CFD modelling of 2-D heat transfer in a window construction

including glazing and frame

Morten V. Vendelboe, Ph.D student
Technical University of Denmark – Department of Civil Engineering
mvv@[Link]

Svend Svendsen, Professor

Technical University of Denmark – Department of Civil Engineering
ss@[Link]

Toke R.. Nielsen, Associate Professor

Technical University of Denmark – Department of Civil Engineering
trn@[Link]

KEYWORDS: fenestration, window, numerical simulation, thermal transmittance, glazing temperature, CFD
SUMMARY:
A numerical model for the simultaneous calculation of 2-D heat transfer and indoor surface temperatures for a
three layer glazing unit and a frame construction based on a commercial CFD code is presented. The calculated
heat flow and the surface temperatures of the inner glazing layer are compared to results obtained using the
finite element code Therm 5.2 by calculations according to EN ISO 10077-2 and ISO 15099. The models
compared produce almost identical results with respect to the overall thermal transmittance of the glazing and
frame, whereas differences in the range 3-7% are found for the thermal transmittance of the frame part only.
The calculated distributions of surface temperatures for the inner pane also agree quite well when convection in
the glazing cavities is accounted for in the finite element calculation.

1. Introduction
Window design affects energy consumption in buildings significantly with respect to both cooling and heating
demand. Sensible design solutions to improve the thermal properties of windows are therefore crucial in the
effort to reduce the overall energy consumption in buildings.
One of the most widely used tools (Gustavsen 2005, Song 2007) for detailed analysis of the thermal properties of
the connection between glazing and window frames is the 2-D finite element code Therm5.2 (Finlayson 1998).
The code can be used for calculation of the thermal transmittance of glazing and frame connections according to
the EN ISO 10077-2 (CEN 2003) and the ISO 15099 (ISO 2003) standards. The EN ISO 10077-2 procedure for
calculation of thermal transmittance is however not suited for determination of condensation risk due to low
inside surface temperatures but a suitable alternative is not prescribed in the European standards (CEN 2001).
A condensation resistance model has however been implemented in Therm for more precise calculation of
indoor surface temperatures (Zhao 1996, Mitchell 2003, Kohler 2003) in compliance with North American
standards (NFRC 2004).
In the following the preliminary investigations regarding the possibility to apply a commercial CFD code for
detailed 2-D analysis of heat transfer and simultaneous calculation of indoor surface temperatures is presented.
Calculated thermal transmittances as well as indoor surface temperature distribution are compared to Therm
models to observe the difference due to more comprehensive modelling of the cavities of the glazing and frame.

-1-
2. Method
2.1 Geometry and materials
The modelled window cross section consists of three 4 mm glass layers and two 12 mm air cavities. The panes
have an emissivity of 0.84 on all surfaces except the outer surface of the inner pane which has a low emissivity
coating with an emissivity of 0.037. Since the focus is on model comparison, and not on calculating the
properties of an actual window, all other emissivities are set to 0.9. For the same reason the detailed geometry
and thermal properties of the spacer are represented by a box of an equivalent fixed thermal conductivity. The
applied thermal properties of the materials used in the calculations are listed in TABLE. 1. The model geometry
is displayed in FIG. 1.

FIG. 1: Figure showing the cross section of the window. Left: The sill and the lower part of the glazing unit.
Right: The entire model.

TABLE. 1: Material properties used for the calculations.

Material Conductivity Emissivities

[W/mK] [-]
Aluminium 160 0.9
Wood 0.13 0.9
EPDM 0.25 0.9
Sealing 0.4 0.9
Spacer 0.246 0.9
Glass 1 0.84
Glass with low emissivity coating 1 0.037

-2-
2.2 Boundary conditions
For all models boundary conditions are chosen according to EN ISO 13788. Constant heat transfer coefficients
of 7.69 W/m2K and 25 W/m2K are assigned to the internal and external boundaries respectively. The indoor
temperature is set to 20 °C and the outdoor temperature to 0 °C.

2.3 Glazing
For the finite element calculations thermal properties of the glazing unit are calculated in Window5.2 (Mitchell
2001) and imported to Therm.
To ensure that the glazing unit, modelled using the CFD code, has the same thermal properties as the glazing unit
imported for the finite element calculations, a comparison is made for the two models of the glazing unit alone.
In both cases boundary conditions according to section 2.2 are applied for the internal and external boundaries
while the top and bottom of the glazing is adiabatic. The height of the glazing is 0.5 m. The calculated thermal
transmittance of the glazing using Window is 1.25 W/m2K.
The calculated thermal transmittance for the CFD model is 1.27 W/m2K. The calculated thermal transmittances
of the two models are in very good agreement with a difference of 1.6 % and are thus used in the integrated
calculations for glazing and frame. The settings for the separate CFD model for the glazing are presented in
section 2.5.
For the finite element calculations two calculations are performed for each model. For calculations of thermal
transmittance (U-value) the air in the glazing cavities is replaces by a solid material with a thermal conductivity
that is calculated based on the heat transfer due to convection and radiation across the cavity. For calculation of
indoor surface temperatures a convection model for the local temperature distribution inside the glazing cavities
developed for Condensation Resistance rating according to NFRC 500 is applied.

