Search

Reverse osmosis (RO) and nanofiltration (NF) membranes are pressure driven, diffusion controlled process. The influence of surface characteristics on membrane process performance is considered significant and is not well understood. Current mass transport models generally assume constant mass transfer coefficients (MTCs) based on a homogeneous surface. This work evaluated mass transfer processes by incorporating surface morphology into a diffusion-based model assuming MTCs are dependent on the thickness variation of the membrane’s active layer. To mathematically create such a surface layer, Gaussian random vectors embedded in a software system (MATLAB) were used to generate a three-dimensional ridge and valley active layer morphologies. A “SMOOTH” script was incorporated to reduce the influence of outlying data and make the hypothetical surfaces visually comparable to the AFM images. A nonhomogeneous solution diffusion model (NHDM) was then developed to account for surface variations in the active layer. Concentration polarization (CP) is also affected by this nonhomogeneous surface property; therefore, the NHDM was modified by incorporating the CP factor. In addition, recent studies have shown that the membrane surface morphology influences colloidal fouling behavior of RO and NF membranes. With consideration of the spatial variation of the cake thickness along the membranes, a fouling model was established by assuming cake growth is proportional to the localized permeate flow. Flux decline was assumed to be controlled by the resistance of cake growth and accumulated particle back diffusion at the membrane surface. A series of simulations were performed using operating parameters and water qualities data collected from a full-scale brackish water reverse osmosis membrane water treatment plant. The membrane channel was divided into a thousand uniform slices and the water qualities were iii determined locally through a finite difference approach. Prediction of the total dissolved solid (TDS) permeate concentration using the model was found to be accurate within 5% to 15% as an average percentage of difference (APD) using the NHDM developed in this research work. A comparison of the NHDM and the modified NHDM for concentration polarization (CP) with the commonly accepted homogeneous solution diffusion model (HSDM) using pilot-scale brackish water RO operating data indicated that the NHDM is more accurate when the solute concentration in the feed stream is low, while the NHDMCP appears to be more predictive of permeate concentration when considering high solute feed concentration. Simulation results indicated that surface morphology affects the water qualities in the permeate stream. Higher salt passage was expected to occur at the valley areas when diffusion mass transfer would be greater than at the peaks where the thin-film membrane is thicker. A rough surface tends to increase the TDS accumulation on the valley areas, causing an enhanced osmotic pressure at the valleys of membrane. To evaluate the impact of surface morphology on RO and NF performance, fouling experiments were conducted using flat-sheet membrane and three different nanoparticles, which included SiO2, TiO2 and CeO2. In this study, the rate and extent of fouling was markedly influenced by membrane surface morphology. The atomic force microscopy (AFM) analysis revealed that the higher fouling rate of RO membranes compared to that of NF membranes is due to the inherent ridge-and-valley morphology of the RO membranes. This unique morphology increases the surface roughness, leading to particle accumulation in the valleys, causing a higher flux decline than in smoother membranes. Extended fouling experiments were conducted using one of the RO membranes to compare the effect of different particles on actual water. It was determined that membrane flux decline was not affected by particle type when the feed water iv was laboratory grade water. On the other hand, membrane flux decline was affected by particle type when diluted seawater served as the feed water. It was found that CeO2 addition resulted in the least observable flux decline and fouling rate, followed by SiO2 and TiO2. Fouling simulation was conducted by fitting the monitored flux data into a cake growth rate model. The model was discretized by a finite difference method to incorporate the surface thickness variation. The ratio of cake growth term (�1) and particle back diffusion term (�2) was compared in between different RO and NF membranes. Results indicate that �2 was less significant for surfaces that exhibited a higher roughness. It was concluded that the valley areas of thin-film membrane surfaces have the ability to capture particles, limiting particle back diffusion.