Research On Fast Differential Evolution Algorithm And Its Blasting Application

Differential evolution algorithm has the advantages of being simple,easy to implement,and fast in convergence.It is widely used to solve various optimization problems.However,for complex engineering problems,the algorithm’s solution speed and performance still need to be improved.In this thesis,two variants of the differential evolution algorithm are designed for low-dimensional and high-dimensional optimization problems to accelerate the optimal solution of the algorithm and save the number of function evaluations,successfully solving engineering optimization problems.The main research contents are as follows:(1)An auto-stop and fast differential evolution algorithm is proposed,and the iteration stop criterion that the average diversity of the population tends to 0 is given,and the optimization process is divided into two stages of global and local search,and a matching evolutionary strategy and sorting selection are formulated to generate offspring.The fast convergence and auto-stop of the algorithm are realized.The performance of the algorithm is verified by 14 2-dimensional or 5-dimensional test functions,and it has strong competitiveness in convergence speed.(2)An adaptive differential evolution algorithm with optional external archiving for the joint search of dominant individuals is proposed.The algorithm adds crossover rate sorting and Subset-to-Subset selection operator to ensure that the dominant individuals in the population have a higher probability of surviving to offspring.Further,by adaptively reducing the search space,the search efficiency of mutant individuals is enhanced,and the fast convergence of the algorithm is realized.Experimental comparison with 7 differential evolution variants on 9 30-dimensional test functions proves that the convergence speed of the algorithm is faster.(3)The convergence speed and optimization performance of the two differential evolution algorithm variants are verified by the design parameter optimization task of bench blasting in open-pit mines and the blasting fragmentation prediction model optimization problem.It also solved the problems of unreasonable optimization of blasting parameter design and inaccurate prediction of blasting fragmentation.