Hybrid models, digital twins, and digital shadows for sustainable membrane technologies: A critical review

Membrane separation technologies are essential for water purification, wastewater treatment, gas separation, hydrogen (H 2 ) production, and carbon (CO & CO 2 ) capture. Despite their wide applicability, large-scale adoption is limited by significant technical challenges, including fouling, the permeability–selectivity trade-off, high energy consumption, and the durability required for long term operation. Integrating machine learning (ML) and artificial intelligence (AI) with the physics-based models, as a hybrid mode, digital shadows (DSs), and digital twins (DTs), offers a promising pathway to overcome these limitations. Hybrid modeling combines physics-based simulations with data-driven learning to produce fast and accurate predictive tools for membrane processes. DSs extend this capability by creating passively updated virtual representations of membrane systems from real-time data, thereby enabling soft sensing, performance forecasting, and anomaly detection. DTs further advance these functions through bi-directional connectivity and higher-fidelity physical models, supporting adaptive control, predictive maintenance, and system-level optimization. This review examines recent applications of hybrid models, DSs, and DTs across major membrane technologies, including desalination and water treatment, membrane bioreactors, gas separation for CO 2 and H 2 , O 2 /N 2 separation, catalytic membrane reactors (CMRs), and proton-exchange membrane fuel cells. The role of computational fluid dynamics (CFD) in generating synthetic data, enhancing mechanistic understanding, and supporting model development is also discussed. Case studies demonstrate improvements in fouling prediction, energy-efficient operating set points, reduced operational costs, and real-time decision support at both module and plant scales. Collectively, hybrid modeling enables high-fidelity predictions, DSs support continuous monitoring, and DTs provide a pathway toward intelligent, self-optimizing, and sustainable membrane-based separation systems, bridging the gap between laboratory innovation and industrial deployment. • Hybrid ML models enhance membrane performance by combining physics-based simulations with data-driven learning. • Digital twins/ Digital shadows enable real-time monitoring, soft sensing, and performance forecasting in membrane systems. • Digital twins support adaptive control, predictive maintenance, and system-level optimization. • CFD-generated synthetic data strengthens mechanistic understanding and model development for membrane processes. • Case studies show improved fouling prediction, energy efficiency, and reduced operational costs.