(PDF) A Visualization Dashboard and Decision Support Tool for Building Integrated Performance Optimization

A Visualization Dashboard and Decision Support Tool for Building Integrated Performance Optimization Mahmoud Gadelhak1 , Werner Lang2 , Frank Petzold3 1,2,3 Technical University of Munich 1,2 {m.gadelhak|sekretariat.enpb}@tum.de 3

Analyzing the results of multi-objective optimization and building performance simulation can be a very tedious process that requires navigating between different software and tools. There is a clear scarcity in visualization tools that combine methods for big data analysis and design decision support tools that integrate detailed information for each design and parameter. Having a single visualization tool that provides methods to both visualize and analyze a large amount of data, understand the relation between objectives and variables, and having the ability to compare and analyze the preferred designs thoroughly can support the process of design decision making. In this paper, previous attempts to develop better data visualization tools for both integrated building simulation and optimization outputs were analyzed, then guidelines and a visualization tool prototype that can be effective in decision making and analyzing multi-objective optimizations results was presented. Keywords: Multi-objective optimization, Building Performance Simulation, Simulation, Visualization tools BACKGROUND Due to global climate change and the related risks simulation results’ visualization and analysis tools are and challenges, the need to reduce carbon emissions not usually given the deserved attention. is more obvious than ever. Planners, architects, and Decision-making and data visualization tools engineers are the key players to create a more en- have a great impact on the final design product. In vironmentally friendly and sustainable built environ- a recent survey, nearly 25% of participants (architects ment. Building Performance Simulation (BPS) is an and engineers) identified graphical representation as important tool to support the move towards more their top priority for the user interface of BPS software energy efficient buildings and is becoming an inte- (Attia et al., 2009). However, another survey showed gral part of the current design decision-making pro- that most users were not satisfied with the Graphi- cess. BPS software and tools are being rapidly devel- cal User Interface (GUI) offered by commercially avail- oped to be more user-friendly and accurate. Despite able tools. In the survey, 75% of the users pointed out the high demand for more user-friendly interfaces, the lack of a graphic interface for post-processing of DESIGN TOOLS - THEORY - Volume 1 - eCAADe 35 | 719 the BPS optimization results (Attia et al., 2013). As PREVIOUS WORK a result, most of the surveyed users had to depend Analysis of optimization results on their own post processing skills or self-developed Several research works investigated the analysis of tools to graphically present the simulation output, building performance optimization results. Brown- and analyze the data. lee and Wright (2012) sought to analyze the rela- The scarcity of efficient data analysis and insuf- tionships between design objectives and variables, ficient quality of visualization tools is even more ev- using a simple ranking order and correlation coef- ident in the case of an integrated (holistic) assess- ficient. They used a combination of scatter plots ment or a multi-objective optimization. In an in- and spreadsheets to graphically present the opti- tegrated design assessment, it can be necessary to mization results. Scatter plots accompanied by par- have a dashboard that gives an overview of all the allel coordinates graphs and graphic images were relevant performance aspects, and summarizing the also used by Chaszar et al., (2016). Such graphs pro- performance of the building while simultaneously vided useful feedback, but it was noted that adding provides detailed information where needed. On the more interactive capabilities could further enhance other hand, analyzing a large number of simulation the workflow. To help designers better understand outputs, such as results from a multi-objective op- the optimization results, Wortmann (2016) presented timization, requires more advanced tools to exam- a novel method to represent the results graphically. ine the whole set of data and to find relations be- His method, called Performance Map, helps in iden- tween different objectives and variables. For multi- tifying the optimization problem, relating parame- objective optimization of integrated building perfor- ters and performance, examining promising designs, mance, there is no single visualization tool that com- and guiding automated design exploration. Other ef- bines methods for big data analysis and design deci- fective methods were also addressed in other engi- sion support through detailed information for each neering disciplines (Pryke et al., 2007; Witowski et al., of the relevant aspects. As a result, analyzing the 2009). Nevertheless, while these methods can sim- multi-objective optimization results becomes a very plify analyzing a large number of cases, it does not tedious process and requires navigating between dif- provide detailed information on each performance ferent software and tools. A single tool, that provides aspect. For instance, while using scatterplots and methods for analyzing and visualizing a large amount parallel coordinates graphs can aid in finding an op- of data, clarifying the relations between objectives timal design for daylighting performance, it does and variables, and having the ability to compare and not show how the daylight is distributed within the analyze the preferred designs thoroughly, can sup- space, or at which hours artificial lighting is needed. port the decision-making process. Such detailed information and context are necessary This paper presents guidelines and a preliminary for the decision-making process. The ability to exam- prototype of a visualization tool that can effectively ine and compare several aspects at the same time is support the decision-making in multi-objective op- also equally important. timization. In a first step, existing attempts to de- velop better data visualization tools for both inte- Integrated performance dashboards grated building simulation and optimization outputs The importance of integrating graphical representa- were reviewed and discussed. In a second step, effec- tions of diverse performance analysis in a single dash- tive techniques for the creation of a data visualization board was highlighted by many researchers. The tool that can aid in the decision-making process were Daylight-Europe project (DLE) presented the “Inte- presented. And finally, a prototype of a visualization grated Performance View (IPV)”, a multi-parameter tool was brought forward as a proof of concept. dashboard to compare reference and as-built cases 720 | eCAADe 35 - DESIGN TOOLS - THEORY - Volume 