Surrogate-Based Decision-Making in Sustainable Lithium-ion Battery Production

This work addresses sustainable decision-making in multistage Lithium-Ion Battery (LIB) production, focusing on balancing production efficiency with environmental and material considerations. The rapidly growing demand for high-performance LIBs necessitates the simultaneous optimization of yield, energy consumption, and material usage across highly interconnected manufacturing processes, where small variations or defects can propagate nonlinearly and affect downstream quality. Digital twins have emerged as a powerful tool for system-level monitoring, process understanding, and optimization in battery production. By providing high-fidelity virtual representations of cell-level electrochemistry and process-level dynamics, they allow detailed scenario analysis and predictive evaluation of process interventions. However, such high-fidelity simulations are computationally intensive, making them impractical for iterative optimization, adaptive control, or real-time decision-support in large-scale production environments. To overcome these limitations, we propose a surrogate-based framework that leverages Machine Learning (ML) models to emulate the behavior of digital twins with substantially reduced computational cost. The surrogate models are integrated within a multi-objective optimization scheme, enabling rapid exploration of trade-offs between key performance indicators, including production cost, CO₂ emissions, and cell yield. This approach allows for dynamic, data-driven decision-making that can adapt to process variations, minimize resource consumption, and support sustainable production practices. We demonstrate the framework’s potential in capturing the complex interactions across multistage LIB production and its ability to guide process parameter adjustments for optimal outcomes. By combining surrogate modeling with digital twin insights, the methodology provides a scalable solution for efficient, environmentally conscious, and quality-driven battery production, bridging the gap between high-fidelity simulation capabilities and the practical requirements of real-time industrial applications.