2.4 Frame cavities

While the thermal properties for the solid domains are identical for the finite element models and the CFD
model, the cavity modelling is different. In the first two finite element models the frame cavities are modelled
according to EN ISO 10077-2 where a simplified radiation model is prescribed. These models are referred to as
‘CEN CR’ and ‘CEN U-value’. In the other two models the frame cavities are modelled according to ISO 15099
including a detailed view factor based radiation model. These models are referred to as ‘ISO CR’ and ‘ISO U-
value’.
In all cases the cavities are defined as non ventilated to match the assumptions made in the CFD model.
The settings for the CFD model for the convective and radiative heat transfer in the frame cavities is presented in
section 2.5

2.5 CFD settings and mesh

2.5.1 CFD settings
A list of settings for the CFD model is presented in TABLE 2. The settings are used for both the separate model
for the glazing and the model including both the frame and the glazing.

-3-
TABLE. 2: Settings for CFD model.

Solver Stationary
Viscous model Laminar
Fluid thermal properties Density modelled according to the Boussinesq approximation. All other
properties are constant.
Discretization schemes Pressure Standard
Momentum 2nd order
Energy 2nd order
Radiation model Discrete Transfer Radiation Model (DTRM)

Though the PRESTO! discretization scheme for pressure is recommended for modelling of natural convection
(Fluent Inc. 2005) convergence is only achieved when the standard scheme is applied for the model including
both the frame and the glazing. A comparison for the separate glazing model does not however show any effect
of enabling the PRESTO! discretization scheme for pressure.

2.5.2 Mesh
The calculation of heat transfer across the window construction including glazing and frame is performed with a
triangular mesh of ~120,000 cells. The solution is grid independent with respect to the total thermal
transmittance of the glazing and frame to less than 1 %. Grid independence could probably be achieved with a
lower number of cells, but due to the complex nature of the geometry a fine mesh is needed to reduce the number
of poor quality cells. Thus critical mass imbalances due to poor mesh quality within key areas of the frame
cavities occurs at both ~30,000 and ~60,000 cells. While calculation time is not comparable to that of Therm
(order of seconds) the model does however converge in less than 1 hour (Intel Centrino Duo, 2 GHz, 4 GB).

3. Results and discussion

3.1 Comparison of total thermal transmittance of glazing and frame
The calculated total thermal transmittance of glazing and frame is presented in
TABLE. 3. Since the modelled thermal transmittance of the glazing in the CFD model is 1.6 % higher than for
the glazing imported in the finite element calculations and the thermal transmittance of the glazing make up the
majority of the total thermal transmittance the results for the CFD model have been corrected for appropriate
comparison. As TABLE. 3 shows the calculated results are essentially identical whether or not the CFD results
are corrected.

TABLE. 3: Comparison of calculated total thermal transmittance of window cross section including glazing and
frame

Model Thermal transmittance Relative difference to CFD

model
(corrected with regard to Ugl)
[%]
[W/m2K]
CFD 1.38 -
CEN U-value 1.40 1.7
ISO U-value 1.39 0.8

-4-
3.2 Comparison of thermal transmittance of the frame
By subtracting the ‘1-D’ thermal transmittance from the separate glazing models calculated in section 2.3 from
the total thermal transmittances presented in section 3.1 the following results are obtained for the thermal
transmittance of the frame including the spacer.

TABLE. 4: Comparison of calculated thermal transmittance of frame

Model Thermal transmittance Relative difference to CFD model

2
[W/m K] [%]
CFD 1.95 -
CEN U-value 2.08 6.7
ISO U-value 2.01 3.0

Though larger discrepancies are found compared to the total thermal transmittance of glazing and frame the
results shown in TABLE. 4 still agree very well. The reason why the results agree so well despite the different
treatment of the heat transfer in the frame cavities is probably that the size and temperature difference across the
cavities result in only very limited convection as illustrated in FIG. 2.

FIG. 2: Calculated velocity distribution for the CFD model for the frame cavities and the lower part of the
glazing cavities.

