Abstract: To solve the difficult problem of weighing the weights of multiple targets in the cloud manufacturing service composition,as well as significantly improve the population diversity of the evolutionary algorithm in the solution process and effectively balances the global and local search capabilities of the evolutionary algorithm,an evolutionary algorithm based on adaptive selection and reverse learning strategy was proposed,while optimizing the time,cost,reliability,availability and credibility.To shorten the time to solve the combined solution,the K-means method was used to cluster the candidate services based on the quality of service,and the poorer services were eliminated.Then,the reverse learning strategy was used to improve the global search performance,and the global and local search capabilities of the algorithm were effectively balanced through selection and probability update strategies.The results of comparative experiments with four advanced algorithms showed that the proposed algorithm had better comprehensive performance.

Key words: service composition, service quality, multi-objective optimization, evolutionary algorithm

摘要： 云制造服务组合问题往往存在多个优化目标且各目标权重难以事先确定,为了显著提高进化算法在求解过程中的种群多样性,并有效平衡进化算法的全局与局部搜索能力,提出一种基于自适应选择和反向学习策略的进化算法,同时优化时间、成本、可靠性、可用性和信誉度。首先为了缩短求解组合方案的时间,采用K-means方法基于服务质量对候选服务进行聚类,去除质量较差的服务;然后采用反向学习策略提高全局搜索性能,再通过选择策略和概率更新策略有效平衡算法的全局与局部搜索能力;最后与4种先进算法进行实验对比表明,所提算法具有更好的综合性能。

关键词: 服务组合, 服务质量, 多目标优化, 进化算法