Research And Application Of Improved Adaptive Multi-objective Genetic Algorithm

In the fields of scientific research,engineering applications,and real life,we often encounter a type of problem called multi-objective optimization: this type of problem requires finding a solution that can meet multiple goals simultaneously under given conditions.Many problems can be abstracted as multi-objective optimization problems,and the solution of this problem has received extensive attention from engineering and academia.The NSGA-? algorithm is a classic algorithm for solving multi-objective optimization problems.It has high solution efficiency and can obtain multiple high-quality solutions in one run.It has become the benchmark for performance comparison of other multi-objective optimization algorithms.algorithm.With the widespread application of NSGA-?,some problems have also emerged.First,the operating parameters of NSGA-? remain unchanged during the entire operation process,and they cannot be adaptively adjusted according to environmental changes.Such settings cause the algorithm to iterate.,It is easy to lose the good solution through crossover or mutation,it is difficult to efficiently search the solution space,resulting in a decline in the algorithm's ability to find the best;Secondly,the NSGA-? algorithm uses a simulated binary crossover operator.When using the binary crossover operator to perform crossover operations,it is necessary to set constant parameters to guide the search of the algorithm.However,the actual situation is that the setting of the constant value is subjective and the constant cannot be set separately according to different stages of the algorithm in an experiment,resulting in prematureness in the search.In summary,this paper addresses the above two problems in NSGA-?,introduces adaptive strategies and normal distribution crossover operators into NSGA-? algorithm,and makes the following two improvements to improve the efficiency and improvement of the algorithm.Algorithm convergence and diversity.First,this paper proposes a non-dominated sorting algorithm based on adaptive strategy for the problem that NSGA-? cannot adaptively adjust mutation and crossover probability.The strategy dynamically adjusts the crossover and mutation probability according to the algorithm's running algebra,improves the algorithm's global search capability,and thus plays a role in suppressing premature,accelerating convergence,and increasing population diversity.Second,the analog binary cross method used for the NSGA-? algorithm has a weak search ability and is prone to fall into local optimal problems.This paper uses a method based on the crossover of normal distribution.The feasibility of this method is proved by reasoning and analysis.The adaptive non-dominated sorting genetic algorithm based on the normal distribution cross operator formed by the fusion of the two improvements is proved by the comparison experiment with NSGA-?.ability.You can get higher convergence and diversity,improve the efficiency of the algorithm,and find a better Pareto solution set.In order to solve the problem of multi-objective flexible workshop scheduling,the equipment needs to be initialized.This paper proposes a distribution method based on uniform design.This method combines workpieces,processes,and equipment through a uniform design table to form a corresponding relationship,and finally completes the initial allocation of equipment.At the same time,the improved NSGA-? algorithm is added to solve the multi-objective(FJSP)with the goal of maximum completion time,total delay,total equipment load and total energy consumption.Through multiple simulation experiments,it is proved that the improved algorithm optimizes the performance of the algorithm and improves the ability to solve the optimal solution when solving the multi-objective flexible job shop scheduling problem.It provides an important theoretical basis for applying to solving practical problems.