Computer simulation of heat loss characteristics of commercial door assemblies.

INTRODUCTION

The 2005 ASHRAE Handbook--Fundamentals (ASHRAE 2005) includes a reasonable range of typical U-factors for standard residential entrance doors but only two values for revolving doors (for "open" and "closed" positions) and only three values for different types of overhead doors. The accompanying text notes that the values given "are generic values, and product-specific values determined in accordance with standards should be used whenever available." This is better than no information at all, but it provides little guidance. The most widely cited national standard for thermal performance of doors in the United States is NFRC 100 (NFRC 2004a), which does not address revolving doors or many types of overhead doors.

The thermal performance of emergency-exit doors and large rolling or sliding cargo doors is not considered at all in the current Handbook. In response to the needs expressed by its members and the HVAC community, ASHRAE issued a request for proposal with the following objectives:

* updating the information in chapter 30 of Fundamentals to include more door types,

* developing correlations to characterize convective flow and related heat transfer in enclosed chambers of revolving doors, for simplified calculation procedures,

* evaluating the variation in thermal performance as a function of temperature conditions and product size, and

* developing recommendations for testing and simulation of various door products, to provide guidance to designers for evaluation of products not in the current scope of work and to assist in the development of national standards to evaluate such products.

Part of the resulting project, ASHRAE 1236-RP, involved determining total product U-factors of seven representative specimens using computer simulation, for comparison to U-factor results from hot-box testing. If the results compared favorably, the computer models were assumed to be validated and were then used to provide a wider range of design values for similar products.

This paper describes the procedures for evaluating the U-factor of the seven specimens using computer simulation. A separate paper (McGowan et al. 2006) has been written to describe the testing procedures conducted in parallel to these simulations. The entire project is described in the final report for this research project, ASHRAE 1236-RP (ASHRAE 2006). The report includes test and simulation results for all seven specimens and design values generated for the ASHRAE Handbook--Fundamentals.

Test Specimens

All specimens were simulated as configured in the test chamber to allow direct comparison between test and simulation results. Therefore, it is appropriate to review the as-tested configurations to describe the simulation conditions. The project report describes all seven test specimens that were simulated, but space considerations restrict this technical paper to a discussion of only three products:

* Specimen 1 was a steel sectional overhead-type door (similar to Figure 1). The door sections were nominally 50 mm (2 in.) overall thickness, constructed of 24-gauge galvanized steel. Meeting rails were rabbeted to form a weather seal (but not an air seal) between sections. Each of the five sections in the door assembly is reinforced with two steel ribs, with end channels wrapped around each section and spot-welded in place. Galvanized hardware provided with the door included adjustable top roller brackets, end-roller, and intermediate hinges and galvanized steel roller channels on either side to represent typical installation. The roller channels normally extend along the ceiling of a garage, to guide the door sections as the assembly opens, but the channels in the test specimen had to be cut at the top of the door to avoid interference with the constant-temperature baffle and the convective flow of room-side air during the thermal testing. Figure 1 shows the installation for Specimen 2, but Specimen 1 was installed in the same way. All specimens were installed in the mask wall and moved with the mask wall from the air-leakage test chamber to the thermal test chamber as a unit.

[FIGURE 1 OMITTED]

The bottom of the door included a weather seal made of a U-shaped extruded vinyl bulb seal, typical of this type of installation. Perimeter seals at the jambs and head of the door assembly are optional accessories but are often not installed as they may interfere with door operation. The perimeter seals were not installed on this specimen.

* Specimen 3 was a three-wing glazed revolving door (Figure 2), comprising three panels, two curved sidewalls, and a canopy manufactured from aluminum extrusions. The sidewall construction utilized a vertical end post on each side and one post in the center (where the door assembly connects to the exterior wall of a building or in this case to the mask wall of the test chamber). The door panels are attached to the center shaft with two hangers on each wing, which allow the panels to be held in position with a preset tension and to be folded outward to allow for emergency egress.

[FIGURE 2 OMITTED]

The hardware that allows the door panels to collapse during an emergency exit is housed in the floor below the central shaft. This hardware cannot be removed from the specimen, as the integrity of the door panels rests on it. This necessitates that the specimen must be elevated by approximately 500 mm (20 in.). Therefore, the mask wall included a "mask floor," made of expanded polystyrene foam insulation. Due to the depth of this specimen, a special thermal chamber was constructed (Figure 2).

