Predicting Material Wear and Degradation: Challenges in Standardizing Data Modalities for AI-Enabled Generative Design

Abstract Accurate material degradation prediction ensures the reliability and safety of systems applied across photovoltaic cells and nuclear reactors. Current predictive capabilities encounter long-term data scarcity and a lack of standardized protocols for data collection and reporting. This Perspective emphasizes the importance of interdisciplinary collaboration between experimentalists and data scientists, particularly in designing data capture practices for processing history, microstructure, surface metrology, and environmental conditions as time-resolved observations, further prompting discussions of data and metadata requirements to support degradation forecasting workflows. This Perspective highlights material wear due to it being strongly influenced by a combination of processing history, surface state, and operating environment. This exemplar material regime provides multimodal data complexity where artificial intelligence (AI) and machine learning (ML) offer promising opportunities for forecasting wear and degradation. Instigating collective action to self-organize and address these challenges promotes the development of more reliable, interpretable, and transferable models for wear and degradation prediction for the community.