ADDITIONAL INSULATION MATERIALS IN A WINDOW FRAME: EXPERIMENTAL AND CFD ANALYSES
Article Category: review
Published Online: Aug 06, 2019
Page range: 149 - 157
Received: Apr 09, 2019
Accepted: May 11, 2019
DOI: https://doi.org/10.21307/acee-2019-031
Keywords
© 2019 Piotr KOPER et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
In recent decades, there has been an increased demand for lowering the energy consumption in the buildings. When it comes to thermal insulation, one of the weakest parts of the structure are windows – they can be a cause of about 35% of the heat energy demands in the building [1]. For this reason, many researches have been conducted to determine ways to increase heat insulation of windows: both their glazing [1–5] and the frame [7–9].
One of the ways to decrease heat losses through the window is to incorporate a layer of thermal insulation into the framestructure. It is especially easy in case of PVC windows, as their profiles have a lot of empty space. While it can be assumed that the additional insulation will decrease the heat losses, it is necessary to determine how big the improvement will be, and if it is economically viable.
One of the most important aspects of these studies is determination of the flux of heat passing through the window in any given condition. The most popular way of obtaining this value is an experimental study using the “hot box” method [10]. Its basic principle is to create artificial environment on each side of the tested sample with controlled temperature and air movement. By measuring the amount of heat supplied or removed from each side, it is possible to determine the amount of heat passing through the sample. While the method is accurate, it has significant disadvantages, the main of which is the complexity and costs of required equipment. The other characteristic of the “hot box” is that the conditions in the chamber are created artificially. It allows to simulate a variety of conditions at any given time, but its disadvantage is that it may be difficult to simulate real conditions at which the windows usually function, especially the natural convection inside a room.
The main aim of the paper is to determine how adding a layer of different insulating material would affect thermal insulation of the window. To achieve this, the authors’ methodics of analysis of the heat flux passing through the window was used. It uses CFD (computational fluid dynamics) simulation to calculate the heat flux, with measurements used only to validate the accuracy of the simulation. We assumed that if the temperature distribution in the numerical model corresponds to the values from the experiment, the heat flux calculated in the simulation also corresponds to the amount of heat passing through the actual window frame.
Experimental analyses were conducted on a piece of 1m long horizontal beam cut out from the window frame (Fig. 1). The window is a standard construction, made in compliance with the present regulations regarding thermal insulation. It has no additional energy saving features, such as third seal between both parts of the frame.
Figure 1.
Tested window – a) cut out beam, b)drawing of the cross – section
The aim of the experiment was to determine the temperature distribution inside the window frame in real working conditions, which would later be used to validate accuracy of the numerical simulations. For this purpose, the cut-out part of the window profile was placed inside the insulated framework mounted inside the real window opening (Fig. 2). To monitor the temperature, seven T-type thermocouples connected to a data recorder were placed inside and on the surface of the profile, in half of its length (Fig. 3). Two more thermocouples were used to monitor internal and external temperature. We used multichannel temperature logger MMT-302 manufactured by Sensor Electronic, with accuracy of ±0.5°C. Furthermore, air velocity was measured near the window on the outside using handheld thermoanemometer. Because the measurements concerned only the frame, glazing was substituted with a board made of polyisocyanurate (PIR). The same material was used to insulate the lower part of the profile normally connected to the wall and its sides. The sides were additionally insulated with aerogel mat to seal the internal spaces of the profile. The room where the experiment was conducted was unoccupied to minimize changes of the internal temperature. The window was facing in the northwestern direction to avoid direct sunlight.
Figure 2.
Beam mounted in the window for measurements – inside and outside view
Figure 3.
Scheme of the measurement station with the location thermocouples (T1 – T9)
As the measurements took place in real conditions, our ability to control external air parameters was limited. The experiments took place on the days when temperature was low (about 5°C), there was no rainfall and no perceptible wind.