The door panels and curved sidewalls of the specimen were made of 1/4 in. clear tempered safety glass. The joints between the door panels and the outer barrel of the assembly are weatherstripped to reduce air and water leakage. Weatherstripping options include felt and horsehair. The latter, which is standard on this particular product, has been used historically as a weatherstripping material because it expands as it absorbs moisture so that the weather seal becomes tighter as the horsehair gets wet.

Specimen 7 was an aircraft hangar door, custom-built for this project (this type of product is typically custom-built to fit a specific application). The specimen was constructed of a 24-gauge hollow structural section (HSS) steel frame, comprising three panels (Figure 3). Each panel was built with a perimeter frame of 4.88 mm (3/16 in.) HSS and two intermediate horizontal mullions of 3 mm (1/8 in.) HSS. The exterior (weather-side) surface was clad with 26-gauge profiled steel, typical of the design of these systems, and the interior was finished with glass-fiber batt insulation and heavy-duty polyethylene. In a commercial application the interior finish would be an opaque (typically white) vinyl sheet product, but the project team wanted to be able to review the batts to ensure that no voids were present in the insulation.

[FIGURE 3 OMITTED]

A single astragal (i.e., where the door panels from either side of the hangar would normally meet in the middle of the rough opening) and a single meeting stile were included in the specimen to represent the typical assembly of these components. The specimen included a steel frame mounted onto the mask wall, as would normally be included in a field installation. All normal operating hardware (rollers, hangers, etc.) was included in the assembly.

The rest of the products analyzed in the RP-1236 project, which are not discussed in this paper, were:

* Specimen 2 was similar to Specimen 1 except that the door sections were insulated with pressure-injected polyurethane-foam and therefore were less interesting than Specimen 1 in terms of heat transfer characteristics and modeling challenges.

* Specimen 4 was a four-wing glazed revolving door, similar to Specimen 3 except with four door panels. Again, this product is less interesting than Specimen 3 from the perspective of computer simulation, as it is symmetrical and therefore requires fewer cross-sectional models to characterize its thermal performance.

* Specimen 5 was a metal coiling slat-type roll-up door, which presented challenges similar to those described for Specimen 1 as both products were uninsulated.

* Specimen 6 was an emergency exit door, which is already covered in the existing NFRC 100 procedures. Therefore, it did not present as many interesting challenges in the modeling procedures and is not discussed further in this paper.

COMPUTER SIMULATION METHODOLOGY AND RESULTS

All products were simulated using THERM (LBNL 2003) to determine thermal performance as characterized by the heat loss coefficient (total product U-factor). The simulation procedures followed the NFRC 100 (NFRC 2004a) protocol as closely as possible, but some modification was required to allow testing of certain products. There is no defined methodology to simulate the thermal performance of revolving doors, for example. In those cases, the underlying principles (described in Enermodal [1996]) were used to develop an appropriate strategy. In general terms, we created computer models of each component of each product and area-weighted the component U-factors to determine a total-product U-factor:

[U.sub.t] = ([U.sub.h] * [A.sub.h] + [U.sub.s] * [A.sub.s] + [U.sub.e] * [A.sub.e] + ...)/([A.sub.h] + [A.sub.s] + [A.sub.e] + ...), (1)

where U is the component U-factor, and A is the associated projected area of the head, sill, end-stile, etc., components, as denoted by the subscripts in Equation 1. Area-weighting was done using as-tested dimensions, with standardized test conditions of air temperature and surface heat transfer coefficients on the room side and weather side of the model components.

Uninsulated Sectional Garage Door

The simulation models used to represent this product (Specimen 1) are shown in Figure 4. These comprise separate models for the head, the left and right jambs, and the sill, with a separate model used to characterize a typical joint between two panels of the sectional door. Another model was used to characterize the center-of-panel condition, as shown in Figure 4.

