Comparison of Computer and Model Simulations of a Daylit Interior with Reality

The lighting simulation programme Radiance is used to predict daylight factors and the illuminance distribution in a room which is 12m x 12m x 3.6m high, with grey tinted solar control glazing. These predictions are compared with measurements taken simultaneously in the real room and in a physical model placed outside. Radiance is also used to predict the lighting conditions under a CIE overcast sky distribution. These latter predictions are compared to the predictions obtained by placing the physical model in a mirror box artificial sky. Specific attention is paid to testing the accuracy of results, and the balance between computer calculation time and Radiance rendering settings. Finally the computer model is used to explore the affect of adjusting light shelf finish. The relative contribution of view window beneath the light shelf and clerestory window above is also examined. INTRODUCTION As stated in “Daylight in Buildings” the brochure that 1 summarises the rationale of the IEA Task 21 project: There seems to be some common barriers throughout the world, that hinder appropriate integration of the daylighting aspects (of building design). Three easily recognised barriers are: C Lack of knowledge on the performance of daylighting systems and lighting control strategies. C Lack of appropriate and user friendly daylighting design tools. C Lack of evidence of the advantages of daylighting. The main thrust of the IEA project is to endeavour to overcome these barriers. This study focuses on one aspect of the second of these barriers: lack of appropriate and user friendly design tools. It compares two design tools; the traditional approach of making a physical model; and the approach that is currently receiving much research and development attention, electronic modelling using computer simulation. This study uses Trad, the UNIX environment graphical user interface of Radiance, as a ‘design tool’. Trad settings are documented to establish how designers may use Radiance to provide meaningful results in reasonable time when analysing the contribution daylight can make to the lighting of architectural scenes. Convergence tests are carried out by adjusting Trad settings then comparing results with the study room measurements and documenting the calculation time. An extensive validation of Radiance’s ability to model actual skies is currently being undertaken in the UK . 2 A sky scanner is being used to measure sky luminance distributions using CIE recommendations which are 3 then modelled in Radiance. Sky luminance distributions of this detail are not readily available. This study uses idealised mathematical distributions for the sky and as such the comparisons are an indication of the performance of electronic modelling combined with these idealised formula. The ability of Radiance to model material properties such as specularity is also being investigated. SIMULATION MODELS Physical Modelling Physical models, provided they are made to scale, provide accurate photometric results and, subject to 4 the model makers ability, provide true to scale representations of the proposed building. These models can be tested in many conditions. All types of sky from clear to overcast can be found if one places the model outside under a real sky. The position of the sun in the sky at different times of the year is conventionally modelled under a real sky by tilting and rotating the model building . 5 Artificial skies inside allow conditions to be controlled. This reduces the error that can occur due to the constantly changing nature of the real sky. The construction of the necessary equipment to simulate the sun and the sky indoors is, however, very expensive. Figure 1 Internal Photograph of Study Room Physical modelling is a time consuming and costly process and is generally only viable once a building’s architectural form is decided. Changes to evaluate new design ideas are not easy to make, which may reduce the use of models as tools to explore design alternatives. Physical models constructed either during conceptual design or as marketing tools are rarely made to the photometric accuracy necessary for the analysis of lighting and daylighting. Electronic Modelling Electronic modelling is beginning to replace physical modelling as a tool for visualising building concepts. However, as with physical modelling the final appearance of the product is often a representation which has no physical basis. The flow of light in the space is illustrative rather than realistic. Recent advances in computer simulation and graphic techniques enable, in principle, the accurate physical modelling of interiors to produce realistic images and quantified photometric output. Radiance is a rendering system that was developed at Lawrence Berkeley National Laboratories in California and Ecole Polytechnique Federale de Lausanne in Switzerland. It is described by its author Gregory J. Ward as “a physically based rendering system ... 6 which blends deterministic and stochastic ray-tracing techniques”. Radiance was used in this study for the following reasons: 1. It is able to handle complicated geometry. 2. Third party CAD translation package are available to create Radiance geometry files. 3. It supports a variety of reflection models. 