Multi-centre cutting service composition in cloud manufacturing: a hierarchical decision model with guillotine cutting evaluation

Abstract

In material-intensive Cloud Manufacturing (CMfg), traditional Service Composition and Optimal Selection (SCOS) approaches often treat cutting services as abstract functional units, which may lead to biased cost estimation and infeasible assignments. This study proposes a hierarchical decision-aid model named DiCACE to integrate macroscopic multi-centre assignment and logistics decisions with microscopic guillotine cutting evaluation, simultaneously optimising completion time, total cost, and overall material utilisation. To make this tightly coupled and strongly constrained problem computationally tractable, we develop a decoder-driven evolutionary framework based on an improved NSGA-III. A novel three-stage decoding strategy is introduced to act as a bridge, efficiently translating abstract assignment decisions into physically executable cutting plans while satisfying complex operational constraints. Experimental results from an industrial case study and dynamic simulations support the practical applicability of the framework. The results show that the proposed framework produces more competitive Pareto trade-offs than representative multi-objective algorithms and maintains better system-level coordination in dynamic scenarios compared to a representative deep reinforcement learning method, while generating feasible assignment schemes for large-scale orders within several minutes. The findings further suggest that CMfg platforms should incorporate process-aware pre-evaluation into the allocation stage, upgrading from passive matchmaking to feasibility-assured decision support for cross-enterprise collaboration.