[FIGURE 4 OMITTED]

The first simulation produced a total product U-factor of 4.39 W * [m.sup.-2] * [K.sup.-1] (0.77 Btu * [h.sup.-1] * [ft.sup.-2][degrees]F-1), well below the tested values ([U.sub.s] or [U.sub.st]), which indicated that the simulation model was not accounting for all of the heat transfer in the test specimen. The standard NFRC procedure for modeling sectional doors includes 63.5 mm (2.5 in.) of edge-of panel in the model, and it was conjectured that this might not capture all of the heat flow in a steel-skinned product (where the skin of the noninsulated door panel acts like a fin). When we increased the height of the edge-of-panel portion in the model, we observed an increase in the U-factor, but increasing the edge dimension beyond 150 mm (6 in.) produced a negligible increase in the result. We therefore determined that the larger edge dimension was necessary in this model to account for all of the heat transfer in the as-tested specimen.

When we calculated the total product U-factor with a 150 mm (6 in.) edge, the result was 4.84 W * [m.sup.-2] * [K.sup.-1] (0.85 Btu * [h.sup.-1] * [ft.sup.-2] * [degrees][F.sup.-1]). At that point in the project, the test and simulation results agreed to within 12.1%, so the results did not meet the validation criterion of the NFRC 100 procedure (NFRC requires agreement to be within 10%). Having exhausted all other possible reasons for the lack of agreement, the test and simulation laboratories were allowed to discuss the results in an attempt to resolve the discrepancy. It became apparent that the simulation assumption of a galvanized surface was incorrect, as the testing laboratory verified that the surface was coated with a primer. The primer-coat mimicked the dull gray finish of a galvanized surface, and it was only evident on close visual inspection that the surface was actually painted.

The simulation results were recalculated using a painted surface, which changes the surface emittance from 0.20 to 0.90. The resulting U-factor was 6.72 W * [m.sup.-2] * [K.sup.-1] (1.18 Btu.[h.sup.-1] * [ft.sup.-2].[degrees]F-1), well beyond the NFRC "equivalence" criterion. To examine the reasons for this difference, the project team looked at the simulation results and separated the total product thermal transmittance (as predicted by simulation) into the specimen conductance and room-side and weather-side film coefficients. Unfortunately, the software used to determine total product U-factors via computer simulation (LBNL 2003) does not provide explicit values for the total room-side film coefficient, so the values had be backed out of the results.

Consider an energy balance on the specimen: the heat flow from the room to the specimen can be expressed as Q = [h.sub.is] * [A.sub.s] * ([T.sub.ia]-[T.sub.is]), and the overall heat transmission through the specimen is Q = U* Ap* ([T.sub.ia]-[T.sub.oa]). Equating these flows, we get [h.sub.is]* [A.sub.s] * ([T.sub.ia]-[T.sub.is]) = U * [A.sub.p] * ([T.sub.ia]-[T.sub.oa]), and solving for the total room-side film coefficient,

[h.sub.is] = U * [[([T.sub.ia] - [T.sub.oa])]/[([T.sub.ia] - [T.sub.is])]] * [[A.sub.p]/[A.sub.s]], (2)

where U is the total product U-factor, either from test or simulation; [T.sub.ia], [T.sub.oa], and [T.sub.is] are the temperatures of the indoor and outdoor air and indoor surface; and [A.sub.p] and [A.sub.s] are the projected and actual wetted surface areas of the specimen.

A similar derivation can be used to produce the exterior film coefficient, but the value of interest is on the room side, as this is the controlling resistance of the non-insulated specimen. Most of the above variables are easily obtained: from the physical configuration of the specimen ([A.sub.p] and [A.sub.s]) or the specified boundary conditions ([T.sub.ia] and [T.sub.oa]). The U-factor is produced from the simulation, and it only remains to obtain a mean surface temperature for the corresponding U-factor. Unfortunately, THERM does not make this easy, especially where the interior surface is irregular. To obtain an approximation of this value, we took advantage of the physics of the situation. As this is a high-conductance specimen that is nearly isothermal, according to the thermocouple readings from the test and the preliminary simulation results, we made the approximation that the room-side surface temperature is equal to the weather-side surface temperature (i.e., [T.sub.is] [approximately equal to] [T.sub.os]). This approximation is seen to be intuitively true for most of the specimen, as it is essentially a single layer of 24-gauge galvanized steel (except at joints between slats and at the perimeter of the specimen).

