SBO project IWT Derde jaarlijks rapport

Transcriptie

1 Faculteit Toegepaste Wetenschappen Departement Burgerlijke Bouwkunde Laboratorium Bouwfysica Kasteelpark Arenberg 40, 3001 Heverlee Universiteit Gent Vakgroep Mechanica van Stroming, Warmte en Verbranding Sint-Pietersnieuwstraat 41, 9000 Gent Universiteit Gent Vakgroep Architectuur en Stedenbouw Plateaustraat 22, 9000 Gent Technische Universiteit Eindhoven Unit Building Physics & Systems P.O. Box 513, 5600 Eindhoven Wetenschappelijk en Technisch Centrum voor het Bouwbedrijf Departement Geotechniek en Structuren Departement Bouwfysica en Uitrustingen Poincarélaan 79, 1060 Brussel Physibel Heirweg 21, 9990 Maldegem Daidalos Bouwfysisch Ingenieursbureau Oudebaan 391, 3000 Leuven Ingenieursbureau Stockman nv Muinklaan 6, 9000 Gent Heat, air and moisture performance engineering A whole building approach SBO project IWT Derde jaarlijks rapport Staf ROELS, Erik DICK, Michel DE PAEPE, Arnold JANSSENS, Peter WOUTERS, Benoit PARMENTIER, Xavier LONCOUR, Filip VAN RICKSTAL, Gilles FLAMANT, Luk VANDAELE, Jan HENSEN, Bert BLOCKEN, Piet HOUTHUYS, Piet STANDAERT, Filip DESCAMPS, Piet DELAGAYE, Demir-Ali KÖSE, Kim GOETHALS, Marnix VAN BELLEGHEM, Mohammad MIRSADEGHI, Daniel COSTOLA, Tadiwos ZERIHUN DESTA, Thijs DEFRAEYE (verslag). September 2009

2 Inhoud Inhoud Wetenschappelijk- technisch verslag Overzicht van uitgevoerde activiteiten... 3 WP1.1 Wind pressure distribution... 4 WP1.2 Driving rain load distribution WP2.1 Development of HAM model WP2.2 Experimental analysis on building enclosures WP3.1 Convective heat exchange and summer comfort WP3.2 Development of CFD-HAM model WP4 Towards an integrated approach WP5.1 Strategic and integrated planning of research activities WP5.2 Strategic implementation of testing and simulation facilities References Bijsturingen in het project Beheer van het project Haalbaarheid van het project Te beschermen resultaten Utilisatieverslag Valorisatiepotentieel: geactualiseerde visie Overzicht van de uitgevoerde valorisatieacties Bescherming projectresultaten Financieel verslag Prestatietabel Prognose voor komende projectjaar Financiële verantwoording... 61

3 1 Wetenschappelijk- technisch verslag 1.1 Overzicht van uitgevoerde activiteiten Het onderzoekswerk is georganiseerd in vijf werkpakketten, welke op hun beurt nog verder onderverdeeld zijn: WP1 Outside boundary conditions WP1.1 Wind pressure distribution WP1.2 Driving rain load distribution WP2 Building envelope WP2.1 Development of HAM model WP2.2 Experimental analysis on building enclosures WP3 Building interior WP3.1 Convective heat exchange and summer comfort WP3.2 Development of CFD-HAM model WP4 Towards an integrated approach WP4.1 Development of a prototype software environment WP4.2 Development of a coupling necessity decision procedure WP4.3 Experimental validation WP5 Establishment of a knowledge platform WP5.1 Strategic and integrated planning of research activities WP5.2 Strategic implementation of testing and simulation facilities Dit verslag beschrijft de onderzoeksactiviteiten van het derde projectjaar lopende van 1 september 2008 tot 1 september Dit deel van het verslag is opgemaakt in het Engels omdat dit de voertaal is voor publicaties en wetenschappelijke rapporten en een aantal onderzoekers Nederlands niet als moedertaal hebben. Derde jaarlijks rapport 3

4 WP1.1 Wind pressure distribution Objectives WP1.1 emphasises the distribution of pressure differences across the building envelope due to wind flow. This subtask is subdivided in two parts: WP1.1.1 Some sets of full-scale measurements will be recorded regarding the wind pressure distribution on the building envelope of the building WINDHouse of the Laboratory Structures of BBRI, located in Limelette. The data will be used to asses the numerical predictions with CFD (Computational Fluid Dynamics). WP1.1.2 Numerical simulations of the air flow around a building will be performed. Different hybrid RANS-LES models to describe the turbulence are tested. The calculations will provide, among other numerical results, the pressure distributions at various locations on the building surface. The obtained data will be compared with the experimental pressure data sets provided by WP Description of work WP1.1.1 Experimental analysis of wind pressure distribution over a building envelope A reliable prediction of the wind pressure around a building is a necessity for a correct simulation of the Heat, Air and Moisture performance of that building. The wind pressures can be calculated with numerical models, they can be determined using scaled models in a wind tunnel or they can be measured on a real scale building. This section describes the measurements that were carried out on the Windhouse, situated at the BBRI-site in Limelette. It first describes how the measurements were organised and then discusses the results. The test configuration in Limelette aims to measure the pressure values around the real scale test house that are caused by wind. Figure 1 shows the Windhouse building (Parmentier 2003). Figure 1. The rotary wind house situated on the research site in Limelette. The locations of the measuring points were determined together with the colleagues from the department of Flow, Heat and Combustion Mechanics from UGent since they will use the results of the experiments to validate their numerical model. The roof contains 6 measuring points, the facades contains 24 points. The measuring points are somewhat concentrated at the front (Figure 2). Because of the ability to rotate the building, this does not limit the amount of Derde jaarlijks rapport 4

