Study On Complex Multi-task Oriented Services Composition In Cloud Manufacturing Base On Crossover And Mutation Particle Swarm Optimization

Cloud manufacturing is a new mode of network manufacturing aims to achievemanufacturing resource sharing and on-demand using based on knowledge, therebyimproving resource utilization and core competitiveness of enterprises. Servicecomposition and optimization is one of the core technologies of the manufacturingresource optimal allocation. Since cloud manufacturing is an open, multi-user andcomplex system, service composition in cloud manufacturing must be multi-taskoriented. However, in cloud manufacturing, most existing studies on servicecomposition and optimization are specified to one-task goal, which is unable to adapt tothe complex needs of the actual situation. Therefore, study on complex multi-taskoriented service composition and optimization technology in cloud manufacturing,which presented in this paper is of great theoretical and practical significant.Compared with previous researches, the paper breaks through the single constraintin the respects of manufacturing task quantity，type and time series. It explores asolution for complex and multi-task service composition and optimization in cloudmanufacturing. The main contents are as follows:①A four-dimensional QoS(Quality of Service) evaluation system and a QoScalculation expression of composite cloud service are proposed as the basis for cloudservices composition and optimization. Firstly, the four-dimensional QoS evaluationsystem，which contains the detailed descriptions and formulas，is established byanalyzing the implementation process of CSCO(Cloud Services Composition andOptimization) and the characteristics of cloud manufacturing. Then, the author studiesfour possible structural models of cloud services composition and their correspondingQoS expression. And the QoS formula for composite cloud service was eventuallygiven.②A crossover and mutation particle swarm algorithm for single-task-orientedCSCO was designed for supporting for the further study. Firstly, a problem model ofsingle-task oriented CSCO was established. Then, in order to solve the problem model,a crossover and mutation particle swarm algorithm was designed. The algorithmincorporates the ideas of crossover and mutation in genetic algorithm into particleswarm optimization. And the local optimization, which is the last step of algorithm,uses greedy choice strategy. ③An overall solution of complex multi-task oriented CSCO was proposed. Thispart studies two kinds of CSCO launched by multi-task asynchronous request andmulti-task synchronous request. The main work are:1) As to solving problem model,breakthrough the restrictions of task request number, type and timing relationships,given the complex multi-task-oriented problem solving algorithm;2) As to theoptimization strategy, a hybrid optimization strategy which considering optimizationresults and time performance was design;3) As to cloud services occupancy rights, afive-priority strategy was designed.Finally, an experimental model based on the motorcycle manufacturing applicationscenarios is established for simulation test of the proposed complex multi-task orientedCSCO solution. The results show that:1) the solution has a high hit ratio of optimalresults, which proves the effectiveness;2) With the expansion of the scale of the taskrequests, the execution time of the solution is within a controllable range, which provesthe efficiency;3) by compared to traditional particle swarm algorithm, the crossoverand mutation particle swarm optimization in the solution has obvious advantages.