4. It has the ability to model sky luminance distributions using mathematical models. Radiance is not generally regarded as a user-friendly programme and it is used primarily in research. The large number of simulation variables is often daunting to the casual user and as a consequence it has not been widely accepted as a design tool in the commercial world. The graphical user interface of Radiance, Trad, was perhaps the first step towards making the programme more accessible whilst still allowing the seasoned user to customise rendering options. Adeline , a product of the IEA has reproduced many 7 features of Radiance and Trad but within a DOS environment. STUDY ROOM AND PHYSICAL MODEL DESCRIPTION Study Room The study room selected is located within the Schools of Architecture & Design building in Wellington. It is north facing and has the following features: C Small amount of extraneous reflections from adjacent buildings. C Little external shading by adjacent buildings. C Sufficient features to check for correlation between models, e.g. Internal column creating shadow. C North facing aspect so that the effects of direct sun and diffuse daylight penetration could be measured. C A window opening designed to reduce solar gain. The glass is of the grey tinted solar control type (with a transmittance of 0.46 ) 8 and is approximately 250mm from the front of the structural opening with a 500mm deep overhang above. An internal photograph of the study room, taken 21 July 1996, at approximately 10am is shown below in Figure 1. Physical Model A scale of 1:20 was selected which resulted in a model approximately 600mm x 600mm x 150mm, large enough to allow the positioning of a light sensor and small enough not to create shadows when positioned within the artificial sky. The physical model was constructed using 5mm card/polystyrene sandwich board. Particular attention was paid to window, cill, overhang, side shading and window frame dimensions. Ceiling beams were also carefully represented as they would affect reflections at ceiling level. Internal reflectances of the study room were measured using a Chroma Meter. These were then simulated Refl.'Spec.%(1&Spec.)(0.265R%0.67G%0.065B) Figure 2 Internal Photograph of Physical Model within the physical model by applying paint of varying ‘plastic’ description. The Radiance definition specified shades of grey in order to achieve reflectances which five material properties ie red, green and blue colour matched those of the coloured surfaces of the room. values (RGB), specularity and roughness. The window The floor of the model was carpeted with a remnant glass was modelled as Radiance ‘glass’ material with from the Study room. Measured internal reflectances a transmittance of 0.46. As with the physical model, were: because only the reflectance of each surface was Table 1: Study Room Measured Surface Reflectances Room Surface Refl Room Surface Refl Kitchen Partition 64 Skirting 28 Blue wall 15 Beams 76 White wall 78 Ceiling 76 Columns 78 Door 6 Window shelf 22 Perimeter Heating 5.5 Window column 13 Floor 4.6 External surfaces 11 Rear Notice Board 48 The average reflectance of the room was approximately 0.45. An internal photograph of the model is shown above in Figure 2. The photograph was taken within a photographic studio with diffuse lighting therefore little shadowing is present. Physical model construction took approximately 24 hours ELECTRONIC MODEL Model Structure A 3-dimensional model was created using AutoCAD Release 13 constructed entirely of 3-D faces . Each 9 10 face was sorted by reflectance into separate AutoCAD layers of unique colour then exported to Radiance using ‘Torad’ , an AutoCAD lisp programme created 11 12 in 1993 by Georg Mischler. As Torad does not recognise the solid modelling features of AutoCAD release 13, 3-D faces were used. The resulting Radiance geometry and material files were identified by colour (they are assigned default AutoCAD layer colours during the export procedure). Each opaque surface was modelled using the Radiance important and not its appearance, a grey scale rendering of the interior would have been acceptable. For example, a grey surface of reflectance 0.5 would be defined by its RGB values as 0.5, 0.5, 0.5. However to add ‘realism’ to the scene some surfaces were assigned colour, achieving the desired surface reflectance by adjusting the red(R), green(G) and blue(B) material descriptions using the following formula relating reflectance and specularity: The ground and adjacent buildings, modelled as simple cubes, were assigned a reflectance of 0.2. The electronic model took approximately 7 hours to construct. Radiance Settings The main user-controllable variables in Trad are the ‘quality’, ‘variability’, ‘penumbra’ and ‘detail’ settings and the number of indirect bounces (the rpict -ab 13 parameter). The ‘quality’ setting affects the overall accuracy and beauty of the renderings produced. This study examined the effects of this setting using a varying number of indirect bounces. ‘Variability’ is a qualitative indication of how light varies in the scene. An artificially lit situation produces a fairly even distribution of light requiring a low setting whereas bright sunlight patches entering a room would indicate a high setting. The greater procesing effort of the high setting has a natural time cost. This project sought to determine what setting was necessary for accuracy. As the study room was empty of furniture etc, a ‘detail’ setting of low was considered appropriate. The ‘penumbra’ calculates for softer shadows from area sources when ‘on’ but in this study was left ‘off’. Time taken to render each scene was stored in a report file designated within Trad. Following the completion of each rendering the ambient file was deleted. The ambient file stores view-independent indirect irradiance values which are shared with further renderings of the same scene. This reduces the time taken to produce each future rendering and would therefore invalidate the results. Computer simulations were carried out on a Sun Sparc20 60 MHZ computer with 96 Mb RAM.