LiQuID: Lighting Quality In Design James Peng1, Ben Liao1, Daniel Glaser1, John Canny1, Ellen Yi-Luen Do2 1 2 University of California, Berkeley University of Washington, Seattle Primary contact address: jpeng9999@hotmail.com Oftentimes, architects and lighting designers do not fully utilize natural daylight when designing buildings and lighting systems. Buildings with artificial lighting systems are simpler to design and analyze. However, doing so would forgo some of the ecological, social, and economical benefits of daylight. Current simulation tools for predicting visual quality and energy use of architectural lighting systems are not very informative or useful when concerning the use of daylight. Photorealistic renderings produced by a number of simulation systems are only relevant for a specific set of variables, including time, season, and sky condition. For instance, the distribution, intensity, and angle of daylight can change rapidly during the day. Also, lighting quality depends upon viewer direction and location within a room, and current tools provide little assistance for managing multiple viewpoints. Tools that aggregate these results do so only at the expense of compromising fine-grained analysis details. LiQuID is an analytical tool for evaluating a comprehensive set of lighting data to account for the complex nature of daylight. The program is designed to generalize large collections of lighting data by performing automatic classification, lighting quality analysis, and a summary component. Automatic classification is a data-driven method for identifying and organizing sets of data, based on similarities in light intensity and distribution. Lighting quality analyses are based on the normative lighting characteristics defined in industry standard handbooks (e.g.(Illuminating Engineering Society et al. 2000; Benya et al. 2001)). These sources contain guidelines on evaluating architectural models for factors like direct glare and light distribution. A summary module integrates with a user interface for a building designer to review. LiQuID works in conjunction with a modeler, simulation program and visualization toolkit. Initially, an architectural design is created with a modeler or CAD program. The three-dimensional geometry is broken down into wall, ceiling, and floor surfaces. Designers can also define viewpoints within the room to evaluate a viewer's actual perspective. Next, a high quality simulation package (such as in (Ward 1994)) is used to generate illuminance and luminance data for the given planes and viewpoints, for a variety of temporal and sky conditions. After the lighting quality analysis, the visualization toolkit displays the results of automatic classification and statistical evaluation. This can enhance existing methods for display of daylight data (Cheng and Pat-Yak Lee 2001; Glaser and Ubbelohde 2002; Papamichael 2002) to also include statistical trends. Through automatic classification, LiQuID detects the major patterns in shape and intensity of lighting within the building design. Classification utilizes a k-means algorithm, which takes large quantities of simulated lighting data and groups them into clusters based on similar patterns. Within each cluster is the subset of all lighting data most similar to one another. Instead of analyzing all the output from the simulation program, these clusters can be examined for their normative lighting characteristics. Examples include point-by-point analysis of lighting intensity and the comparison of adjacent surfaces and perspectives for discontinuities. After classification and lighting quality analysis, LiQuID can present the evaluation of the architectural model. Reports of problematic lighting, such as high glare or low ambient lighting, are accompanied by relevant information, such as frequency or the reasons for occurrence. In addition, the program may give suggestions on improving the lighting quality of the model, like installing shading devices to reduce glare. By analyzing numerous lighting simulations across multiple variables, LiQuID will extract useful lighting information from an architectural model. LiQuID aims to relieve architects and lighting designers from the tedious work required to analyze countless potential lighting situations. Also, it aims to remove some of the oversimplifications and generalizations about daylight that are utilized when evaluating lighting systems. Therefore by doing careful analysis of a comprehensive amount of simulations, LiQuID aims to provide reliable, accurate and legible feedback to building designers to encourage them to incorporate lighting, energy, and economic concerns of daylighting into their designs. REFERENCES: Benya, J., L. Heschong, T. McGowan, N. Miller and F. Rubinstein, Eds. (2001). Advanced Lighting Guidelines. White Salmon, WA, New Buildings Institute. Cheng, N. and E. Pat-Yak Lee (2001). Depicting Daylighting: Types of Multiple Image Display. Proceedings of the Association for Computer Aided Design in Architecture (ACADIA '01), Buffalo, New York. 282-291 Glaser, D. and S. Ubbelohde (2002). "Techniques for Managing Planar Daylight Data." Building and Environment 37(8-9): 825-831. Illuminating Engineering Society, M. S. Rea and Illuminating Engineering Society of North America (2000). The IESNA lighting handbook : reference & application. New York, N.Y., Illuminating Engineering Society of North America. Papamichael, K. (2002). A Web-based Virtual Lighting Simulator. Proceedings of the Association for Computer Aided Design in Architecture (ACADIA '02), Pomona, CA. 269-277 Ward, G. (1994). "The radiance lighting simulation system." Computer Graphics 28(7): 459-72. LiQuID Lighting Quality In Design James Peng1, Ben Liao1, Daniel Glaser1 John Canny1, Ellen Yi-Luen Do2 1University of California, Berkeley 2University of Washington, Seattle http://iolanthe.cs.berkeley.edu/~peng simulation tools for predicting visual quality and energy use of Problem: Current architectural lighting systems are not very informative or useful. P1. Photorealistic images are easy to understand, but only describe a single moment of time in simulation space, thus requiring many images to account for all possible conditions. Jun 21st 9am Sky Condition: Clear Dec 21st 9am Dec 21st 12pm Sky Condition: Overcast Sky Condition: Clear P2. Lighting quality also varies by viewer perspective. Current tools provide little assistance for selecting a comprehensive set of viewpoints. East View N South View Sample Images from Lawrence Berkeley National Laboratory Virtual Lighting Simulator Solution: Automatic classification can help users quickly understand lighting problems and characteristics of their architectural designs. S1. Cluster the vast amounts of time-varying lighting data according to normative lighting characteristics (such as glare level or minimum illuminance) to identify significant trends. GROUP: very dim lighting GROUP: fair lighting S2. Create representative slices of the architectural model based on automated and user-specified conditions of location and viewer perspective. GROUP: high glare lighting 3D Model Surface Lighting Data Lighting Configuration #3 Day: Dec 21 Time: 12:00 pm Sky Condition: clear Surface Slices Insert Viewpoints Perspective Lighting Data Perspective Slices 1 1 Will require excessive electric lamps to light most of the year Efficient daylighting – saves energy during springtime These clusters were generated with LiQuID 2 Shading devices recommended for early morning 2 Wall 1 Illuminance -5 fc (min - ... Future: We are currently: • Prototyping interfaces to summarize lighting characteristics and suggest recommendations • Generating appropriate slices and views • Refining normative lighting characteristics • Integrating with existing research programs LightSketch and Scythe & Sew • Testing user interface with professionals Jan Dec 9am Low Ambient Lighting - Winter: 66% Fall: 45% - 9am: 55% High Glare - Summer: 55% Fall: 35% 12pm: 40% 3pm: 25% Predicted Energy Usage – 3 kw/ft2 (excessive) Recommendation: Install Lightshelf on south window 6pm Average Ambient Lighting - 10 footcandles Conceptual sketch of LiQuID integrating with Scythe and Sew