To simplify our analysis, we focused on the center-of-panel, which comprises two-thirds of the specimen's area. This allows us to approximate the room-side film coefficient without the geometric complications that occur at the perimeter of the specimen (including the mounting track and mask-wall interface). The convective portion of the room-side film coefficient is assumed in THERM to be constant for a given type of frame, so the radiative component can also be isolated. The room-side film coefficient is determined in the test according to Clause 8.2.9.1 of NFRC 102 (NFRC 2004b):

[h.sub.STh] = 1.46[[[([T.sub.ia] - [T.sub.is])]/H].sup.0.25] + [sigma][[epsilon].sub.1][[[[([T.sub.ia] + 2.73.16)].sup.4] - [[([T.sub.is] + 273.16)].sup.4]]/[([T.sub.ia] - [T.sub.is])]] (3)

where H is the height of the specimen, [sigma] is the Stefan-Boltzmann constant, and [epsilon] is the mean surface emittance of the specimen.

It is interesting to note that, although NFRC 102 requires that the emittance is to be used in the above equation, it does not specify anywhere whether to measure this value or use look-up tables. For the purposes of this project, the test facility followed its standard protocol, which is to use a typical value of 0.84, but the procedure should be specified more clearly in the standard.

The first term in Equation 3 is convective and the second part is radiative, so the components of the film coefficient as determined from testing can also be isolated. Thus, total and component parts of room-side film coefficients from test and simulation can be compared, as is summarized in Table 1.

The documentation provided with THERM provides no explicit description of the models used but notes that the software follows the methods described in ISO 15099 (ISO 2002). In this standard, the convective portion of the room-side film coefficient is calculated from a correlation based on the Rayleigh number (which in turn depends on surface temperature, air temperature, and the height of the specimen). The radiative component of the film coefficient is determined using a modified radiosity model, with view factors calculated using Hottel's crossed-string method. Hottel's method provides a two-dimensional approximation of the view factors between three-dimensional surfaces that exchange heat via radiation and introduces a small error for the sake of increasing computational speed. The three-dimensional nature of the large specimens considered in this project is only important at the perimeter of the specimens, where some of the surfaces of the specimen exchange heat via radiation with other surfaces in the specimen. As this is a relatively small proportion of the overall specimen area, the error introduced by the Hottel approximation should be relatively small.

As noted, the THERM program assigns a fixed value to the convective portion of the film coefficient, depending on the frame type selected. For this case, the value is 3.29 W * [m.sup.-2] * [K.sup.-1] (0.58 Btu.[h.sup.-1] * [ft.sup.-2] * [degrees][F.sup.-1]), based on an assumed specimen height of 1 meter (3.3 feet) and an assumed mean surface temperature typical of a given frame type.

Although the simulation and test methods provide two different approaches to determine the film coefficient, they should produce similar results. Table 1 indicates that they do not, however, and it would be useful to determine why this is so. As we do not have a clear explanation of how THERM assigns its values, we have used Equation 3 to investigate the effect of the base assumptions on the results. We do this by adjusting each of the variables in the test procedure until they exactly match the conditions in the simulation to see whether agreement is obtained.

Table 1. Film Coefficients for Specimen 1

                                       Simulation (Galvanized)

                                Test      Result     % Diff.
                               Result                  From
                                                       Test

Surface temperature            -2.65   -12.6 (9.3)     N/A
[T.sub.is], [degrees]C         (27.2)
([degrees]F)

U-factor,                       5.43   3.78 (0.67)   -43.7%
W*[m.sup.-2]*[K.sup.-1]         (0.96)
(Btu*[h.sup.-1]*
*[ft.sup-2]
[degrees]][F.sup.-1])

Weather-side film coefficient   30.0   25.77 (4.54)  -16.4%
[h.sub.0],                     (5.28)
W*[m.sup.-2]*[K.sup.-1]
(Btu*[h.sup.-1]*
[ft.sup.-2]*
[degrees][F.sup.-1]F)

Room-side film coefficient      6.75   4.13 (0.73)   -63.4%
[h.sub.i],                     (1.19)
W*[m.sup.-2]*[K.sup.-1]
(Btu*[h.sup.-1]*
[ft.sup.-2]*
[degrees][F.sup.-1]F)
- total