During the measurements, aside from the unmodified profile, three different variants were tested: Profile with additional insulation made of PIR board (Fig. 4a). The stripes of PIR were placed in six outer air chambers of the frame. Profile with additional insulation made of aerogel mat (Fig. 4b). The mat used was Porogel Medium Spaceloft manufactured by Aerogels Poland Nanotechnology. The stripes of insulation were placed in six outer chambers and one inner air chamber of the frame. Because the aerogel mat was elastic, it adjusted its shape to fill the chamber, in contrast to the solid PIR board. Profile with added no load-carrying partitions that divide the larger air chambers into smaller spaces (Fig. 4c). The baffles were placed in five largest air chambers to decrease air circulation. The elements were made from Acrylonitrile Butadiene Styrene (ABS) using a 3D printer.
Figure 4.
Different variants of the tested insulation – photo and drawing of the cross – section: a) PIR board, b) aerogel mat, c) elements made using a 3D printer
Each series of measurements lasted about three hours, until the temperatures were stabilized (Fig. 5). As a result of measurements, we have obtained a set of stabilized temperature values for different thermocouples – those values for different types of insulation are listed in Table 1. However, as each of the measurement series was made for slightly different external and internal temperature, the direct comparison of the results could cause errors. The measured averaged in time external air velocity for all tested cases was close to 0.2 m/s.
Figure 5.
Temperature distribution over time during the measurements for different thermocouples (thermocouples location – Fig. 3; unmodified frame)
The results of the measurements
| Temperature of thermocouple, °C | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Variant | T1 (outside air) | T2 | T3 | T4 | T5 | T6 | T7 | T8 | T9 (inter nal air) |
| 1. unmodified frame | 4.2 | 5.4 | 5.8 | 6.9 | 9.1 | 9.9 | 12.3 | 16.2 | 18.7 |
| 2. PIR insulation | 7.5 | 8.4 | 8.8 | 9.8 | 11.1 | 11.8 | 13.5 | 16.4 | 18.3 |
| 3. aerogel insulation | 5.0 | 5.4 | 5.7 | 7.5 | 9.6 | 10.5 | 12.5 | 16.2 | 17.6 |
| 4. 3D printed elements | 9.5 | 10.2 | 10.6 | 11.2 | 13.1 | 13.8 | 15.2 | 17.1 | 18.4 |
The numerical simulation of window frame was made using Ansys CFX 14.5 software, integrated with Workbench 14.5 platform. Finite volume method is used to solve the energy balance equations for each of the elements in a disretized model.
In the horizontal beam of window frame, both air and heat flows along its length should be negligible, so two-dimensional simulation was chosen. As the CFX software doesn’t allow direct modelling of 2D geometry, geometrical model of the frame was made as a 5 mm thick slice of the window profile (Fig. 6). Aside from the frame, certain amount of the indoor and outdoor space was modeled. On both slice planes, a boundary condition
Figure 6.
Geometrical model of the window frame without additional insulation. Boundary conditions on the interior and exterior wall: green – opening, blue – air inlet, red – air outlet
To simulate conditions inside the room, boundary condition
On the outside, air movement is caused mainly by the wind. The thermoanemometer used in measurements does not measure direction of the airflow. Because the influence of the air direction on the heat flow is minor, for the simplification of the simulation, it was assumed that it is horizontal, parallel to the surface of the profile. At two opposite surfaces of the model boundary conditions
The materials used in the simulation along with their thermal conductivity are listed in Table 2. All the materials except the air were modeled as solid domains. Emissivity of all the surfaces was equal to 0.9 [11]. The inside and outside temperatures for different models are listed in Table 1.
Thermal conductivity of materials used in the numerical model [11, 12, 13]
| Material | Thermal conductivity, W/(m•K) |
|---|---|
| PVC | 0.17 |
| Rubber | 0.24 |
| Steel | 50.0 |
| PIR board | 0.023 |
| Aerogel mat | 0.014 |
The simulations were made in a steady-state, non-isothermal conditions. The high-resolution discretization scheme and the Rhie Chow redistribution algorithm for coupling pressure-velocity were used.
