DWM - 2006 - September

Volume 7, Issue 8 - September 2006

IG Lessons: Desiccants
Computer Aided Prediction of Drying Out Performance of IG Units
by Dr. Hans Höfer, Dr. Horst Thamm and Andreas Keinath

A computer program for predicting the drying-out performance of insulating glass (IG) units has been developed by our company, GRACE GmbH & Co. KG. Based on a gas diffusion model, we learned that the experimentally determined water diffusion through the spacer bar perforation can be used to simulate the time dependent development of the spatial dew point distribution in an IG unit. The results of the simulation suggest that for IG units, non-symmetrically filled with desiccants, a dew point test executed at the center of a double lite might be misleading. This becomes critical when the time elapsed between manufacturing and shipment is short and the outside temperatures are low.

Introduction
The IG manufacturing industry uses a broad variety of spacer bars in terms of profile shape, bar width and kind of perforation. The perforations may differ in number and sequence of holes as well as hole diameters. Individual fabrication concepts require spacer bars that allow a fast water pick up for IG units, so they are ready for shipment just after completion. Alternatively, bars with a slow water pick up are also used. These have to be stored under temperature-controlled conditions for a decent amount of time prior to transportation. The need to have dew points below minus 30°C in the IG units prior to shipment is common to all manufacturing concepts. The spacer bar specifies the key characteristics of an IG unit. It determines the geometry of a window and, in combination with the desiccant, its drying out characteristics.

Figure 1 (below) depicts the experimentally determined water vapor uptake as a function of both various spacer bars and amounts of desiccant fillings leading to different dry-down rates.

Figure 2 (see page 93) shows the experimentally found dry-down rates of a fast and a slow profile measured with a 1-m by 1-m double lite, 16-mm spacer and a two-sided desiccant filling. The profiles differ in the number of holes/cm. The information about the dry-down rates and, in more detail, the time dependent development of the spatial dew point distribution after assembling the IG unit, is of pivotal importance to the window manufacturer in order to avoid numerous effects caused by water condensation that eventually lead to product quality issues.

Experimental Data
In order to qualify the computer program for dew point simulations, comprehensive experimental data was gathered to determine the drying-out behavior of an IG unit under industrially relevant conditions. Basically, two series of experiments were conducted: measurements of water diffusion rates through various profiles and time-dependent dew point development evaluations.

The water ingress into spacer bars was investigated by cutting the spacer bars into pieces approximately 40 cm long, filling these specimens with active desiccant, such as zeolite type 3A, and sealing the ends tightly with polyisobutylene, usually called “butyl.” In some experiments the spacer bars were connected by corner keys and/or profile junctions in order to study the influence of these elements on water ingress. Furthermore, complete frames were investigated. Then the spacer bars were stored in a climate chamber at exactly defined ambient temperature and humidity. The time dependence of the water ingress into the spacer bar or frame, respectively, was found by weighing the spacer bar or frame immediately after sealing and then again in pre-determined time periods after storage in the climate chamber. In most cases the atmosphere in the climate chamber was maintained at 25°C and 50 percent relative humidity (RH). The water ingress was observed for at least 24 hours. In all tests, a minimum of five spacer bars or frames of the same type was investigated (note: all curves shown in the profile test graphs represent the averages of at least five spacer bar or frame specimens).

The dew point measurement according to DIN 52345/EN 1292-2 comprises a controlled cooling down of a small outside area of the lite at the center of an IG unit. Heat flow from the inside to the outside surface of the lite takes place. The dew point inside the IG unit was reached when water vapor began condensing at the corresponding point inside of the lite (dew formation). To practice, a plastic beaker was cut in half and glued with butyl to the center of the lite and filled with acetone. A mirror for observing the dew formation was placed next to the outside of the lite. Solid carbon dioxide added to the acetone formed a cooling mixture. The progress of the downward cooling was monitored with a thermometer. The lower the dew point, the dryer the atmosphere inside the IG unit.

Background of the Computer Model
The program is based on a model that describes the diffusion of water molecules within the gas atmosphere inside the IG unit (Fick’s second law), the water transfer rate through the profile, as determined by the profile test, and the water vapor adsorption characteristics of the desiccant. In the course of the experiments it was found that the type of desiccant had little influence on the above-mentioned water transfer rate. Therefore, a separate term for differing desiccants was not built into the program.

