Title: Accelerating Graphite Anode Development Through AI-Driven Iterative Feedback Loops

Abstract:

This research introduces a novel, iterative workflow that leverages artificial intelligence to expedite the development of graphite-based anodes, simultaneously enhancing both process robustness and formulation viability. Utilizing the Citrine Platform, the study employed sequential learning and AI/ML-guided multiobjective inverse design to optimize anode performance. The process began with a dataset characterized by noise and incompleteness; however, early surrogate models generated by the platform effectively identified critical gaps in process constraints, despite their initial low predictive certainty. Through the iterative incorporation of feasibility labels and data regarding boundary condition failures, the system rapidly narrowed its focus toward high-performance, manufacturable formulations. This approach yielded significant operational improvements: fabrication reliability saw a dramatic shift from frequent failures to a 100% success rate in cell production. Furthermore, the proportion of cells achieving a capacity of at least 350 mAh g⁻¹ surged from 28.4% to 84.8%, and capacity retention improved substantially from 42.1% to 97.3%. These findings underscore the potential of structured, feedback-centric AI methodologies to convert imperfect industrial data into precise, actionable insights, thereby facilitating a faster and more reproducible optimization of battery electrode manufacturing.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC