Adaptive multi-objective service composition reconfiguration approach considering dynamic practical constraints in cloud manufacturing

Dynamic uncertainty factors such as equipment faults are common in practically implemented cloud manufacturing (CMfg) environments, often causing the manufacturing service to be invalidated. In that case, efficient reconfiguration of the original service composition under practical constraints is critical; however, existing research scarcely focuses on it. This paper proposes a dynamic service composition reconfiguration model to bridge the gap by considering practical constraints (DSCRPC) in a real-life cloud manufacturing environment. Based on the constraints considered in this study, the DSCRPC model redefines three objectives: time (T*), cost (C*), and product service quality (Q*S*). To optimize the DSCRPC model, this study developed an adaptive multi-population multi-objective whale optimization algorithm (AMPOWOA) based on the Pareto strategy. The algorithm adopts four balancing strategies and adaptively optimizes and adjusts the key parameters under various balancing strategies through well-designed reinforcement learning models. Finally, we conduct numerical experiments and actual application case tests to compare the performances of AMPOWOA and other algorithms (MOWOA, MOHHO, NSGA-II). The results show that DSCRPC can continuously tackle the cloud manufacturing service composition (CMSC) reconfiguration issue with constraints until an order is completed. Moreover, AMPOWOA is superior to the other algorithms optimizing the DSCRPC model. This significantly enhances the robustness of service composition reconfiguration in real-life CMfg. (C) 2021 Elsevier B.V. All rights reserved.

Automated summary

This paper proposes a dynamic service composition reconfiguration model to address the service reconfiguration issue in real-life cloud manufacturing environments, redefining three objectives and utilizing an adaptive multi-population multi-objective whale optimization algorithm for optimization. The results show that the proposed model can effectively tackle the issue and outperforms other algorithms, enhancing the robustness of service composition reconfiguration in cloud manufacturing.