A modular Digital Twin architecture for the data-driven reactive planning of disassembly systems

Abstract

Disassembly is key to circular production, enabling end-of-life products to be recovered through reuse, remanufacturing, or recycling. However, disassembly is characterized by high uncertainty about product condition and process feasibility, which requires dynamic, product-specific planning. While predictive disassembly planning anticipates uncertainties, disassembly systems must additionally be able to adapt their plans in real time when deviations occur. This is addressed by Reactive Disassembly Planning (RDP), which enables the dynamic adaptation of disassembly processes and schedules during disassembly. However, unlocking RDP’s full potential requires complex decision-making based on comprehensive real-time data. To support such data-driven RDP, a modular Digital Twin architecture is presented that integrates all decision-making steps into an automated, real-time planning pipeline. The architecture is instantiated and deployed as a Digital Twin in a learning-factory disassembly system to demonstrate the feasibility and end-to-end operation of the proposed modules and connectors. Its potential is illustrated in a representative scenario, in which the instantiated Digital Twin achieves a 78% reduction in lead time and a 2% increase in profit compared to operating without RDP, and a 529% profit increase compared to a rule-based process-termination strategy. These results serve as illustrative proof of applicability, indicating that data-driven RDP via Digital Twins can support efficient, resilient disassembly operations and thereby strengthen the business case for circular factories and the transition to more sustainable production.