Study On Several Evolutionary Algorithms For Smart Cloud Manufacturing Resource Service Composition

The increasing maturity of emerging information technologies such as Cloud computing,Internet of things,Cyber physical system and Big data have promoted the development of Smart cloud manufacturing.Manufacturing Cloud Service Composition(MCSC)dynamically assembles services offered by different providers through certain rules,which aims at implementing complex functions that are difficult to be completed by a single service,leading to an on-demand service capability for cloud manufacturing systems.The MCSC optimization problem has dynamic,coupling,large-scale and multi-objective features,and is a typical NP-hard problem.MCSC optimization algorithms have a direct impact on the efficiency and effectiveness of cloud resource allocation.This thesis extracts four types of MCSC problems arising from multi-task requirements,service domain knowledge reuse,the dynamic optimization,and the multi-objective optimization,and further constructs corresponding problem models,optimization strategies and implementation algorithms.The main contents of this thesis are as follows:(1)First,the development trend of manufacturing industry is retrospected,the characteristics of cloud manufacturing are briefly outlined,and the challenges for MCSC in the Cloud manufacturing environment are analyzed.Next,the research progress of Cloud manufacturing related fields is reviewed with focus on the MCSC optimization problems in complex cloud environment,and the research contents and overall framework of this thesis are drawn.(2)From the perspective of social cyber-physical-system,the abstract system model and architecture of Smart cloud manufacturing are discussed.From the point of data acquisition,analysis and intelligent application,the operation mechanism of Smart cloud manufacturing system is discussed.At last,the basic theoretical knowledge regarding Smart cloud manufacturing MCSC optimization is given.(3)In view of MCSC optimization for concurrent tasks in cloud environment,the overall optimization strategy is proposed by constructing a new model which can achieve a balance allocation of limited service resources among multiple tasks,and a hybird Artificial Bee Colony(ABC)algorithm with social cognitive ability and chaotic excitation mechanism is proposed.The proposed algorithm inherits the good exploration ability of ABC,and the introduction of distribution estimation model makes onlooker bees have social cognitive ability,while scout bees driven by chaotic sequence can better traverse the solution space.Numerical experiments show that,compared with the traditional single task optimization strategy,the proposed overall optimization strategy for multiple tasks can better allocate the cloud service capability,and solutions obtained by the proposed algorithm have better quality and higher stability.(4)Taking into consideration of service domain knowledge,a service domain knowledge guided pollination algorithm is proposed to solve MCSC problems with higher efficiency.First,the domain knowledge from the cloud service ecosystem is introduced into the design of flower pollination algorithm with the purpose of improving the initial population and search efficiency of heuristic operators.Then,a search strategy enhanced by individual dependency mechanism is proposed so that the search mode can be transfered automatically according to the fitness value of individuals.Excellent individuals are updated using the “dimension by dimension improvement” based self pollination strategy with focus on exploitation.Inferior individuals are updated by the Lévy random walk with concentration on exploration.They are cordinated to balance the exploration and exploitaion ability of the proposed algorithm.Numerical experiments show that domain knowledge can accelerate the convergence speed of the algorithm,and the search mechanism based on individual dependency mechanism can improve the search ability of the algorithm.(5)A multi-swarm adaptive differential ABC algorithm for dynamic MCSC optimization is proposed to address MCSC optimization in the dynamic cloud environment such as service addition,service deletion and quality of service changes.First,a multi-dimensional queue model for Skyline services is introduced to reduce the size of the candidate service set and shield the unrelated disturbance.Then,a dynamic multi-swarm differential ABC algorithm which can effectively balance exploration and exploitation is proposed,in which a state trigger mechanism is desinged to adjust the utilization rate of evolution strategies accroding to the changes of environment,a multi-subgroup exclusion strategy is adopted to avoid the over-crowded aggregation of individuals in the same region,and an aging mechanism is applied to activate individuals that may fall into local optimum.Experimental results show that the proposed algorithm has good search robustness and adaptability in solving dynamic MCSC problems.(6)To handle multi-objective MCSC problems,a so-called EsMOEA/D is proposed within the framework of basic MOEA/D,with introduction of a dual co-evolution strategy that combines multiple evolution operators and multiple neighborhood sizes,and the fusion of differential evolution,ABC and Teaching-Learning based optimization.The historical performance of different evolutionary operators and neighborhood sizes is recorded and used to adjust the utilization rate of different evolution strategies,which is expected to improve the search ability of EsMOEA/D on complex problems and achieve better allocation of computational resources.Comparison experiments show that the proposed algorithm has obvious advantages in solving multi-objective MCSC optimization problems.