- convective component          2.44   3.29 (0.58)   +25.8%
                               (0.43)

- radiative component           4.31   0.84 (0.15)    -413%
                               (0.76)

                                Simulation
                                 (Painted)

                               Result  % Diff.
                                         From
                                         Test

Surface temperature            -10.0     N/A
[T.sub.is], [degrees]C         (14.0)
([degrees]F)

U-factor,                       6.72   +19.2%
W*[m.sup.-2]*[K.sup.-1]         (1.18)
(Btu*[h.sup.-1]*
*[ft.sup-2]
[degrees]][F.sup.-1])

Weather-side film coefficient  30.93    +3.1%
[h.sub.0],                     (5.45)
W*[m.sup.-2]*[K.sup.-1]
(Btu*[h.sup.-1]*
[ft.sup.-2]*
[degrees][F.sup.-1]F)

Room-side film coefficient      7.97   +15.3%
[h.sub.i],                     (1.40)
W*[m.sup.-2]*[K.sup.-1]
(Btu*[h.sup.-1]*
[ft.sup.-2]*
[degrees][F.sup.-1]F)
- total

- convective component          3.29   +25.8%
                               (0.58)

- radiative component           4.68    +7.9%
                               (0.82)

The base-case test conditions ([T.sub.ia] = 21.29[degrees]C or 70.3[degrees]F, [T.sub.is] = -2.65 [degrees]C or 27.2[degrees]F, H = 3048 mm or 120 in., and [epsilon] = 0.84) were varied, one parameter at a time, to examine the resulting changes in the total room-side film coefficient and its components. The results are listed in Table 2. As Table 2 shows, the total film coefficient from testing can vary from ?8.5% to +7.3%, relative to the simulation result, depending on the values chosen for the variables. There could be a combination of values for these variables that would provide an exact match between test and simulation, but that would be begging the question. Even the closest agreement between test and simulation in Table 2 is somewhat misleading, as it is the result of errors in the components (?25.8% on the convective portion and +12.5% on the radiative portion) canceling each other to produce a net difference of -4.6%.

Table 2. Adjustment of Film Coefficients for Specimen 1

                         Total Film Coefficient
                              [h.sub.STh]

                    W*[m.sup.-2]*[K.sup.-1]      Diff. from
                       (Btu*[h.sup.-1]           Simulation
               [ft.sup.-2]*[degrees][F.sup.-1])

Base case                           6.75 (1.19)       -8.5%

Air                                 6.73 (1.19)       -8.8%
temperature =
21[degrees]C
(simulation
standard)

Emittance =                         7.04 (1.20)       -4.6%
0.90 (default
value)

Specimen                            7.82 (1.38)       +6.0%
height = 1 m
(THERM
assumption)

[T.sub.is] =                        7.88 (1.39)       +6.8%
-10[degrees]C
(to match
simulation
result)

[epsilon] =                         4.43 (0.78)       +7.3%
0.20
(to match
galvanized
simulation)

                          Convective Component

                   W*[m.sup.-2]*[K.sup.-1]       Diff. from
                       (Btu*[h.sup.-1]*          Simulation
               [ft.sup.-2].[degrees][F.sup.-1])

Base case                           2.44 (0.43)      -25.8%

Air                                 2.44 (0.43)      -25.8%
temperature =
21[degrees]C
(simulation
standard)

Emittance =                         2.44 (0.43)      -25.8%
0.90 (default
value)

Specimen                            3.22 (0.57)       -2.1%
height = 1 m
(THERM
assumption)

[T.sub.is] =                        3.45 (0.61)       +4.9%
-10[degrees]C
(to match
simulation
result)

[epsilon] =                         3.45 (0.61)       +4.9%
0.20
(to match
galvanized
simulation)

                          Radiative Component

                   W*[m.sup.-2]*[K.sup.-1]       Diff. from
                       (Btu*[h.sup.-1]*          Simulation
               [ft.sup.-2].[degrees][F.sup.-1])

Base case                           4.31 (0.76)       +5.4%

Air                                 4.29 (0.76)       +4.9%
temperature =
21[degrees]C
(simulation
standard)

Emittance =                         4.60 (0.81)      +12.5%
0.90 (default
value)

