Dynamic Multi-objective Squirrel Search Algorithm And Its Application

In order to further extend the application of the Squirrel Search Algorithm(SSA)in evolutionary computing and engineering field,this paper proposes a dynamic multi-objective Squirrel Search Algorithm,the proposed algorithm is used to solve the Flexible Job Shop Scheduling Problem(FJSSP).Generally,dynamic multi-objective optimization algorithm is composed of dynamic processing technology and static multi-objective optimization algorithm in transient environment,while static multi-objective optimization algorithm is composed of multi-objective framework and core evolutionary strategy.The core evolutionary strategy is used to update the population,which is the basis of the dynamic multi-objective optimization problem and its convergence directly affects the capability of the dynamic multi-objective optimization algorithm;the multi-objective framework is used to preserve the non-dominated solutions,which directly affects the convergence and distribution of the Pareto Optimal Front(POF)in transient environment;dynamic processing technology is used to deal with the environment changes,which directly affects the timeliness of solving dynamic multi-objective problems.Therefore,this paper improves the single-objective SSA,combines multi-objective framework and dynamic processing technology to construct dynamic multi-objective Squirrel Search Algorithm,the proposed algorithm is applied to solve FJSSP,the details are as follows.Firstly,in order to improve the convergence of the core evolutionary strategy SSA,an Improved Squirrel Search Algorithm for global function optimization(ISSA)is proposed.ISSA contains two searching methods,one is the jumping search method,the other is the progressive search method.Besides,the "escape" operation and the "death" operation are introduced into the jumping search method,the mutation operation is introduced into the progressive search method.The practical method used in the evolutionary process is selected automatically through the linear regression selection strategy.Experimental results show that the convergence of ISSA is much better than the other four algorithms.Secondly,in order to improve the convergence and distribution of the obtained POF when SSA in used to solve multi-objective optimization problems in transient environment,this paper proposes the Multi-objective Improved Squirrel Search Algorithm based on Decomposition with External Population and Adaptive Weight Vectors Adjustment(MOEA/D-EWA-ISSA),the proposed algorithm takes the Multi-objective Evolutionary Algorithm Based on Decomposition(MOEA/D)as the multi-objective framework and takes SSA as the core evolutionary strategy.MOEA/D-EWA-ISSA establishes external populations,and external individuals participate in the individuals' updating of ISSA,each weight vector is adaptively adjusted by the actual evolutionary direction of POF and its neighbor weight vectors.The experimental results show that the convergence and distribution of POF obtained by MOEA/D-EWA-ISSA are much better than other three algorithms,especially in solving complex multi-objective problems.Thirdly,in order to improve the adaptability of the environments when SSA is used to solve dynamic multi-objective problems,this paper uses multi-objective Squirrel Search Algorithm to solve multi-objective problems in transient environment and integrates dynamic processing technology,proposes Dynamic Multi-objective Squirrel Search Algorithm based on Decomposition with Evolutionary Direction Prediction and Bi-Directional Memory Populations(DMOISSA/D-P&M).DMOISSA/D-P&M uses the modified vector to predict the evolutionary direction in the new environment and uses the memory population to retain the evolutionary information in the historical environments.The modified vector and memory individuals participate in the individuals' updating of ISSA at the same time.The experimental results show that convergence and distribution of POF obtained by DMOISSA/D-P&M are better than other three algorithms,and the adaptability of the environments of DMOISSA/D-P&M is better as well.Fourth,the application of dynamic multi-objective Squirrel Search Algorithm in FJSSP.In this paper,the characteristics of FJSSP are analyzed comprehensively and a new mathematical model is established.The new model takes the completion time,the load balance,the average deviation between new and old scheduling schemes as the optimization objectives,and the dynamic multi-objective SSA algorithm is used to solve the established model.The experimental results show that,compared with other mathematical model,the optimized scheduling scheme of the established model has higher efficiency and better stability,especially when the machines are out of order or the breakdowns are recovered,the established model keeps better balance between efficiency and stability.