5 information that can be read from the tests. In addition to these external pressure taps, the internal pressure in a nearly air tight box inside the house was measured. Figure 2. The locations of the measuring points in the facade and the roof. Because pressure values are low when relatively low wind speeds are measured, some uncertainty is expected on the absolute level of the pressures. In order to overcome this problem, the Cp-values should be calculated in a different way. The method used is based on the analysis of the variations of the pressure and wind velocities. A collection of measurements consists of the data obtained from the peripherals: the wind properties and the position of the house and the wind pressure values. Figure 3 shows the pressure history measured at points 19 and 22. The average wind velocity is 5.8 m/s, but the peak velocity is nearly 50% higher. Variations in wind velocity are common. More striking are the variations in the wind direction. The house is rotated in a way that the desired angle of attack of the wind is realised compared to the (average) wind direction measured during the last standby period. However, since the wind direction is seriously varying during the record, the average angle of attack might differ from the aimed one. On a series of 20 runs, the deviation of the angle of attack is as high as 9. This deviation brings about significant divergence from expected theoretical values. Obviously, the real angle of attack is used for the analysis (and not the supposed one). As stated before, the analysis is based on the variations of the measured pressure values rather than on the absolute pressure. A run contains the wind properties and the pressures caused by the wind load on the house. A measuring run lasts 10 minutes. For each angle of attack, a number of runs is carried out. They are analysed one by one and then the individual results are averaged. The runs are realised on different days and are related to different wind directions. The Cp-value for a point is given by: P Cp = 2 v 2 ρ Derde jaarlijks rapport 5

6 Figure 3. Wind pressure history for a 10 minute run. This equation is valid for a constant wind velocity. Since the velocity varies during the 10 minutes lasting measurements, one should use an average value. When calculated in the classical way, the average wind velocity would be the average of the mean wind velocity measured by the three anemometers in the middle of the measuring mast. Then, the square of this average velocity is used. The pressure value P that is used is the average of the pressure for a certain point. The density of the air is considered to be 1.2 kg/m 3. All the runs for the same angle of attack are averaged to obtain the final result. One should realise that this is not completely correct. One should use the average of the square velocity-values instead of the square of the average velocity. Although the result differs only slightly in case the wind velocity variations are small. Further, one should remember that the results are analysed using the variation of the pressure and wind velocity rather than the absolute values to calculate Cp. Figure 4. Cp-values corresponding to an angle of attack of 0, based on pressure and wind velocity variations. When the wind is blowing perpendicular to the front, there is a 0 angle of attack. The wind causes overpressure on the front and thus positive Cp-values. On all other facades negative Cp-values will be present. In total, 38 runs are carried out. Results are shown in Figure 4. The spread on the results is rather big. This can be explained by the variation on the angle of attack. The most critical points, those located near the front corners, indeed show the biggest variation. When calculated in the classical way, their sign is continuously changing, Derde jaarlijks rapport 6

7 confirming the fact that overpressure is alternating with underpressure. This hinders the comparison of the experiments with the numerical simulations. To get an idea of what might be expected, a simulation with a small angle of attack, say 10, could be interesting. When the wind is blowing parallel to the front, perpendicular to the façade containing the steps, there is a 90 angle of attack. The measuring points are asymmetrically spread in this case since they are concentrated at one side of the house. Figure 5 provides the results for the 90 angle of attack. Finally, other measurements (not presented here) were achieved for 45 and 180. Figure 5. Cp-values corresponding to an angle of attack of 90, based on pressure and wind velocity variations. WP1.1.2 Development of a hybrid RANS-LES technique of flow over buildings Motivation of the subtask From a practical point of view it is not feasible to perform high quality Large Eddy Simulations (LES) of flows around buildings (LES = resolving the large turbulence structures in the flow and modelling the small structures). Since anyhow in the vicinity of walls turbulent structures are small, an extreme fine mesh near building walls is required for sufficient resolution. This leads to very large computational costs. On the other hand, simulations with Reynolds Averaged Navier-Stokes equations generally result in very poor predictions of the flow field over a building (RANS = modelling all the turbulence structures). This is due to the large-scale unsteadiness of the flow. A possible solution is a hybrid RANS/LES model. In such a model, an unsteady Reynolds Averaged Navier-Stokes (RANS) model is used in the near-wall regions, while far from walls a sub-grid scale (SGS) model is used within a LES-formulation. The motivation is the belief that turbulence in wall vicinity is reasonably universal, so that RANS description is valid, while far from walls it is not, so that LES description is more appropriate. The hope is that by using RANS in wall vicinity, the grid can be much coarser than necessary for LES, so that sufficient quality of the flow prediction can be obtained at an acceptable computational cost. Obtained results During the first and second project years, many hybrid models have been investigated, including the more-or-less standard Detached Eddy Simulation (DES) model of Spalart, the DES model based on the SST k-ω turbulence model (DES-SST), the k-l model of Davidson, the k-l model of Tucker and the hybrid model developed by our own research group (C. De Derde jaarlijks rapport 7

