Data-Efficient Value Chain Digital Twin (VC-DT) Architecture
In modern manufacturing ecosystems, achieving end-to-end visibility remains a critical challenge due to supplier performance variability, logistics disruptions, and fragmented enterprise data silos. Instead of rigidly centralizing disparate datasets, this methodology—developed under the HORIZON DMaaST project—introduces a data-efficient Six-Layer Value Chain Digital Twin (VC-DT) framework that preserves data sovereignty while enabling dynamic simulation and structural flexibility.
Methodology & System Architecture
The proposed framework is built upon three core technical pillars designed to handle high-dimensional enterprise data while maintaining the dynamic fidelity required for systemic resilience assessment.
To provide end-to-end (E2E) visibility, the system is divided into 6 functional layers, each representing a The Six-Functional Layer Structure stage of the operational flow. At the heart of this structure is a dynamic Bill of Materials model that manages the material flow between layers:
- L1-Supplier: Models raw material procurement, supplier reliability, and lead-time variability.
- L2-Inbound Logistics: Captures transportation and shipping dynamics from suppliers to the facility.
- L3-Warehouse: Manages inventory states, safety stock policies, and material availability.[1] [2]
- L4-Production: Simulates floor activities, multi-level Bill of Materials (BOM) structures, and operational bottleneck dynamics using a Discrete Event Simulation (DES) engine.
- L5-Outbound Logistics: Models downstream distribution networks and finished goods shipping.
- L6-Customer: Represents final market demand points, service level requirements, and feedback loops.
Semantic “Soft-Linking” Approach
Rather than merging fragmented ERP, MES, and WMS data into a single centralized database, which compromises data ownership, the architecture employs a semantic “soft-linking” protocol.
- Data Sovereignty: Federated datasets remain in their native environments without undergoing rigid database joins.
- Bridge Keys: Disparate enterprise tables are connected via shared semantic identifiers, such as Configuration IDs and Shipment IDs.
- Future-Proof Foundation: This semantic association establishes end-to-end traceability while seamlessly preparing the underlying data structure for integration into a Decentralized Knowledge Graph (DKG) to ensure ultimate semantic interoperability.
Dynamic Simulation & Probabilistic Modeling Engine
To capture the real-world stochastic behavior of manufacturing flows, the VC-DT integrates a data-driven execution engine:
- Discrete Event Simulation (DES): Captures time-varying operational behaviors, including resource contention, lead-time variability, and inventory dynamics.
- Kernel Density Estimation (KDE): Supply and demand uncertainties are captured via KDE—a non-parametric, data-driven approach that extracts realistic, probabilistic distributions directly from historical data rather than assuming rigid theoretical models.
- Two-Stage Push-Pull Strategy: The product structure utilizes a hybrid production strategy. Stage 1 (Push) involves regular procurement and pre-production of lower-level sub-components independent of immediate demand. Stage 2 (Pull) triggers final assembly only upon receiving active customer orders. This dual-tracking allows the architecture to distinguish between upstream component shortages and downstream configuration bottlenecks.
Stress-Testing & “What-If” Scenario Analysis: This methodology provides a global interface allowing users to adjust operational parameters, such as supplier lead times, defect rates, safety stock multipliers, and arrival delays. By modulating these variables based on historical bootstrap data or expert input, users can execute end-to-end “what-if” analyses to safely evaluate risk policies and stress-test the value chain against severe shock scenarios before real-world deployment.
