Intensifying the Modeling of Membrane Processes Via a Multi-Scale Modeling Approach

Membrane processes have become a well-developed class of water treatment technologies. Accurate and accessible modeling software is critical to making informed decisions for encouraging the implementation of membrane technologies. Increasing the utilization of innovative water recovery and treatment approaches such as membrane separation technologies will be critical to addressing water scarcity. Modeling membrane processes often demands a trade-off between simulation speed and fidelity when the focus of these models shifts from micro-scale to system-scale investigations. This work demonstrates a multi-scale modeling approach developed to bridge the gap between fundamental physics-based models and system- scale models for real-world analysis that can be applied to multiple types of membrane systems.An improved approach to the module-scale modeling of osmotically driven membrane processes by accurately representing the features of a spiral-wound membrane was proposed. The model captures important membrane features such as the cross-flow stream orientation, membrane baffling, and channel dimensions unique to spiral-wound membranes. This model was implemented at the system-scale to compare system designs for hybrid forward osmosis-reverse osmosis (FO-RO) and seawater reverse osmosis-pressure retarded osmosis (SWRO-PRO) systems, most notably, a multi-stage recharge design. Results indicate that a multi-stage recharge design can allow FO systems to achieve a wastewater utilization of 90%, and PRO systems increase specific energy recovery up to 75%. Experimental data from an engineering-scale SWRO-PRO system was used to verify a solution-diffusion-with-defects model that indicates defects in the membrane structure. A solution-diffusion with defects (SDWD) model proposes that a membrane with defects, where the defects operate like a filtration membrane, would allow for a pressure-driven convective flow through the membrane. The SDWD model explains the lower-than-expected water permeability and salt rejection often seen in experimental PRO results. Integrating this model into future software will allow for more accurate simulation of larger systems and aid in investigating PRO scale-up. Turbulence-promoting spacers are an integral component of most membrane systems for providing structural integrity and increased mixing to maximize water recovery. However, including these spacers comes at the cost of increased pressure losses in the membrane channel. Computational fluid dynamics (CFD) simulations were used to investigate the impact of the geometric features of spacers on pressure loss and mass transfer in membrane channels. Neural network and regression models were trained on the data collected from CFD simulations to produce more representative and flexible models that better consider the geometric parameters of a given spacer design compared to existing empirical models. These models were then utilized with an optimization algorithm to uncover a spacer design that best balances the trade-off between reduced polarization and increased pressure losses in membrane channels