8 Langhe): the ε-l model. The test case was a wind-tunnel experiment of the flow over a cube on a flat plate. The Reynolds number based on the cube size is This is about a factor 40 lower than for atmospheric boundary layer flow, but it is generally accepted in the community that for this difference in Reynolds number, flow patterns are not much different. The conclusion was that, based on the prediction of velocity profiles, the DES-SST model results are in best agreement with the experimental data, especially in the vicinity of the cube. Therefore, we recommended this hybrid model at the end of year two for the prediction of the pressure distribution on a building. We identified test data for flow over a cubic building of 6 m x 6m x 6 m in real atmospheric conditions. The Reynolds number is 4x10 6. The pressure distribution is available on two cuts of the cube: one with the vertical symmetry plane in flow direction and one with a horizontal plane. The data were used in year three for further validation of RANS-models, hybrid RANS/LES models and LES models, with the aim to come to a final conclusion on the use of hybrid methods for analysis of flows over buildings. We used two grids. Mesh A is a typical RANS-grid with clustering of nodes in wall vicinity, with approximately 8x10 5 cells. Mesh B has a little more than 1.2 million cells and is a grid particularly adapted for LES. It has a coarse background grid, which is refined twice in cube vicinity. Both grids are rather coarse, with a size typical for RANS calculations, with the objective to keep computational costs low. At the inlet of the domain a logarithmic velocity profile, derived from the experimental profile, is assumed. We tested RANS, hybrid RANS/LES and LES. The grids are much too coarse for a genuine LES. Therefore, we introduced a wall model to derive the shear stress at walls. We used the best quality wall model of FLUENT (enhanced wall treatment) and an own developed wall model. The results of the hybrid RANS/LES simulations are poor and almost equal to the results of the RANS simulations. The main reason is that the grids are by far not fine enough for reliable simulations with typical hybrid models. These models activate LES when the grid size is small enough to resolve a significant part of the turbulent structures. On very coarse grids, this only happens in limited areas. In our simulations, LES is only active in a very small part in the wake of the building. This is not enough to benefit from the LES features of the model. So, we suffer here a drawback of the much higher Reynolds number with respect to the earlier wind tunnel tests. The grids have about the same size in both applications, but due to the much higher Reynolds number, in real atmospheric conditions, the scale range of the turbulence structures is much larger, so that with the used grid size we only resolve but the very largest turbulence structures. As a consequence, the hybrid algorithm decides to stay in RANS mode almost everywhere. Much better results are obtained by forcing LES to be active everywhere. Principally, this is not justified due to the coarseness of the grid. So, it means that we perform underresolved LES. Due to the coarse grid near walls, a wall model is necessary to determine the wall shear stress. Overall, the results with such a crude LES method are quite good. Summary of the work in year 3 RANS-simulations. We tested the standard k ε and the k ω SST models. As expected, the RANS results are poor since the flow is highly unsteady but it is represented steady with the RANS models. Derde jaarlijks rapport 8

9 Hybrid RANS/LES-simulations. In this case we only tested the DES SST model. Based on the results of the second project year, we expected that this hybrid model would perform the best. However, we observe that there is hardly any difference between the hybrid model and the k ω SST predictions. Plotting the relative DES length scale reveals that the LES region is very small and located in the wake of the building. This means that there is very low LES activity and the simulation is RANS in a large vicinity of the building. This explains the resemblance of the RANS and hybrid results. LES-simulations. Here we tested the standard Smagorinsky model with C s = 0.1. We also ran calculations without a SGS model by putting C s equal to 0. Furthermore, we derived a wall model (WM) based on the log law and used this instead of the no slip boundary condition. The WM is programmed with user defined functions (UDF s) and is used on both grids. Remark that when a no-slip boundary condition is activated in FLUENT, actually on coarse grids also a wall model is used (enhanced wall treatment). The LES simulations are far much better than the hybrid predictions. There is very good agreement with experimental profiles at the front, on the top and at the back of the building. The predictions on the side walls follow the experimental profiles, but there is some deviation. There is also very little influence of the wall model. With our own wall model and the wall model of FLUENT, the results are almost the same on mesh A. On mesh B we observe some differences in particular at the top and side walls of the building. We observe that the overall quality of the predictions is not really better on mesh B than on mesh A. This means that both meshes are much too coarse to resolve enough in wall vicinity. The differences in the results are due to the different mesh size in wall vicinity, but there is no conclusion on the best gridding strategy. The observation is that any LES formulation gives reasonably good results, taking into account the very coarse grids used. Practical conclusion Based on the results of the wind-tunnel experiment, our expectation was that the DES-SST model would again perform the best for the cubical building in the atmospheric boundary layer. This is not what we observe. The results of the hybrid RANS/LES simulations are poor and almost equal to the results of the steady state RANS simulations. The main reason for the poor predictions is that the grids which we use are not fine enough for high quality hybrid simulations. Plotting the relative DES length scale reveals that the LES region is very small and is located in the wake of the building. This means that there is very low LES activity. One could make use of much finer grids. This would lead to better predictions but at a very large computational cost. This is what we try to avoid here. With LES, we observe only little sensitivity to the quality of the predicted pressure distribution from grid resolution in wall vicinity and from the wall stress calculation method. For practical prediction of the pressure distribution on real size buildings, we therefore now have to recommend LES with a wall model. Derde jaarlijks rapport 9