Specimen                            4.60 (0.81)      +12.5%
height = 1 m
(THERM
assumption)

[T.sub.is] =                        4.43 (0.78)       +8.3%
-10[degrees]C
(to match
simulation
result)

[epsilon] =                         0.98 (0.17)      +16.7%
0.20
(to match
galvanized
simulation)

The largest difference in successive changes to the total film coefficient occurred when the specimen height was changed from 3048 mm or 120 in. (the actual specimen height) to 1 m or 39.4 in. (the assumed height in THERM). This suggests that the one-meter assumption in THERM is a significant source of the differences between the test and simulation results for these products. This might not be apparent for typical residential products, which are closer to the assumed 1 m (39.4 in.) in height and for which the controlling thermal resistance is not the room-side film coefficient. For typical residential specimens, the error introduced by the one-meter assumption may not be significant, but it appears to be a problem for large, high-conductance products such as Specimen 1. It would be useful if THERM were modified to allow users to input the actual specimen height.

Changing the surface emittance from 0.84 (the value used in the test) to 0.90 (the default value in NFRC 101, Appendix A) made a difference of approximately 4% in the total film coefficient, relative to the simulation. The relative change in the radiative component of the film coefficient was rather large, as the difference between the test and simulation results went from +4.9% to +12.5%, but the difference in the total-product U-factor would be less than 4%. The difference between = 0.90 (painted surface) and = 0.20 (galvanized surface) is quite significant, as noted previously, and the difference between the radiative component of the film coefficient as predicted by Equation 3 and the value predicted by simulation doubled from +8.3% to +16.7%.

Changing the room-side surface temperature [T.sub.is] from the value measured in the test (-2.65[degrees]C or 27.2[degrees]F) to the value predicted by simulation (-10[degrees]C or 14[degrees]F) produced a significant increase in the difference between the test and simulation results for the convective portion of the film coefficient and a slightly smaller decrease in the radiative component. The net result was a slight increase in the difference between test and simulation results for the total film coefficient.

In summary, the test and simulation methods use two different approaches to determine film coefficients. The derivation used in the test method is based on an empirical correlation, whereas the simulation method uses a model based on fundamental heat-transfer principles but with some simplifying assumptions. In the case of a noninsulated specimen (e.g., Specimen 1), where the controlling resistance is the room-side film coefficient, the two different approaches produce results that are too dissimilar to provide a sufficient level of comfort to rely on the computer simulation as being representative of the test.

Three-Wing Revolving Door

As might be expected, the unusual configuration of this door type introduces more complexity to the simulation procedure. Apart from the natural convection inside the pie-shaped chamber formed by the closed revolving door, convective and radiative heat transfer on the weather-side and room-side of the specimen will be unusual. The interesting geometry formed by the intersection of the door cylinder with the wall in which it is mounted and by the center V formed by the closed door panels will produce complex airflow patterns on the exterior surface. This is not apparent in the U-factor, however, because the calculation of film coefficients in ASTM C1199 (ASTM 2005) assumes that they are applied uniformly over the (room-side and weather-side) surfaces of the specimen. The same uniform application of film coefficient occurs in the simulation procedure, so local variations in these conditions (which almost certainly occur in the test specimen) are not considered in the overall heat transfer evaluation. Thus, the procedure would not be appropriate for evaluating local surface temperatures for condensation potential.

The method for determining the U-factor of a revolving door using computer simulation follows the general underlying philosophy behind the original NFRC simulation procedures. As noted above, the basic concept of the calculated U-factor involves an area-weighted average of all component U-factors, where an individual component model (frame and "edge-glass" assembly) defines a unique cross section of the total product. The difficult part is in determining which cross sections should be used to characterize the component parts of the overall product.

In the case of a three-wing revolving door, for example, the left and right head sections would represent unique components of the overall assembly. Representative two-dimensional models of these sections would include a portion of the mask wall above the product, the canopy, and the upper part of the door panel (incorporating the edge-glass region of the door panel, at least 63.5 mm or 2.5 in. away from the sightline). Similarly, the left and right sill sections and left and right jambs (Figure 5) would present different surface areas on the warm and cold sides of the test chamber, and separate simulation models are therefore required. The spindle at the center of the door is another section that requires a separate model to properly characterize heat flow through this unique portion of the door.

