AILD: An Interpretable AI Framework for Designing Logical Descriptors in Predicting CO Adsorption on Catalysts

A rigorous understanding of structure-property relationships is pivotal for the rational design of high-performance catalysts, which requires descriptors that are both predictive and physically interpretable. To overcome the limitations of empirical metrics and the opacity of machine-learning "black boxes", we propose a 4-stage Artificial Intelligence-guided Logical Descriptors (AILD) framework: (1) Knowledge-driven feature generation, (2) Computational data and preprocessing, (3) Feature engineering and predictive modeling, and (4) Mechanistic interpretation and experimental guidance. By uniting domain knowledge, first-principles simulation, and interpretable machine learning, this framework links mechanism to design and advances catalyst development beyond empirical trial-and-error toward a knowledge-driven, science-based paradigm. To validate these logical descriptors, we have implemented them in the open-source AILD software, enabling reproducible research and accelerated rational catalyst design.