Study On Grey Wolf Optimizer For Service Composition Optimization In Cloud Manufacturing

The deep integration of manufacturing and advanced information technology has facilitated the emerging of cloud manufacturing(CMfg).As a key component of CMfg system,service composition and optimal-selection(SCOS)problem bear new challenges in application scenarios and optimization algorithms.On the one hand,to follow the high quality and sustainable development of manufacturing,the energy consumption and robustness of service in CMfg should be extended to service composition scenarios.For another,in terms of the extended application scenarios,efficient service composition throws new requirements for the performance of optimization algorithm.In this paper,three types of service optimization scenarios are formalized in view of the increase of service combination scenarios,the high-dimensional expansion of optimization problems,and the demand for efficient optimization algorithms.Then,several evolutionary algorithms based on the grey wolf optimizer are developed to address these problems.More contents are as follows:(1)Firstly,the connotation and characteristics of CMfg is introduced,the key research problems in CMfg system are summarized,and the importance and complexity of SCOS problem is emphasized.Then,the research status of SCOS problem and optimization algorithm are surveyed,the deficiencies in existing studies are pointed out,and the research contents of this paper are given.(2)Firstly,manufacturing resources of CMfg is introduced,the process of cloud service combination optimization is analyzed,the definition of SCOS problem is given,and the mathematical model of SCOS problem is formulized based on quality of service(Qo S)evaluation in Web services.Then,the basic principle of grey Wolf algorithm and the coding scheme corresponding to SCOS problem are studied,and the scenarios of SCOS problem are analyzed,which provides theoretical support for subsequent research.(3)For the energy consumption during service composition,based on Qo S evaluation,the single objective optimization SCOS model considering manufacturing energy consumption and logistics energy consumption is established,and an improved grey wolf optimizer is proposed.Firstly,an adjusting control parameters is designed to enhance the exploration of the algorithm,so as to improve the accuracy of the solution.Then,an improving position-updating strategy is proposed to increase the diversity of population,which helps to escape from the local optimal.Finally,comparative studies on SCOS problems are carried out to verify the proposed approach.(4)Considering the subjectivity of weight in single objective optimization,single objective SCOS problem is extended as a multi-objective optimization problem.A bi-objective SCOS model involving maximum Qo S and minimum energy consumption is established,and an enhanced multi-objective grey wolf optimizer is developed.Firstly,a new population initialization strategy integrating opposition-based learning is employed to improve the diversity of the initial population.Then,a nonlinear control parameter is utilized to balance of exploration and exploitation.In addition,an enhanced search strategy is designed to enhance the exploration of leaders,so as to avoid local optimization.Finally,comparative studies on the benchmark problem and SCOS problem are performed to verify the proposed approach.(5)In view of the robustness of service,a tri-objective SCOS model considering maximum Qo S,maximum robustness of service,and minimum energy consumption is established,and a new multi-objective grey wolf optimizer based on decomposition inspired from MOEA/D is proposed.Firstly,the population initialization strategy integrating opposition-based learning is used to improve the diversity of the initial population.Then,a crossover operator based on differential evolution is proposed to conduct the position-updating of population.Moreover,the adaptive crossover rate is designed to balance exploration and exploitation.Finally,comparative studies on the benchmark problem and SCOS problem is conducted to verify the proposed approach.