Computational fluid dynamic analysis of RO membrane performance with novel feed spacer geometries

A finite element based numerical model has been employed to describe momentum and mass transfer in open and spacer-filled crossflow membrane channels. Simulations considering operating conditions typical of seawater (SW), brackish water (BW), and reuse water (RW) reverse osmosis (RO) applications are employed for optimization of hypothetical spiral wound element feed spacer designs. Preliminary results suggest that for lower TDS waters, energy consumption is minimized at smaller spacer-to-channel height ratios. Conversely, larger spacer-to-channel thickness ratios result in minimal energy consumption when processing higher TDS waters. Spacer shape has little impact on concentration polarization (CP), but spacer shape has a large effect on axial pressure losses. A variety of spacer shapes and designs are simulated and evaluated for each water source. INTRODUCTION Concentration polarization is an important factor that limits separation performance in nearly all crossflow membrane filtration processes (Mulder, 1991). For example, polarization of rejected solutes in reverse osmosis (RO), nanofiltration (NF), and ultrafiltration (UF) separations causes elevated concentrations at the membranesolution interface, which increases solute passage and trans-membrane osmotic pressure. In addition, CP exacerbates all forms of surface fouling phenomena including scale formation by sparingly soluble mineral salts, cake formation by colloids, gel formation by organics, and biofilm formation by bacteria. Therefore, accurate description of CP phenomena is critical for design, optimization, and operation of RO membrane separations. MATERIALS AND METHODS A previously developed finite element model has been applied to study momentum and mass transfer in crossflow membrane filtration systems with open and spacer-filled channels (Subramani et al., 2006). A commercially-available finite element solver (FEMLAB 3.0, © COMSOL AB), was used to construct the model geometries and perform numerical simulations. The model employs the steady-state Navier-Stokes, continuity, and convection-diffusion equations as shown in eq (1-3), respectively. ( ) v g v v ⋅ ∇ ∇ + + ∇ + −∇ = μ ρ μ ρ 3 1 2 p Dt D (1) 0 = ⋅ ∇ v (2)