution but cannot provide effective help to the main population,a dynamic population reduction mechanism is used to dynamically adjust the size of the auxiliary population,and reduce the computational resources consumed by the auxiliary population,so more resources are provided for the main population to increase diversity.In addition,in order to improve the diversity of the main population in the later stage,an external archive is added to save the feasible individuals found by the auxiliary population.The comparative experimental results on the benchmark function test sets demonstrate the excellent performance of the designed algorithm.Finally,the designed algorithm is used to solve a practical optimization problem,namely the welded beam design optimization problem.The final experimental results demonstrate that the designed algorithm can show better competitiveness in solving real CMOPs.3.Aiming at the problem that the auxiliary population in dual-population CMOEAs cannot provide effective help to the main population in the later stage,a dual-population evolutionary algorithm based on dynamic auxiliary population is designed.A dynamic constraint processing mechanism is used in this algorithm to update the auxiliary population.By processing constraints one by one,the auxiliary population can eventually perform the same task as the main population.Since the way of convergence of the auxiliary population to the optimal solution is different from the main population,the auxiliary population can provide more effective information for the main population,and can still help the main population to improve the diversity in the later stage.In addition,in order to reasonably allocate limited computing resources to the two populations,a dynamic resource allocating scheme is designed.The comparative experimental results on the benchmark function test sets confirm the performance of the designed algorithm.Finally,the designed algorithm is used to solve the vibrating platform design optimization problem,and the experimental results demonstrate that the designed algorithm can better solve real-life CMOPs than other CMOEAs.