Multi-Criteria Attributed Graph Embedding-Enabled Decision Support System for Manufacturing Resilience

Abstract

Manufacturing resilience is critical for achieving mass personalization paradigms, with disruption management being a key aspect to ensure continuous collaboration between manufacturing value chains. Past research primarily focuses on mitigating material shortfall either through supplier selection or material substitution with minimal historical considerations from both areas when devising recovery strategies. Additionally, the lack of an automatic decision support system often results in lengthy discussions and experience-led decision support. This study introduces a graph-embedded decision support system (GEDSS) designed to mitigate disruptions by recommending viable solutions. An event-centric industrial knowledge graph is constructed which integrates supplier networks, shop floor, and product configuration components to provide a holistic view of the manufacturing value chain. To facilitate solution generation, this study proposes a multi-criteria attributed graph embedding (MCAGE) approach to encode both business and operation considerations in a consistent embedding space. As such, feasible solutions can be inferred via link prediction. The experimental results reveal that the MCAGE approach outperforms baseline graph embedding models and the effectiveness of the GEDSS is further demonstrated through an automotive case study.