-5-
3.3 Calculated indoor surface temperature distribution of the glazing
The calculated surface distribution for the inside of the innermost glazing for the 5 models is shown in FIG. 3.
The x-axis shows the vertical distance from the top of the glazing (x=500 mm denotes the lower edge of the
daylight opening). Since the results from the CEN and ISO U-value models as well as the CEN and ISO CR
models are practically identical, only one plot is included for each model category.

100
D istan ce fro m to p o f g lazin g [m m ]

200

CFD
300 CEN/ISO U-value
CEN/ISO CR

400

500

600
12.5 13.5 14.5 15.5 16.5 17.5
Surface temperature [deg-C]

FIG. 3 : Calculated surface distribution on the inside of the innermost glazing for the 5 models as a function of
the vertical distance from the top of the glazing.
As FIG. 3 shows, the results for the CFD model and the CR models are very similar regarding the overall
tendency as well as local values with a predicted minimum temperature of 13.3 °C and 12.9 °C respectively. The
U-value models on the other hand display similar surface temperatures on average but poor agreement
concerning both overall tendency and local values in certain points. Due to the omission of the convective effects
in the glazing cavities the predicted minimum temperature at the lower daylight opening is higher than predicted
by the CFD and CR models with 13.8 °C.

4. Conclusions and further work

The conclusions are as follows:
• There is excellent agreement between the calculated values for the total thermal transmission whether
calculated using the CFD model, the CEN U-value model or the ISO U-value model.
• There is good agreement between the three model types for the thermal transmittance of the frame with
differences of less than 7 % in all cases.

-6-
 • There is good agreement between the calculated indoor surface temperatures of the glazing for the CFD
model and the CR models whereas the U-value models perform poorly with respect to both overall
tendency and local values.

Given the overall agreement with the results based on the simplified finite element modeling of the glazing and
frame cavities the application of the more comprehensive and time consuming CFD modelling is not justifiable
for the given case. This is however believed to be attributed to the limited convection occurring in the glazing
and frame cavities. Further work will therefore include similar comparisons for window constructions where
convection is more significant for the overall heat flow as for window solutions with larger glazing distances and
cases where non laminar flow conditions may occur.
In addition the comprehensive modelling capabilities of the CFD software facilitate further investigation of the
impact of detailed modelling with respect to variable radiative and convective indoor boundary conditions for
more precise indoor surface prediction in particular.

5. References
European Committee for Standardization (2003). EN ISO 10077-2: Windows, doors and shutters – Calculation
of thermal transmittance – Part 2: Numerical method for frames.
European Committee for Standardization (2001). EN ISO 13788: Hygrothermal performance of building
components and building elements - Internal surface temperature to avoid critical surface humidity and
interstitial condensation - Calculation methods.

Finlayson, E., Mitchell R., Arasteh D., Huizenga C. and Curcija D. (1998). THERM 2.0: Program description. A
PC program for analyzing the two-dimensional heat transfer through building products. University of
California, Berkeley, CA, USA.

Fluent Inc. (2005). Fluent 6.2 Users Guide.

Gustavsen A., Uvsløkk S. and Jelle P. (2005). Numerical and experimental studies of the effects of various
glazing spacer on the windows and the glazing temperature, Proceedings of the 6th Symposium on
Building Physics in the Nordic Countries, Reykjavik, Iceland, June 13-15 2005.

International Organization for Standardization (2003). ISO 15099:2003(E) - Thermal performance of windows,
doors and shading devices – Detailed alculations.

Kohler C., Arasteh D. and Mitchell R. (2003). THERM simulations of window indoor surface temperatures for
predicting condensation, ASHRAE 2003 Winter meetings CD, Technical and SymposiumPapers,Vol.109,
Issue 1, p. 587-593.

Mitchell R., Kohler C., Arasteh D., Carmody J., Huizenga C. and Curcija D. (2003). Therm5/Window5 NFRC
Simulation manual, Lawrence Berkeley National Laboratory, Berkeley, CA, USA.

Mitchell R., Kohler C., Arasteh D., Huizenga C., Yu T., and Curcija D. (2001). WINDOW 5.0 User Manual.
Lawrence Berkeley National Laboratory, Berkeley, CA, USA.

National Fenestration Rating Counsel (2004). NFRC 500: Procedure for determining fenestration product
condensation resistance values.

-7-
Song S., Ho J.,Yeo M., Kim Y. and Song K. (2007). Evaluation of inside surface condensation in double glazing
window system with insulating spacer: A case study of residential complex, Building and Environment,
Vol. 42, p. 940-950.

Zhao Y., Curcija D. and Goss P. (1996). Condensation Resistance Validation Project – Detailed Computer
Simulations Using Finite-Element Methods. ASHRAE Transactions, Vol.102 Issue 2, p. 508-515.

-8-