Mathematical-Physical Basics of the Drying-Out Computer Program
The drying-out performance of an IG unit caused by physical adsorption of water vapor on the desiccant accommodated in the spacer bar may be split into three parts theoretically:

Diffusion of the water vapor molecules from the gas volume between the window lites to the perforation openings of the spacer bar;
Transfer of the water vapor molecules through the perforation openings of the spacer bar; and
Uptake (adsorption) of the water vapor molecules by the desiccant within the spacer bar1 (see page 96).

In the following charts, experimental and calculated data are compared in order to prove the validity of the gas diffusion model. Figure 3 (see page 93) shows the dew point development of a 1.0 m x 1.0 m x 0.016 m, two-side filled IG unit with a water diffusion rate of 1.43 g/m/24h referring to a “slow” profile. The dots represent the measured data; the line represents the calculated values. The measured and calculated dew points refer to the conditions in the center of the lite. Figure 4 shows identical conditions, but with a four-side filling. In both cases the calculations satisfyingly match the measured data.

Figure 5 relates to the same experimental set up but with just a one-side filling. Also, in this more critical case, experimental and calculated data are in agreement. It can be seen that the desiccant, even in a one-side filling, has enough water pick up power to meet the dew point specifications of minus 30°C, but after a much longer period of time (double that of a two-side filled unit). Figure 6 represents the same experiment as given in Figure 5, but with a doubled water diffusion rate of 2.86 g/m/24h, now referring to a fast profile. Again experimental and calculated data are in agreement. As expected, the doubled water diffusion rate leads to half of the dry-down time of the IG unit as given in Figure 5.

Figure 7 shows the dew point development of a 0.6 m x 0.6 m x 0.016 m four-side filled IG unit with a water diffusion rate of 1.43 g/m/24h again referring to a “slow” profile. This diagram also shows the measured and calculated data in agreement.

This program, however, has its limits. A number of test series and corresponding simulations revealed that the program becomes less accurate when the IG units are very small. In these instances the real drying out times are slower than predicted.

Using the Computer Program
In the following series of tests, the computer program has been used for the time-dependent development of the spatial dew point distribution. Since dew point measurements are time consuming, even for quality control reasons, only one set of time-controlled dew point development measurements per IG unit is normally executed according to the standards. In contrast, the computer program allows the space within the IG unit to be divided into an array of volume elements. It is possible to calculate the dew point development for such a volume element resulting in a spatial dew point distribution. Each color on the graphs represents a given temperature range. Three industrially relevant cases have been simulated and depicted in Series A, B and C in the Appendix. In all cases we calculated air filling with 55 percent relative humidity at both 23°C processing and storage temperatures. The following parameters were varied:

Number of filled sides;
Arrangement of filled sides;
Geometry of the IG unit; and
Water diffusion rates through the profiles.

We chose a slow profile with a rate of 1.43 g/m/day and a fast profile with a rate of 2.86 g/m/day. The input data are given in the box on page 130. For the sake of simplification, concentration of filling gas, temperatures and relative humidity were kept constant for all simulation runs.

Results and Conclusion
The simulations show that a dew point below – 30°C across the entire lite of glass can be reached in Series A already after 60 minutes. In contrast, in Series C, this is the case after ten hours starting from sealing the IG unit. The calculated time and spatial dependent dew point developments as given in Series C disclose that windows with asymmetrically filled spacer bars might have asymmetric dew point distributions. This might lead to misleading quality assessments when the dew point would be measured in the center of the lite. When in doubt, the dew points for such windows should be measured non-centrically. In case of adjacent two-sides filling the measurement should take place close to the opposite corner of the window.

Dr. Hans Höfer is the global technology manager, adsorbents, for GRACE GmbH & Co. of Worms, Germany. Dr. Horst Thamm is retired from GRACE, where he served as senior principal scientist. Andreas Keinath is regional technical customer service manager.

Acknowledgement
The authors wish to thank Professor Dr. Matthias Suchow, Department of Chemical Engineering of the Fachhochschule Lausitz, for his assistance in the programming.