10 Examples A view of mesh A and mesh B is given in Figure 6. In Figure 7 to Figure 9 we present the prediction of the pressure distributions on both sections and mesh A. Finally, the results for the LES and ILES simulations on mesh B are shown in Figure 10. Figure 6. Mesh A (left) and close up of mesh B (right). Figure 7. Pressure coefficient profiles on the vertical (left) and horizontal (right) sections, RANS results on mesh A. Black squares are experiments. Figure 8. Pressure coefficient profiles on the vertical (left) and horizontal (right) sections, HYBRID results on mesh A. Black squares are experiments. Derde jaarlijks rapport 10

11 Figure 9. Pressure coefficient profiles on the vertical (left) and horizontal (right) sections, LES and ILES results on mesh A. Black squares are experiments. Figure 10. Pressure coefficient profiles on the vertical (left) and horizontal (right) sections, LES and ILES results on mesh B. Black squares are experiments. Deliverables WP1.1.1 The test setup for the Windhouse has been used to obtain various data sets, which can be used for validation purposes in WP WP1.1.2 Several RANS, hybrid and LES techniques were evaluated for a full-scale building in atmospheric conditions. Attention was given to wall modelling and grid quality. LES together with a wall model is recommended for surface pressure and flow calculations around buildings. Planning WP1.1.1 Further processing of the measured data and transfer of the results to WP In a technical meeting the agreement of the simulations of WP and the measurements of WP will be discussed. WP1.1.2 In the literature, experimental pressure data on a 13.7 m x 9.1 m x 4 m building with a roof (the Texas Tech experiment) have been found. Measurements have been Derde jaarlijks rapport 11

12 performed at angles of attack of 90º and 60º. We will use LES and ILES and compare the results with the available data. At the BBRI, a test building has been built with pressure taps for pressure measurements. We will simulate the flow around this test building and validate it with the experimental data. Derde jaarlijks rapport 12

13 WP1.2 Driving rain load distribution Objectives WP Laboratory and in situ experiments of contact phenomena of driving rain impinging on different building materials. The final amount of water that may enter the building enclosure (the boundary conditions for the building envelope model) will be determined by splashing effects, adhesion, evaporation, run-off and capillary absorption. All these phenomena strongly depend on the material properties of the building enclosure. Parameters are porosity of the material, moisture capacity, capillary activity,. WP Development of a water uptake and run-off model for building materials. Numerically, the transformation of individual rain drops, absorbed by the building material towards a smeared wetting as assumed in building envelope models, will be investigated. In addition, a run-off model for rain water run-off on capillary active materials will be developed. The research of the previous two sections of this subtask (WP and WP1.2.2) was conducted in strong collaboration with a PhD student working on a KUL OT-project where the contact phenomena (bouncing, splashing and spreading) and surface phenomena (absorption, evaporation and run-off) are also of interest. Out of this research it has become clear that the outside boundary conditions, namely convection and radiation, have a large impact on the evaporation at building facades wetted by wind-driven rain. Therefore the focus of the PhD student that is involved in this SBO project (PhD1) will be mainly on these topics and an additional subtask (WP1.2.3) has been introduced below. WP Numerical and experimental analysis of convective heat and mass transfer at exterior building surfaces. More accurate predictions of the convective transfer coefficients are obtained, considering the influence of wind speed, wind direction and building surroundings and spatial distribution across the surface. Also the influence of specific materials and surface texture is investigated. Description of work WP Laboratory and in situ experiments of contact phenomena of driving rain impinging on different building materials. The objectives of this subpackage are achieved. No further actions are planned for the coming year. WP Development of a water uptake and run-off model for building materials. The objectives of this subpackage are achieved. No further actions are planned for the coming year. WP1.2.3 Laboratory experiments and numerical modelling of the influence of outside boundary conditions on the heat and mass transfer at building facades wetted by wind-driven rain. As mentioned before, the effect of the outside boundary conditions, such as convection and radiation, was found to be of significant importance in the assessment of the response of building walls after a rain event for simulations such as described in WP Therefore, WP1.2.3 was created. In a first stadium, only convective heat and moisture transfer was considered and only forced convection was taken into account. Derde jaarlijks rapport 13

