Composing existing services into one value-added composite service for a cooperative business process is especially attended by both academia and industry in the cloud manufacturing (CMF) environment from the perspective of industrial information integration. Various service composition methods based on the QoS have been proposed to satisfy the user's requirement. However, the traditional methods fail to consider interrelation among various services and change in QoS. The unstable QoS may affect the reliability of service composition, which may also affect reliable collaboration as cloud services need to cooperate with each other to accomplish the business process efficiently. In this paper, we propose an efficient two-phase approach by integrating clustering and Chaos-Gauss-based PSO to solve the above problem. In phase one, the K-means clustering algorithm is adopted to improve the quality of candidate services with the consideration of QoS stability for reducing the search space. In phase two, a novel multi-objective PSO algorithm based on Chaos-Gauss named CGPSO is proposed to find the optimal service composition. The Two-Phase approach considers both the QoS stability and service collaboration ability to reduce the probability of service composition failure. Further, we conduct a comprehensive analytical and experimental study to show that our approach has better performance and effectiveness for service composition than other approaches in the CMF environment.

从工业信息集成的角度来看，在云制造（CMF）环境中，学术界和工业界尤其需要将现有服务组合为一项用于合作业务流程的增值复合服务。已经提出了各种基于QoS的服务组合方法来满足用户的需求。但是，传统方法无法考虑各种服务之间的相互关系以及QoS的变化。不稳定的QoS可能会影响服务组合的可靠性，也可能会影响可靠的协作，因为云服务需要相互协作才能有效地完成业务流程。在本文中，我们提出了一种有效的两阶段方法，该方法将聚类和基于混沌高斯的粒子群优化算法相集成来解决上述问题。在第一阶段，在考虑QoS稳定性的前提下，采用K均值聚类算法提高候选服务的质量，以减少搜索空间。在第二阶段，提出了一种基于混沌高斯的新型多目标PSO算法CGPSO，以找到最佳的服务组合。两阶段方法同时考虑了QoS稳定性和服务协作能力，以减少服务组合失败的可能性。此外，我们进行了全面的分析和实验研究，表明与CMF环境中的其他方法相比，我们的方法在服务组合方面具有更好的性能和有效性。