Air dehydration membranes offer a simple, cost effective solution to humidity control. Membrane units may be installed in-line and require no- auxiliary utilities - only a portion of the feed gas is lost and a small pressure drop is incurred.

To produce desired dew points, a portion of the product gas typically is used as sweep in the module. Sweep lowers the water concentration in the permeate to permit sufficient reduction of the water concentration in the retentate (non- permeate) stream.

The literature reports numerous ways to create the sweep stream including: 1) making the fibers non-selective at the product end, 2) introducing the sweep from an external collar around the module, and 3) inserting tubes through the tubesheet that allow communication between the product header and the shell of the module.

We report simulations of the sweep distribution within the shell and its effect on module performance. Two types of simulations are considered: 1) simulations that explicitly predict flow fields within the shell based on how the sweep gas is introduced and 2) simulations that assume the sweep flow around each fiber is distributed in a Gaussian manner.

The use of fibers that are non-selective at the product end is most efficient based on module capacity and dry gas recovery. Introducing the sweep through an external collar or internal tubes is poorer due to poorer gas distribution in the shell.

Predictions based on explicit calculation of shell flow fields are in good agreement with those based on a Gaussian sweep distribution using a standard deviation in sweep flow equal to ~15% of the average sweep flow rate. We believe the results of this work may be used to evaluate alternative methods providing sweep in a module.