14 The numerical study (with CFD) on convective heat transfer of an isolated cubic building in an atmospheric boundary layer was continued. The focus was mainly on boundary-layer modelling issues. Two different techniques were considered: (1) low-reynolds number modelling (LRNM), which was found to give accurate CHTC predictions but which required a large computational grid in the boundary-layer region; and (2) wall functions (WF), which are less accurate in predicting heat transfer in the boundary layer but which allow for much coarser grids and therefore these WF are commonly used in building aerodynamics. Two optimisation approaches are proposed to allow for accurate convective heat transfer predictions with CFD (RANS) without increasing the computational cost significantly. The first approach considers evaluating convective heat transfer at lower wind speeds, based on the concept of flow similarity by Reynolds number independence. Afterwards the obtained data at these low wind speeds are successfully extrapolated to higher wind speeds. The advantage of using low wind speeds in the simulations is that it leads to a reduction of the computational grid resolution in the boundary-layer region. A significant reduction of the wind speed, used in the simulations, was found to be possible, up to a factor of about 5, depending on the conditions. The second approach proposed the use of customised thermal wall functions (CWF), derived from LRNM simulations. These wall functions showed a significant improvement of heat transfer predictions compared to standard wall functions (SWF) but still allowed for coarse grids in the boundary-layer region (see Figure 11). Note that two journal papers on this work will be submitted soon to international journals. CHTC [W/m²K] CHTC [W/m²K] (a) CHTC - LRNM CHTC - SWF CHTC - CWF y* - CWF x/h [-] CHTC - LRNM CHTC - SWF CHTC - CWF y* - CWF (b) x/h [-] y* [-] y* [-] x x Wind Wind Figure 11. Convective heat transfer coefficient (CHTC) distribution on the surfaces of a 10 m high cubic building (for LRNM, SWF and CWF) and distribution of y* in the wall-adjacent cell (for CWF) in a vertical (a) and horizontal (b) centreplane for a reference wind speed of 0.5 m/s at a height of 10 m. In the previous simulations, heat transfer is entirely solved within the CFD package. Solving moisture transfer in porous materials (which involves both liquid and vapour transport) is not possible with the CFD software that is used. Therefore the program is coupled (explicitly) with HAMFEM, which is an in-house HAM modelling tool developed by the Laboratory of Derde jaarlijks rapport 14

15 Building Physics. The air flow is entirely solved within the CFD package whereas the heat and moisture transport in the porous material is solved within HAMFEM. This coupled program (CHAM) is used to evaluate coupled convective heat and moisture transfer for flow parallel to a porous building material. The influence of the air flow on the heat and moisture flows in the porous material is found to be limited if the material is not very wet. This is due to the resistance of the material itself to moisture transport, which is usually much higher than that of the boundary layer. For wet materials however (e.g. after a rain event), a significant difference in drying rate is found with the conventional approach, with spatial and temporal varying transfer coefficients over the surface. This study shows that the use of constant (spatial and temporal) transfer coefficients can lead to a significant simplification of the drying behaviour of porous materials. As a realistic case study, the drying of a building wall, in an atmospheric boundary layer and wetted by wind-driven rain, was also modelled with the coupled model (Figure 12 and Figure 13). (a) y Interface x 7H U 10 = 2.5 m/s Symmetry (slip wall) T e = 10 C RH e = 60% Building walls (AD&IP&NS) 9.8 m 10 m (H) 20H Ground (AD&IP&NS) 20H Zero static pressure (b) Interface Outer wall 9 cm XPS (not included in model) Inner wall (not included in model) Interior boundaries model (AD&IP) Figure 12. Model for numerical analysis with boundary conditions: (a) Environment modelled with CFD, (b) Wall composition (AD = adiabatic, IP = impermeable for moisture, NS = no-slip wall with zero roughness) (Fig. not to scale). 5.E-05 4.E-05 CHAM - y=9.8m CHAM - y=7.5m CHAM - y=5m CHAM - y=1m CHAM - Average HAM (b) qc,m,w (kg/sm²) 3.E-05 2.E-05 1.E-05 0.E+00 (a) Time (h) Figure 13. Comparison between HAM and CHAM at different positions on the exterior wall surface: (a) Drying rate, (b) RH (%) distribution (with CHAM) over exterior wall surface as a function of time and location on surface. Derde jaarlijks rapport 15

16 Deliverables WP1.2.1 The objectives are achieved WP1.2.2 The objectives are achieved WP1.2.3 Two optimisation approaches are proposed and evaluated to increase the computational economy for CFD simulations of convective heat transfer calculations at high Reynolds numbers. The coupled CFD-HAM model was used to calculate drying of porous materials for different configurations. Planning WP1.2.1 No further actions are planned for the coming year. WP1.2.2 No further actions are planned for the coming year. WP1.2.3 More advanced modelling techniques (e.g. LES) will be evaluated for the isolated building. The coupled CFD-HAM model will be further verified and validated and applied to various configurations. Derde jaarlijks rapport 16

17 WP2.1 Development of HAM model Objectives The aim of this subtask is to develop a comprehensive building envelope model for the coupled analysis of heat, air and moisture transport through building enclosures. The model will be based on existing scientific models available from the partners. Because of the different time-scales between heat, moisture and air transport, and the associated numerical obstacles, a stabilised solution method could not be achieved when dealing with air transport. Therefore the main objective of this subtask is to tackle this problem. Description of work The heat and moisture transfer phenomena through the VLIET test wall in the NE-façade of the row house are modeled using Delphine software. Boundary conditions are obtained from the experiments. The moisture content of the Celit3D board (the hygroscopic buffering external wind barrier) has been measured on a regular basis on nine removable specimens (see Figure 14). The evolution of the moisture content with time is simulated and compared with the measured data. Figure 14. The removable Celit3D samples configuration and cross-sectional view of the test wall (right, central and left from the room inside view). The HAM-response of the walls is simulated for the period from December 2007 until July 2009 with time stepping of one hour. The model is validated in terms of temperature and humidity prediction at several sensor locations. The obtained results show a good prediction of the flow system. Figure 15 compares the predicted moisture content of the Celit3D specimen with the measured data. As can be seen, the simulation underestimates the moisture content of the Celit3D material in the winter season because liquid water transport, a phenomenon occurring due to interstitial condensation, is not yet included in the model. Derde jaarlijks rapport 17

