An adaptive recognition method for reliable collaboration of manufacturing services based on edge-aggregated graph convolutional network
提出一种基于改进图卷积神经网络的RCEN方法，利用制造过程高频历史数据构建动态服务协作图，提升制造服务协作链可靠性评估的准确性和全面性。
Current methods for assessing the reliability of Manufacturing Service Collaboration Chain (MSCC) primarily rely on service quality metrics and numerical calculations. However, these traditional methods face significant challenges in practical applications. This is due to idealised assumptions inherent in the definition of reliability, the often-low quality of historical data, and factors such as the coupling of service physical attributes, the flexibility of collaboration patterns, and the dynamic changes in MSCC topology. To overcome the aforementioned problems and enhance the accuracy and comprehensiveness of MSCC reliability assessment, this paper proposes an Reliable Collaboration Adaptive Recognition Network (RCEN). The RCEN introduces an improved graph convolutional neural network architecture, integrating a deep feature extraction mechanism with an optimised node information aggregation strategy. This method fully leverages the advantages of graph data in modelling complex service dependencies. It constructs a dynamic MSCC graph based on high-frequency historical operational data from the manufacturing process, where nodes and edges characterise the physical and operational attributes of manufacturing shop-floor services, respectively. Specifically, the graph construction method captures the inherent service collaboration patterns within manufacturing workflows. Combined with domain-specific quality metrics, it exhibits strong generality and scalability. Experimental results on multiple benchmark datasets demonstrate that the proposed method significantly outperforms three representative state-of-the-art methods in the task of MSCC reliability identification.