[FIGURE 5 OMITTED]

The most difficult part of characterizing heat flow in a revolving door using a two-dimensional model is that there is no way to capture the effect of convection and radiation within the enclosed door chamber. In the special case of a revolving door, the unusual configuration of the door type introduces more complexity to the simulation procedure. Natural-convection cells will develop inside the pie-shaped chamber formed between the closed door panels and the surrounding barrel (the barrel is formed by the curved panels that enclose the central revolving component). The computational software used in the NFRC procedures is a simplified two-dimensional program that does not explicitly model convective heat transfer. Rather, all components are modeled as solid components with a constant thermal conductivity. Increased heat transfer due to convective motion in a fluid can be modeled by assigning an "effective thermal conductivity" to the component. The difficulty is in knowing what value to assign to the conductivity that will properly account for convective heat transfer.

To solve this problem, an explicit three-dimensional model of the enclosed air chamber in the door was developed using a computational fluid dynamics (CFD) software program. The software models convective and radiative heat transfer and can be used to examine characteristic heat flows for a range of geometries and boundary conditions. A CFD model of the door chamber and enclosing solid components (e.g., door glazing, aluminum framing elements) was run with typical room-side and weather-side boundary conditions to obtain overall heat flow. Then the CFD model was run with a constant value of thermal conductivity for the door chamber, [[lambda].sub.eff]. The value of [[lambda].sub.eff] that produced the same heat flow rate as the explicit CFD model was considered to be the average effective thermal conductivity of the air cavity. This value was used in a simplified two-dimensional model to obtain average heat flow (and thus the component U-factor) for all appropriate cross sections. Although this method would not be acceptable if we were using computer simulation to predict surface temperatures, it was considered reasonable for determining U-factors, as the latter value is representative of average heat flow through the entire specimen.

The simulation produced a total product U-factor of 4.53W*[m.sup.-2]*[K.sup.-1] (0.80 Btu*[h.sup.-1].[ft.sup.-2].[degrees][F.sup.-1]), which is 2.9% lower than the test result Using the area-weighted method and 19% higher than the test result for the CTS method. As stated previously, the ASTM procedure specifies the area-weighting method. Thus, this approach appears to provide a very accurate representation of the thermal performance of the three-wing revolving door. It would appear that this simulation approach is appropriate for this type of product but, because of the approach Used in modeling the "center-of-glass" area Using a CFD model specific to that size and boundary condition, it cannot be extrapolated to other door sizes or temperature regimes.

Rolling Warehouse/Aircraft Hangar Door

The simulation models Used to represent Specimen 7 are shown in Figures 6 and 7. The initial simulation produced a U-factor of 3.31 W*[m.sup.-2]*[K.sup.-1] (0.58 Btu.[h.sup.-1].[ft.sup.-2].[degrees][F.sup.-1]), much higher than the tested values ([U.sup.s] or [U.sup.st]), indicating that the simulation model accounted for too much heat transfer in the test specimen.

[FIGURE 7 OMITTED]

At several cross sections, the weatherstripping provided the only separation between the interior and exterior chambers in the hot box. In terms of heat transfer, the significant element of thermal resistance (i.e., the "controlling resistance") was the room-side film coefficient. Therefore, the room-side film coefficient was suspected as a source of error for the simulation model. The model automatically applies the room-side film coefficient to the room-side surface of the cross section, as shown on the left-hand side of Figure 6. This is not realistic, as it is more likely that pockets of still air would be trapped in the smaller recesses between the door and the frame, as suggested on the right-hand side of Figure 6. Also, the design drawings provided by the manufacturer indicated a center-ofpanel thickness of 100 mm (4 in.) for the batts, but field measurement of the specimen indicated that the actual thickness was 150 mm (6 in.).

[FIGURE 6 OMITTED]

The simulation model was modified to better represent the trapped air cavities at the warm-side surfaces, as shown at right in Figure 6. Also, the batt insulation thickness was adjusted to match the measured value of the test specimen. With these adjustments, the simulation total-product U-factor was 2.27 W*[m.sup.-2].[K.sup.-1] (0.40 Btu*[h.sup.-1].[ft.sup.-2].[degrees][F.sup.-1]). This is 7% higher than the areaweighted test result and 15.4% higher than the CTS test result. The area-weighting method is preferred in ASTM C1199 (ASTM 2005), so simulation appears to provide an accurate representation of the thermal performance of this specimen.

