TEMPORAL GRAPH LEARNING FOR CASCADING FAILURE PREDICTION, BUSINESS IMPACT ASSESSMENT, AND RECOVERY OPTIMIZATION IN SMART MANUFACTURING SYSTEMS

Smart manufacturing systems contain interconnected machines, production processes, and operational resources. A failure in one component may affect dependent components, increase downtime, and create additional business losses. This paper presents a conceptual framework for temporal graph learning, cascading failure risk assessment, business impact estimation, and recovery optimization in smart manufacturing systems. The framework consists of five stages: synthetic data specification and preprocessing, temporal graph construction, cascading failure prediction, business impact assessment, and recovery optimization with explainability. A six component illustrative numerical example is used to demonstrate the business impact and resource constrained recovery decision. Under a resource capacity of five units, the example selects components M2 and M4 for recovery. The calculated total modeled cost decreases from $5,552.50 to $2,442.50, representing a 56.01% reduction under the specified assumptions. These values are analytical example results rather than outputs from a trained model or an experimentally validated simulation. The proposed framework provides a basis for future evaluation using real industrial data, repeated simulation trials, and comparative experiments.