Physics-Informed Machine Learning for Smarter Design and Manufacturing Part I

Abstract Manufacturing systems are undergoing a fundamental transformation driven by the convergence of cyber-physical infrastructure, ubiquitous sensing, and artificial intelligence. At the heart of this transformation lies the challenge of embedding physical knowledge into data-driven models to achieve both predictive accuracy and engineering interpretability. Physics-informed machine learning (PIML) has emerged as a paradigm for integrating conservation laws, constitutive equations, and domain constraints into neural architectures. Despite significant advances in smart design and manufacturing, three persistent gaps remain. First, multimodal sensor fusion in manufacturing environments continues to struggle with heterogeneous data types and varying sampling rates, where data-driven approaches often fail to generalize across operating conditions. Second, intelligent fault diagnosis under data scarcity, label noise, and cross-domain shifts demands methods that can encode physical priors while maintaining sample efficiency. Third, dynamic production planning under disturbances requires optimization frameworks that respect physical constraints while adapting in real time. This special issue brings together 10 contributions that address these challenges through novel PIML models. These works demonstrate how the systematic integration of physical knowledge can enhance model robustness, reduce data requirements, and improve generalization across manufacturing applications.