Macroscopic fluid flow conditions in spiral-wound membrane elements

Abstract

Reverse osmosis (RO) finds increasing applications as separation technique in chemical and environmental engineering where desalination, selective separations in the agro-industrial processes or wastewater purification are well-established examples. To fully evaluate the potential of RO and facilitate scale-up procedures, the modelling of the process is an important tool and literature models analyze the separation efficiency in terms of mass transfer with material balances, pressure drop through the module and mass transfer coefficients as dominant parameters. Important underlying factors are the geometry of the module and the hydrodynamic flow regime since mass transfer and pressure drop are a function of these factors, as witnessed by several publications. Since the concentrate-side (which is also the feed side) of the membrane plays the key-role, measurements should focus on the concentrate-channel of the spiral-wound membrane element. The determination of the channel flow regime and hydrodynamics of spiral-wound RO-channels can be carried out through the measurement of the residence time distribution (RTD). The present paper describes our experimental investigations on RTD through the application of a step change in tracer concentrations and relates the RTD-response curves to the regime of flow through the concentrate channels. This tracer-technique is common in characterizing flow regimes in chemical reactors. Its application to RO flow channels is innovative and results obtained stress its applicability to this specific geometrical layout. Results indeed demonstrate that the experimental average residence time of the concentrate channel is smaller than the theoretically calculated residence time, with differences between both values gradually decreasing with increasing liquid flow rate. This observation corresponds with findings for traditional packed bed applications where the presence of dead volumes reduces the real average residence time. The presence of dead zones in the spiral-wound membrane element is therefore evident. Since in our experimental procedure, residence time distributions are measured both between the tracer injection point and respectively the inlet and outlet of the spiral-wound element, the nett RTD-contribution of the membrane element itself can only be determined by eliminating the influence of the inlet flow equalization zone. This is performed by applying the convolution principle. Experimental and calculated RTD-curves for values of the Péclet-number of approximately 20 are in very good agreement. In analogy with fixed bed applications, the interpretation of the findings corresponds to a laminar flow profile with a limited dispersion. The definition of this flow profile in the concentrate channel is important in the use of transport models to characterize the membrane performance, as will be shown in a further paper.