Modeling of fluid flow in spiral wound reverse osmosis membranes

Abstract

The research performed here is motivated by the need to understand and quantify the phenomena that underlie the purification of impure water by the reverse osmosis process. Despite the fact that reverse osmosis is a well-established method of water purification, the design and implementation of the process has been based on vastly oversimplified models. Oversimplified reverse-osmosis (RO) models lead to inefficient RO element performance estimation.Reverse osmosis is based on profoundly interacting fluid flow and mass transfer phenomena. These phenomena are modeled without approximation, and the model is implemented by numerical simulation. The numerous oversimplifications of prior models have been eliminated. In particular, the linearity that marked those models has been demonstrated to be invalid. Although the flow is laminar, it is not friction dominated. Instead, pressure drop non-linearity exists because of inertial effects. A second factor promoting non-linearity is the continuous bleeding-off of product water from the salt-containing feed stream. The flow phenomena at the entry and exit of the individual reverse osmosis elements have been clarified. The associated pressure drops were found to be remarkably small.The simulations spanned the entire range of operating conditions of actual reverse osmosis installations. The species conservation for salt took account of both advection and diffusion. Mass transfer coefficients at the membrane surface were determined, again for all practical operating conditions. The main outcomes of the simulations were the true portrayal of the in-element pressure losses and mass transfer coefficients. Experiments were performed to support the simulation model. The attainment of marginal levels of agreement motivated careful examination of the physical interactions of the feed spacer and the RO membrane. Upon investigation, it was found that the feed spacer penetrated into the membrane, with the outcome that the dimensions of the actual flow passage were less than that based on the dimensions of the feed spacer alone. When the simulation was repeated with the actual flow passage dimensions, good agreement was achieved between the simulations and the experimental data.Based on the new information extracted from both the simulations and the experiments, a new methodology was developed for the accurate simulation of typical reverse osmosis elements. The new methodology supersedes that which has been standard in the past. All of the oversimplifications and omissions have been avoided in favor of a logic-based application of the underlying physical phenomena. The outcome of the research work is a thorough understanding of the reverse osmosis desalination operation.