Toward Intelligent Thermal Fluids: AI-Driven Design of Zeta-Engineered Nanofluids and Self-Organized Thermal Transport Networks

Abstract The development of next-generation nanofluids requires a shift from empirical formulation approaches toward data-driven design strategies that simultaneously optimize thermal conductivity, dynamic viscosity, stability, and system-level performance. This perspective highlights the emerging convergence of artificial intelligence (AI), zeta-potential (ZP) engineering, and advanced nanomaterials for the design of intelligent thermal fluids. Particular emphasis is placed on ZP-engineered nanofluids, in which electrostatic interactions are deliberately tailored to control nanoparticle dispersion, stability, and the formation of self-organized thermal transport networks. While interconnected nanoparticle networks can enhance thermal transport, excessive network connectivity, especially in high-aspect-ratio nanomaterials such as carbon nanotubes and graphene derivatives, may increase the dynamic viscosity and hydraulic resistance. Therefore, optimal nanofluid design requires balancing competing thermophysical and operational objectives under application-specific conditions. An AI-driven framework integrating machine learning, physics-informed machine learning, explainable AI, active learning, and multi-objective optimization is presented to accelerate nanofluid discovery and optimization. A representative case study of ZP-engineered iron oxide-carbon nanotube-graphene oxide nanofluids demonstrates the use of data-driven modeling, feature-importance analysis, SHAP interpretation, and Pareto optimization. The integration of AI-guided optimization with ZP-controlled network engineering provides a promising pathway toward intelligent thermal fluids and autonomous materials discovery for advanced thermal management systems.