Research On Krill Herd Optimization Algorithm And Its Application In Flexible Job Shop Scheduling Problem

With the increasingly fierce competition in the global market,the Flexible Job-shop Scheduling Problem(FJSP)has been concerned by more and more people in the field of intelligent manufacturing system.Due to effective production scheduling method,we can make rational allocation of limited resources to get better economic benefits.FJSP is a complex combinatorial optimization problem,which is highly difficult to solve.The krill herd(KH)algorithm,with its superior local development and global exploration ability,has a certain advantage in dealing with scheduling optimization problems.Therefore,this paper uses the krill herd optimization algorithm to solve the single target and multi-target FJSP.The contents of the study are as follows:(1)A discrete krill herd algorithm is proposed to solve the single objective FJSP for minimizing the maximum finish time.First,based on in-depth study of the characteristics of FJSP and KH algorithm,the priority encoding strategy of process and workpiece is presented;Considering the problem of processing conflict at the same time,workpiece is selected according to proximity.Then,temporary resource pool matrix,resource state matrix,process work related matrix are defined,in order to express the information of the machine,the workpiece,the process.Secondly,the insertion,reversal and cross operation are defined to enhance the search ability of individual neighborhood,and the mutation mechanism is proposed to jump out of the local optimal solution.Finally,the simulation results show that the improved algorithm is feasible and effective in solving this problem.(2)In order to solve the problem of local search ability and population diversity in multi-objective FJSP,The KH algorithm based on reverse learning and local learning is proposed.First of all,compared with the ordinary individuals,elite individuals have better search information space,so elite individuals are selected for adaptive dimension reverse learning.Then the local learning of the improved elite krill herd is proposed instead of the random diffusion motion of the standard KH algorithm,the local learning includes the Levy flight distribution of the adaptive step length and the differential mutation operator with the optimized design.Then,compare with the latest improved algorithm on the standard test functions,the experimental results show that the proposed algorithm has better convergence speed and convergence accuracy.At last,a model of FJSP with the goal of makespan,processing cost and processing quality is established,the two layer coding structure of the process and the machine is given,and the superiority of the improved algorithm in solving this problem is verified.