Modularization of smart product service (SPS) emerges as a priority strategy to agilely satisfy the dynamic customization requirement and to flexibly response to the rapid market change. Compared with the traditional product service, the SPS have more complicated interactions between the service components due to the novel characteristics caused by the application of smart technologies. The SPS modularization presents great differences from the identification of service component, correlation evaluation and module partition with the traditional product service modularization. However, most the existing research mainly focuses on the context of traditional product service, while containing scant study of smart product service. Therefore, this study proposes a hybrid framework for SPS modularization. In the framework, a cyber-physical product service blueprint is firstly proposed to represent the SPS operation process and identify the SPS components. Then, a rough-fuzzy correlation matrix is presented to determine the comprehensive interdependence between all pairs of SPS components with fully considering the hybrid decision uncertainties involved in the evaluation process, i.e., intrapersonal linguistic vagueness and interpersonal preference diversity. After that, the complex network theory is used to construct the SPS network and a modified Girvan-Newman algorithm is adopted for the SPS module partition. Finally, an illustrative modularization case of smart gearbox maintenance service and some caparisons with other methods demonstrate the feasibility and validity of the proposed approach.

智能产品服务（SPS）的模块化已成为一种优先策略，可以灵活地满足动态定制需求并灵活地响应快速的市场变化。与传统产品服务相比，由于智能技术的应用所带来的新颖性，SPS在服务组件之间的交互更为复杂。SPS模块化与传统产品服务模块化相比，在服务组件的标识，相关性评估和模块划分方面存在很大差异。然而，大多数现有研究主要集中在传统产品服务的背景下，而对智能产品服务的研究却很少。因此，本研究提出了用于SPS模块化的混合框架。在框架中，首先提出了一种网络物理产品服务蓝图，以表示SPS的运行过程并识别SPS的组成部分。然后，提出了一个粗糙-模糊的相关矩阵来确定所有SPS组件对之间的全面相互依赖性，同时充分考虑了评估过程中涉及的混合决策不确定性，即人际语言模糊性和人际关系偏好多样性。之后，采用复杂网络理论构建SPS网络，并对SPS模块分区采用改进的Girvan-Newman算法。最后，以智能变速箱维修服务为例的模块化案例以及与其他方法的一些对比证明了该方法的可行性和有效性。