18 Figure 15. Comparison of simulated and measured moisture content at the left-middle section of the wall. Deliverables The modeling of the HAM-transport is ongoing. The first results are promising, but so far air transport is not the dominant factor. Planning Inclusion of air transport in the model. Derde jaarlijks rapport 18

19 WP2.2 Experimental analysis on building enclosures Objectives This subtask focuses on the analysis of HAM transport through building enclosures based on measurement data. It was decided to limit the experimental analysis of this workpackage to the VLIET building of KULeuven. The experiments are conducted in a wall of the row house constructed in the previous years (see the report of 2007). The obtained data can be used to get physical insight about the HAM transport phenomena and as a validation tool for the numerical model to be developed under WP2.1. Description of work To study HAM-transport through building components a test set-up has been constructed at the VLIET building, as mentioned in the previous annual reports. The test set-up consists of a terraced house with in the north-east façade a light weight wooden wall structure partitioned in three parts, each with a similar set-up but different air tightness. In the room, a forced ventilation system is operational, since March The measurement data are not only used for studying the HAM-flow phenomena through the wall but also to validate a numerical model. In this subsection the experimental results of the VLIET test wall will be analysed and interpreted. Figure 16 shows the evolution of the moisture content of the nine removable moisture capturing Celit3D specimen with time. From March 2009, a forced ventilation system is installed in order to study the effect of air on the prevailing heat, air and moisture transport through the wall (see Figure 17). Figure 16. Moisture content evolution of Celit3D specimens from June 2008 until July To quantify the air tightness of the left, central and right parts of the wall, a tracer gas experiment was conducted. While Sulphur Hexafluoride, a tracer gas, was injected in the room, the injection rate and the gas concentration in the room were recorded. Simultaneously, the concentrations in the cavities, just in front of the three partition of the test wall are logged (Figure 18). Derde jaarlijks rapport 19

20 Figure 17. a) Ventilation system b) Room pressure build-up c) Incoming air flow rate. Based on the mass balance equation of the tracer gas, the air permeability of the left, central and right parts of the wall are deduced as , and m 3 /m 2 /h/pa respectively. The conclusion of the results was that all three parts of the wall are at the moment too air tight and air flow effects do not prevail. Therefore at the start of autumn 2009 a grid of small holes will be drilled in the Celit3D to create a diffuse air-open wind barrier. Figure 18. Tracer gas concentration profile in the room and in front of the three parts of the wall (in the cavity). Deliverables Temperature and humidity experimental data are available since December Moisture evolution data of the celit3d material are collected since June Tracer gas and ventilation data are available. Data processing tools are developed. Some material property data of Celit3D are available. An internal report that describes experimental analysis and result interpretation is available. Planning The wooden finishing at the left part of the wall will be drilled in order to study the effect of air flow through it. Measurement of material properties will be conducted. Derde jaarlijks rapport 20

21 WP3.1 Convective heat exchange and summer comfort Objectives WP3.1 will focus on the convective heat transfer in rooms and offices and its impact on summer comfort performances by solar processing, intensive (night) ventilation and thermally active flooring. Engineering the summer comfort performances depends on the other hand to a large extent on a reliable prediction of the local heating through solar irradiation and the prediction of wind pressures around buildings for ventilation (WP1.1). WP3.1 is subdivided into four parts: WP3.1.1 An experimental analysis of convective heat transfer between the ventilation air and the building surface will be performed. This work focuses on a single zone and will be done in the rotating PASLINK-cell available at BBRI, a highly instrumented test facility. WP3.1.2 An experimental analysis of solar heating and intensive ventilation and its impact on summer comfort will be conducted. WP3.1.3 A numerical prediction of ventilation and local heating through solar irradiation inside buildings will be performed. The solar illuminated patterns on inside surfaces through direct solar radiation will be calculated and converted to local heat powers. For the ventilation, input data from WP1.1 will serve as boundary conditions. The CFD simulations will be coupled with a multi-zone building simulation model and the predictions will be validated by comparison with the test results. WP3.1.4 The numerical methods that are developed will be validated with data measured in projects realised by the industrial partners. Description of work WP3.1.1 Experimental analysis of convective heat transfer between the ventilation air and the building surface High computational costs and the need of considerable empiricism limit the use of computational fluid dynamics (CFD) in building design, in favour of multi-zone energy simulation (ES). Yet, a key parameter of building performance analysis is the prediction of interior convective heat transfer as shown in Goethals and Janssens (2008). This study showed that the choice of the convection algorithm is of the same importance as the choice of the design parameters, such as internal heat gains or sun blind control. However, this last finding is predetermined by, amongst others, the modelling of the incoming direct solar radiation, since most of the convection algorithms depend, partially, on the temperature difference between the air and the surface. Therefore, the above-mentioned study is extended to assess the sensitivity of the predicted convective heat transfer, and thus building performance, by ES to the modelling of the distribution of solar radiation described in Goethals and Janssens (2009a). Simulations of summer comfort in a night-cooled office room in Belgium are carried out in TRNSYS using three convection algorithms and four methods to model the distribution of direct solar radiation. The influence is evaluated for the summer comfort - weighted temperature excess method (GTO) and adaptive temperature limits indicator (ATG) - and the cooling demand with and without night ventilation. To model the distribution of direct radiation, TRNSYS uses, by default, absorptance-weighted area ratios. Conversely, constant fractions of the total entering solar radiation that strikes the surfaces can be defined. Finally, the distribution can also be calculated, based on the position of the sun, Derde jaarlijks rapport 21

