Cross-Scale Modelling and Validation for Thermowettability in Graphene Channels: Measurement and Comparator Design

Abstract

Thermowettability must be understood across molecular interactions, nanoscale channel transport, device measurements, and operational outcomes, yet evidence at these levels is not directly interchangeable. This review develops a cross-scale validation design for AI models that connect them. Physics-informed surrogates encode temperature-dependent wetting and transport constraints, while hierarchical probabilistic models propagate uncertainty from simulation parameters and sensor error to flux or separation decisions. Comparators are selected to isolate what each AI component adds: mechanistic baseline, empirical regression, learned representation, and hybrid model are tested on the same regimes and shifts. The synthesis requires agreement not only in average trends but also in local sensitivities, failure cases, calibration, and transfer to changed channel geometry or surface condition. Provenance is retained for simulations, specimens, measurement protocols, preprocessing, and checkpoints. No new experimental dataset is introduced. The article contributes a measurement-and-comparator framework that prevents apparent cross-scale consistency from being inferred merely because different models produce similar curves, and makes AI-supported translation conditional on explicit validation evidence.