SUMMARY AND DISCUSSION

This paper presents simulation results for several nonresidential door assemblies that are among the most difficult to test or simulate using conventional methods. Indeed, modifications to existing test procedures are necessary to measure the thermal transmittance of these products, as described in a companion paper. The research project upon which this paper is based suggests modifications to the existing procedures for simulating large nonresidential door assemblies and provides new procedures for modeling doors that were not previously considered in the relevant standards. These procedures appear to be reasonably accurate even for complex products, such as revolving doors and aircraft hangar doors, but not for highconductance products, such as uninsulated garage doors. The computer model that is currently used in the US standard for evaluating windows does not appear to be able to accurately model the room-side film coefficient, which is the controlling resistance in high-conductance products.

The interested reader is directed to the report for the research project upon which this paper is based (ASHRAE 2006). It is intended that the air-leakage and U-factor data arising form this project will be included in the 2009 ASHRAE Handbook--Fundamentals.

ACKNOWLEDGMENTS

The RP-1236 project team acknowledges the participation of the sponsors who donated test specimens and in some cases paid for shipping and customs brokerage: Boon Edam Corp., Salt Lake City, UT; C J Rush Industries Ltd., Markham, ON; de la Fontaine Doors, Buckingham, QC; Garaga Industries Ltd., St. Georges, QC; Overhead Door Corporation, Dallas, TX; Steel-Craft Door Products Ltd., Edmonton, AB; and A T Spec-Dor, Candiac, QC. We also appreciate the sponsorship of ASHRAE Technical Committee 4.5 (Fenestration) and the collaboration of the TC4.5 Project Monitoring Subcommittee: Jeff Baker, WESTLab, London, ON; Dr. D.C. Curcija, University of Massachusetts at Amherst, MA; Marcia Falke, Keystone Certifications, Inc., York, PA; Joseph R. Hetzel, DASMA International, Cleveland, OH; John Hogan (subcommittee chair), City of Seattle Department of Planning and Development, Seattle, WA.

We also gratefully acknowledge the financial support of ASHRAE, Inc., and its generous donors, whose contributions enable projects such as this one to succeed.

REFERENCES

ASHRAE. 2005. 2005 ASHRAE Handbook--Fundamentals, Chapter 30. Atlanta: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.

ASHRAE. 2006. Heat transfer through roll-up doors, revolving doors, and opaque nonresidential swinging, sliding and rolling doors. Final report, ASHRAE Research Project RP-1236. Atlanta: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.

ASTM. 2005. ASTM C1199-00, Standard Test Method for Measuring the Steady-State Thermal Transmittance of Fenestration Systems Using Hot Box Methods. West Conshohocken, PA: ASTM International.

Enermodal. 1996. Procedures for modeling the energy performance of building components: Doors and walls. Enermodal Engineering Ltd., Kitchener, ON.

ISO. 2002. UISO/FDIS 15099, Thermal Performance of windows, doors and shading devices--Detailed calculations. Geneva, Switzerland: ISO Central Secratariat.

LBNL. 2003. THERM 5.2, A PC Program for Analyzing Two-Dimensional Heat Transfer through Building Products. Berkeley, CA: Regents of the University of California.

McGowan, A.G., R. Jutras, and G. Riopel. 2007. Testing of air leakage and heat loss characteristics of commercial door assemblies (ASHRAE RP-1236). ASHRAE Transactions 113(2).

NFRC. 2004a. NFRC100-2004, Procedure for Determining Fenestration Product U-Factors. Silver Spring, MD: National Fenestration Rating Council.

NFRC. 2004b. NFRC 102-2004, Procedure for Measuring the Steady-State Thermal Transmittance of Fenestration Systems. Silver Spring, MD: National Fenestration Rating Council.

Alex McGowan, PEng

Member ASHRAE

Alex McGowan is the Practice Area Leader in the Building Science Division of Levelton Consultants Ltd., Victoria, BC, Canada.