Component-level multi-lifecycle end-of-life framework, enhancing sustainability and profitability

Effective implementation of Closed-loop strategies through End-of-Life (EoL) management is fundamental to the Circular Economy (CE). However, current EoL models predominantly address single or second use, failing to capture the dynamic evolution of individual component states and the potential of having multiple lifecycles for components. This paper presents a dynamic multi-lifecycle modelling framework to optimize EoL actions at the component level, accounting for their individual degradation and inter-cycle dependencies where decisions in one lifecycle directly constrain the feasible options and performance of all subsequent cycles. Employing multi-objective optimization, the framework allows balancing environmental impact (EI) and economic profit using Deep Reinforcement Learning Multi-Objective Optimization (DRLMO), as benchmarked against the evolutionary algorithms. A smartphone case study was applied to demonstrate the framework’s capability to identify optimal EoL strategies, revealing component recyclability and degradation, particularly of the mainboard and battery, as critical sustainability drivers. Results show DRLMO matches the solution quality of both evolutionary benchmarks while offering superior computational efficiency and significantly higher solution consistency across multiple runs. Furthermore, a sensitivity analysis highlights actionable priorities for CE stakeholders, identifying mainboard recyclability and battery degradation as the most critical drivers of both profitability and environmental impact. By validating sub-millisecond decision latency, this work establishes a computationally feasible foundation for integrating adaptive EoL decision-making into industrial digital twin ecosystems.