The design of experiments method was applied to create empirical models to evaluate performance of membrane humidifiers. Experiments were run and the results analyzed to develop a combined metric for humidifier performance, provide information regarding effects, and compare geometries and membranes.

A range of experimental designs were evaluated and a central composite design selected to produce a response surface model. The numeric factors chosen were wet and dry stream flow rates and temperatures and the categoric factor was pressure. Four humidifiers were tested with two distinct geometries and two membranes.

The experimental analysis provided factorial models for response surfaces in six performance metrics:​ water transfer rate;​ water recovery ratio;​ humidity ratio;​ relative humidity;​ dew point;​ and dew point approach temperature. All of these models displayed overall model significance and a high signal-to-noise ratio, but also an overall lack of fit to the second-order model. This demonstrated the suitability of the design of experiments approach but suggests the use of a higher-order model.

Based on these models and a priori information regarding performance metrics, a combined metric was proposed incorporating relative humidity, water recovery ratio, and water transport rate. Performance bounds are set using relative humidity, and water recovery ratios inside these bounds are plotted.

Dry side flow rate was found to have a positive effect on total water transport but a negative one on saturation-based metrics. Wet side flow rate had a positive effect on total water transport and a negative effect on relative transport metrics. Dry side temperature had an overall negative effect. Wet side temperature had a similar effect to wet side flow rate. Low pressure had generally higher performance, but high pressure mitigated negative effects in dry side temperature and flow rate. The two geometries showed similar relative performance. The ionic membrane displayed consistently better water transport than the porous polymer membrane.

Based on these results simplified central composite designs for humidifier characterization are proposed, eliminating dry side temperature and pressure as a factor and allowing the possibility of varying flow rates in tandem. These tests have respectively 17 and 11 test runs for full humidifier characterization.