22 using uniform or discretized surfaces. In case of discretized surfaces, the movement of the sun patches along the surfaces is modelled in a more detailed way. As shown in Table 1, Table 2, Table 3, the influence of the modelling of the distribution of solar radiation on the predicted thermal comfort and energy demand is inferior to the choice of the convection correlations. However, the relative importance of advanced modelling of the convective heat transfer and of the incoming direct solar radiation is case-specific. Table 1. Results of building simulation with CHTC correlations of pren ISO Table 2. Results of building simulation with CHTC correlations of Awbi and Hatton-natural. Table 3. Results of building simulation with CHTC correlations of Beausoleil-Morrison. cooling demand Incoming solar ATG (h) GTO (h) (kwh/m².year) radiation - -C -B A +B +C - AC AC+NV Diffuse radiation Constant (85%) (100%) 2.41 (100%) Time-dependentuniform (85%) (101%) 2.48 (103%) Time-dependentdiscretized (93%) (104%) 2.46 (103%) cooling demand Incoming solar ATG (h) GTO (h) (kwh/m².year) radiation - -C -B A +B +C - AC AC+NV Diffuse radiation Constant (101%) (100%) 5.29 (100%) Time-dependentuniform (101%) (101%) 5.23 (99%) Time-dependentdiscretized (109%) (102%) 5.46 (104%) cooling demand Incoming solar ATG (h) GTO (h) (kwh/m².year) radiation - -C -B A +B +C - AC AC+NV Diffuse radiation Constant (102%) (101%) 4.97 (100%) Time-dependentuniform (100%) (101%) 4.99 (101%) Time-dependentdiscretized (91%) (104%) 5.17 (105%) To elaborate on the impact of the modelling of incoming direct radiation, simulations of summer comfort in an office room in the Unilin Flooring-Quickstep building are performed using TRNSYS and VOLTRA. Using the measurement data obtained in the summer of 2009, both simulation models are validated. Given the accuracy of the temperature sensors, the predicted operative temperature corresponds considerably well with the measurement data - as shown in Figure 19. Besides, since the time-invariant convective heat transfer coefficients mentioned in pren ISO 13791, are used, the importance of the modelling of convective heat transfer should be mitigated - for this specific case at least. Derde jaarlijks rapport 22

23 Deviation [ C] Occurrence [%] ± ± ± ± ± ± ± ± ± ± Figure 19. Comparison of measured and predicted operative temperature in the small office. Hereafter, the impact of the approximation methods used in TRNSYS to model direct solar radiation is compared with the results obtained with VOLTRA, which is equipped with a solar processor. The best agreement is found for the time-dependent distribution of the solar radiation. Especially, the total absorbed solar radiation predicted by the simulation model with discretized surfaces comes close to the VOLTRA results - as shown in Figure 20. Further, a sensitivity analysis, similar to the above-mentioned study, is performed, resulting in similar conclusions. Figure 20. Comparison of absorbed total solar radiation predicted by TRNSYS (time-dependent, discretized surfaces) and VOLTRA. Finally, following the research on diffuser modelling in CFD (Goethals and Janssens 2009b), the applicability of correlations for a range of flow regimes will be studied. These simulations are based again on the work performed during the IEA Annex 20-project. For a given Reynolds number, results will be obtained in the range of Richardson number 0 Ri 10 and a fixed Prandtl number of Correlations, extracted from literature, will be compared with the predicted CHTC values. Beforehand, however, CFD simulations of the two-dimensional case used in Annex 20 are performed to assess appropriate turbulence models. Furthermore, Derde jaarlijks rapport 23