The first simulations were made with no turbulence model, because we assumed that the air movement inside the chambers would be laminar, as suggested in [8]. However, the resulting temperature distribution inside the frame was different from the measurement results, so the Shear Stress Transport (SST) turbulence model was used. Thanks to the combination of k-ε and k-Ω models, it can model the turbulence accurately with both high and low Reynolds numbers [13]. With the use of turbulence model, accuracy of the simulation was greatly improved.
Thermal radiation was modeled using the Discrete Transfer model. It simulates thermal radiation by tracing the rays between the surfaces exchanging heat [13].
The unstructured discretization grid was made using CFX-MESH software as an extrusion of 2D mesh (Fig. 7). The insides of the fluid domains and the solid domains are made of prismatic elements. In fluid domains near the walls five layers of the tetrahedral elements are added to better simulate both the air flow in the vicinity of the wall and the heat transfer between two bodies. The maximum edge length of mesh elements was set as 0.5 mm. Numbers of elements for meshes made for different models are listed in Table 3. A grid independence test was performed for a mesh with smaller elements, but further decrease of the element size did not affect the simulation results in a noticeable way.
Figure 7.
Fragment of the discretization mesh
Discretization mesh parameters for different models
| Model number | Additional insulation type | elements | nodes |
|---|---|---|---|
| 1 | None | 592484 | 762270 |
| 2 | PIR board | 589848 | 760005 |
| 3 | Aerogel mat | 586312 | 757353 |
| 4 | 3D printed elements | 621424 | 782904 |
The calculations were carried out using iteration method with minimum number of 4000 iterations. To control the convergence, monitor points for temperature were set in places corresponding to the locations of thermocouples. For each simulation, after the iteration process, the temperature values in these points were stabilized. Achieved convergence was in the range of 1.0E-8 for momentum and mass, and lower than 1.0E-10 for heat transfer.
As presented in Fig. 8, the numerical model can recreate both the insulating properties of the profile and the air movement in the chambers caused by thermal convection. Measured and calculated values of temperature in each of the points for all the cases are collected in Table 4.
Figure 8.
Temperature and velocity distribution inside the unmodified window frame for experimental conditions
Comparison of measurements and simulations results for the experimental conditions
| Point number | Temperature from simulations, °C | Temperature from measurements, °C | Difference between measurements and simulation, °C |
|---|---|---|---|
| Unmodified profile | |||
| 2 | 5.3 | 5.4 | 0.15 |
| 3 | 5.4 | 5.8 | 0.4 |
| 4 | 8.3 | 6.9 | -1.381 |
| 5 | 10.7 | 9.1 | -1.61 |
| 6 | 11.4 | 9.9 | -1.53 |
| 7 | 12.5 | 12.3 | -0.21 |
| 8 | 15.7 | 16.2 | 0.49 |
| PIR board | |||
| 2 | 7.8 | 8.4 | 0.6 |
| 3 | 7.9 | 8.8 | 0.9 |
| 4 | 11.4 | 9.8 | -1.6 |
| 5 | 12.5 | 11.1 | -1.4 |
| 6 | 13.3 | 11.8 | -1.5 |
| 7 | 14.1 | 13.5 | -0.6 |
| 8 | 16.9 | 16.4 | -0.5 |
| Aerogel mat | |||
| 2 | 5.4 | 5.4 | 0 |
| 3 | 5.4 | 5.7 | 0.3 |
| 4 | 10.0 | 7.5 | -2.5 |
| 5 | 11.3 | 9.6 | -1.7 |
| 6 | 12.2 | 10.2 | -2 |
| 7 | 13.1 | 12.5 | -0.6 |
| 8 | 16.2 | 16.2 | 0 |
| 3D printedelements | |||
| 2 | 9.8 | 10.2 | 0.4 |
| 3 | 9.9 | 10.6 | 0.7 |
| 4 | 12.7 | 11.2 | -1.5 |
| 5 | 13.8 | 13.1 | -0.7 |
| 6 | 14.3 | 13.8 | -0.5 |
| 7 | 15.0 | 15.2 | 0.2 |
| 8 | 17.3 | 17.1 | -0.2 |
To compare thermal transmittance of all the window profile types, simulations for the numerical models were carried out with unified external and internal air temperatures. The external temperature was equal to -20°C [14], and internal temperature was 20°C for all the cases.
Knowing the flux of heat passing from the internal environment to the window frame in the simulations, Linear thermal transmittance of the whole window profile was calculated (Table 5). The temperature distribution in the cross-sections of all the profiles is presented in Fig. 9.
Figure 9.
Temperature distribution inside all the tested profiles for uniform inside and outside temperatures (20. -20°C); a)unmodified profile; b) PIR board; c) aerogel mat; d) 3D printed elements
Heat flux and thermal transmittance of the profile for all the tested variants
| Additional insulation type | Heat flux per unit profile lenght with temperature difference 40K, W/m | Linear (per meter length of profile) thermal transmittance, W/(m•K) | Heat flux difference compared to the unmodified frame |
|---|---|---|---|
| none | 5.405 | 0.1351 | 0.00% |
| PIR board | 5.034 | 0.1258 | -6.85% |
| aerogel mat | 4.877 | 0.1219 | -9.76% |
| 3D printed elements | 5.173 | 0.1293 | -4.29% |
The comparison between simulations and measurements results (Table 5) shows that the temperatures on the surface of the profile (point 2, 3, 8) are modeled very accurately – the differences between measured and calculated temperatures are in most cases smaller than 0.5°C, which is the value of accuracy of the used thermometer. In case of the temperatures inside the profile, this difference is slightly greater – about 1.5°C, and it is smaller for points closer to the interior of the building (points 5, 7). This may suggest that the heat transfer through the metal parts is modeled with slight inaccuracy. The simulations with different kinds of steel varying in thermal conductivity were tested, but it did not affect the simulation results in a significant way. However, as the differences in temperature between the numerical models and the measurements results were in most cases very small, we can assume that the results of the simulations accurately represent temperature distribution inside the window frame. Moreover, the values of errors were similar in corresponding points for all the tested cases. It is possible to assume that the errors in calculated heat flux will also be similar for different models, decreasing the significance of inaccuracy during comparison between them.
As our experiment shows, use of the proposed method to determine the heat flux passing through the window profile is much simpler and uses much less equipment than the hotbox method. However, it has some disadvantages, the most prominent of which is strong dependency on natural external conditions. This limits the time of measurements to cold seasons and to days with acceptable weather. Also, even though there was little difference between the measured and simulated values of temperature, it is difficult to determine the accuracy of heat flux calculation. It would be useful to compare those results with the measurements made using hotbox method, which should be the aim of the further studies.
Results of the simulations made for unified conditions (Table 5) show that the heat flux passing through the profile strongly depends on thermal conductivity of the additional insulation layer. The insulating properties improved greatly for aerogel mat (about 10%), which suggests that it is the best insulating material. Its additional advantage is flexibility, which makes fitting it inside a profile easy. Its price, however, is very high compared to other materials, which may disqualify this material from commercial use.
In case of 3D printed elements, heat flux drop compared to the unmodified profile was low (about 4.3%). This may be caused by fact that the window profile is already divided into many closed spaces, probably in a way optimized by the manufacturer to achieve maximum thermal insulation.
Considering the cost of the material and the insulation properties, PIR board seems to be the best of all the tested insulation types. Even though the heat flux drop (about 7%) was lower than in case of an aerogel mat, the material was very cheap compared to the other types of insulation. This would allow to fill even more air chambers with insulation without significant increase of the product price.