24 the research will be extended with different configurations of inlet and outlet to optimize the design of night ventilation. Currently, however, no agreement between the CFD results and the experimental data is found yet. WP3.1.2 Experimental analysis of solar heating and intensive ventilation and its impact on summer comfort The experimental study will be performed in the PASLINK cells at Limelette. The results and conclusions drawn from this will provide necessary information for multi-zone models. During the last months, one PASLINK test cell, located at BBRI in Limelette, has been adapted into a testing infrastructure allowing a detailed monitoring aiming to a better understanding of the convective heat transfer between the ventilation air and the building surface. Originally the test cell was used to determine the thermal and solar properties of building facade components under real weather conditions in a highly standardised environment. A PASLINK test cell is basically a prefabricated, well-insulated structure, including a service room and a test room. The test room is equipped with both heating and cooling systems that allow the precise control of the indoor climate based on test schemes designed according to the component type. The PASLINK test site is equipped with a meteorological data acquisition system, continuously monitoring the outdoor climate (solar and long-wave irradiance, air temperature, relative humidity, etc.) and the indoor climate in the test cell. A full description of the original PASLINK test cell is available on As the ventilation unit is located in the original test room, a well-insulated wall has been constructed in the test room creating a second service room which comprises the ventilation unit and a new test room as shown in Figure 21. By this, the geometrical characteristics of the test room are made simpler and the air inlet and outlet can be created more easily. The separation wall is made of 200 mm EPS and is well-sealed at all joints and borders in order to assure maximum air tightness. To be able to enter the new test room, an openable element (door) is created in the separation wall. Figure 21. Plan of the modified configuration of the PASLINK test cell. Meanwhile, in the upper part of the wall, at 200 mm from the ceiling, two openings are foreseen, as shown in Figure 22 : one in the symmetry plane and one at 200 mm from the right side wall seen from inside the test room. At the lower part of the wall, one opening in the symmetry plane is located at 200 mm above the floor. Each opening can be used as an extraction opening, as an inlet or can be closed. In the closed position, the opening is filled with an EPS block, which is perfectly sealed at the edges. In case of an extract or inlet, a grille diffuser, Trox Type AT-DG/225mmx125mm/A1, is installed in the opening. This type of diffuser is chosen because it can be used to study both displacement and mixing ventilation. Moreover, this simple type can be modelled with both the momentum and box model, i.e. Derde jaarlijks rapport 24

25 simplified geometry descriptions to be used in CFD. The horizontal aerofoil blades are adjusted to a horizontal position. When the grille is used as an inlet, it is connected by an insulated duct with the fan exit. Moreover, a flow straightener, in front of the inlet grille, has been installed to obtain a representative, symmetric flow as shown in Figure 23. In case of an extract, the air is sucked from the test room into service room 2 via an identical grille. Figure 22. Location of the openings in the separation wall. Figure 23. Detailed view of the grille located in the separation wall and of the flow straightener. The measurement bay where normally the façade component to be tested is installed is filled with a well-insulated wooden frame construction. The wall is a copy of the current side walls and is equipped with a heating foil in order to be able to create a constant surface temperature. Meanwhile, in the test room the air temperature is measured at different positions by thermocouples type T (Co-Cu). As shown in Figure 24, in zone 2 and 4 of the test room, sensors are fixed on a vertical rope at three heights: at 200 mm above the floor, at mid height and at 200 mm from the ceiling. In zone 3, more detailed temperature profiles over the height of the room are measured: one sensor at mid height, four in the lower part at 20 mm, 40 mm, 100 mm and 200 mm from the floor and analogously in the upper part. In all zones, the above-mentioned configurations are installed at three horizontal locations: 150 mm away from the side walls and in-between. Also the air temperature at the inlet grille is measured by a thermocouple. Meanwhile, thermocouples, installed at the surface of the walls, measure the average surface temperature for a whole zone of a specific wall, corresponding to the zones defined in Figure 24. Derde jaarlijks rapport 25

26 Figure 24. Location of the sensors for the measurement of the air temperature and illustration of the vertical ropes. Preliminary measurements have been carried out to check the airflow rates given a welldefined supply temperature: an air change rate of 1.53h -1 and 8h -1 and a supply temperature of 15 C while keeping the surface temperature of the wall constant at 20 C or 45 C. Finally, the air velocity profile near the grille has been measured for an air change rate of 8h -1 and a supply temperature of 15 C. However, the measured velocity profiles showed a strong asymmetric character as shown in Figure 25. Therefore, a flow straightener has been incorporated. Further measurements are regarded necessary. Figure 25. Velocity profile at a distance x=0.6m away from the grille for three horizontal locations z. WP3.1.3 Numerical prediction of ventilation and local heating through solar irradiation inside buildings On the measurement equipment built in 2008 an outdoor experiment was carried out in August 2008 and September The 5-day measurement period from 24 until 28 September with clear sky and still wind conditions was selected for the experimental validation with the program VOLTRA with its solar processor. The boundary conditions to be used for the outdoor climate were the ones recorded by the climate station of the Laboratory of Building Physics on the site. However, as no diffuse solar radiation and no infrared sky radiation are recorded on this site, these data were obtained from the climate station at the BBRI in Limelette. This method was validated, for example by a comparison of the global solar radiation recorded at both sites. Generally spoken there is a good correspondence between measured and simulated temperatures. The differences can be Derde jaarlijks rapport 26

27 explained by the simplified empirical convection model, and by the inaccuracy of some of the boundary conditions. For example, VOLTRA uses a constant external (forced) convective surface heat transfer coefficient, while in reality a variable coefficient occurs due to variable wind speed and different orientations. The main conclusion is that the simulation of the transient 3D heat conduction, the infrared and solar radiation, and the convection using the program VOLTRA allows reliable results. A report (document PPB_Voltra_validation_SBO_box.pdf) with a full description of the validation tests (the internal power test carried out in June 2008 and the outdoors test mentioned above) is available. The report contains a detailed description of the box, the material properties as measured and as assumed in the simulation, the complete results (temperature courses for all thermocouple positions, isotherm snapshots and animations). Site picture Isotherms at 9h Isotherms at 14h Horizontal global solar radiation: Heverlee vs Limelette. Air (Heverlee) and sky (Limelette) temperature. Measured and simulated temperatures (3 positions) Figure 26. Test setup information and results. Derde jaarlijks rapport 27