1 (Hensen et al., 1996). This dashboard proved useful in INTERACTIVITY AND VISUALIZATION comparing the overall performance of design cases, TECHNIQUES as it integrated different charts and graphs for heat- Information visualization can be defined as “the use ing load, energy consumption, visual comfort, ther- of interactive visual representations of abstract data mal comfort and glare index. to amplify cognition” (Ware, 2012). Scientific re- The IPV tool was further developed to provide search on information visualization has resulted in more flexibility and customization as well as sev- several best practices and guidelines for visualiza- eral enhancements for better communication with tion design (Cleveland, 1985; Few, 2006; Tufte, 2001; users (Prazeres & Clarke, 2005). Struck et al., (2012) Ware, 2012). Interactivity plays a major role in the vi- built upon this concept with a special focus on hu- sualization tools. Yi et al., (2007) presented seven in- man cognition and more innovative graphs, such as teractive techniques that can be effectively applied temporal maps and motion charts. Other research to the case of building performance optimization. works and commercial software also offer an inte- The first four techniques, Select, Explore, Reconfigure grated performance dashboard. However, most of and Encode, can help the user explore the whole set these tools lack the ability to deal with a large amount of data by switching between different graphical rep- of data, and thus cannot be efficiently used to analyze resentations of data and marking preferred designs. the results of multi-objective optimizations. The other three techniques, Abstract/Elaborate, Fil- ter, and Connect, can be used to provide detailed in- Figure 1 A screenshot of the visualization tool. A- Context and design parameters panel, B- Explorer panel, and C-Integrated dashboard DESIGN TOOLS - THEORY - Volume 1 - eCAADe 35 | 721 formation for selected cases. Adding two other tech- 3- Detailed results and comparison be- niques, such as Compare (for directly comparing se- tween favorite designs lected cases) and Advice (as a tool for guiding further At the final level, an integrated dashboard is pre- enhancements) allows for quick, yet thorough, com- sented with detailed performance data, which pro- parison between preferred cases and supports in- vides all the needed information about each selected formed decision-making. These two additional tech- design. To ensure an informed decision-making pro- niques were also suggested by Haeb et al. (2014), cess, the visualization tool should offer the ability to who highlighted the importance of spatial context compare the detailed performance and contextual and visual feedback as an essential component in the reference (images and 3D model of the cases) of fa- field of building performance simulation. vorite cases. VISUALIZATION DASHBOARD: GUIDE- VISUALIZATION DASHBOARD: PROTO- LINES TYPE Building on the reviewed literature, the following As a proof of concept, a preliminary prototype guidelines, and requirements for a new tool for vi- was developed according to the above-mentioned sualizing the results of integrated building perfor- guidelines. The prototype was built using the vi- mance optimizations were defined. The suggested sual language programming tool Grasshopper and visualization tool can provide better ways to inves- HumanUI, a plugin for Grasshopper that enables tigate the building optimization results by offering the creation of graphical user interfaces. Additional three levels of data analysis: Grasshopper user objects and code functions were written to overcome limitations in the HumanUI Plu- 1- Design space overview and exploration gin. The visualization tool consists of the following At the first level, the full set of simulation results panels. should be explored. Multi-dimensional graphs, such Context and design parameters. The left panel as parallel coordinates and scatter plots, are useful in shows the names of the selected design alternatives this case. Switching between plot types, filtering the as well as a zoomable and rotational 3D visualization results and selecting favorite cases help in clarifying and the corresponding design parameters. The user basic relations between the objectives and variables, can change the design parameters to specify they de- in addition to highlighting optimal and preferred de- sign alternative (Figure 1-A). signs. Explore. Alternatively, the user might choose to select the design from the Explore section. The 2- Sensitivity analysis and parameter rela- Explore section contains a parallel coordinates chart tions that shows the design variables and results of the On the second level, the direct relation between any complete design set with the selected design high- two variables or objectives can be investigated. The lighted. Additionally, it also contains a radar or bar use of sensitivity analysis and 2D charts can indicate chart for showing the performance of the selected the variables that drive the optimization process, the design as well as a data table. The user can filter expected enhancement in each objective, and the the design alternatives by limiting the values of any relative importance of the design variables. of the variables or the objectives. It is also possible to mark cases in order to be compared later to each other (Figure 1-B). Variable-objective relations. In this panel, the user can choose a variable(s) and objective(s) to see 722 | eCAADe 35 - DESIGN TOOLS - THEORY - Volume 1 Figure 2 The workflow of the optimization process and the corresponding visualizations panels in the visualization tool. the direct relation between them, which is rendered discussed earlier in the literature, the integrated in the shape of a 2D scatterplot chart in the case of a dashboard houses more details for all the perfor- single variable and single objective, or as a matrix of mance objectives for the selected design alternative. scatterplots in the case of several variables and ob- Performance objectives tabs. For each perfor- jectives. mance objective, a separate section that includes al- Compare. The compare panel offers a bar chart ternative ways of result visualizations an even greater to compare the marked design alternatives as well as detail of result analysis is provided. simple visualization of each performance objective. To make it easier for the user to explore and Integrated Dashboard. Similar to the IPV tool choose preferred designs, the ability to show or hide DESIGN TOOLS - THEORY - Volume 1 - eCAADe 35 | 723 Figure 3 Different design alternative with similar energy savings. panels was provided. This enables the user to focus Parametric model on a specific panel or several ones, e.g., the Explore The optimization was carried out for the south façade & Integrated Dashboard or the Explore & Compare. of a single office space in Munich, Germany (48°8’N Figure 1 shows a screenshot of the visualization tool 11°34’E). The office room was assumed had the di- prototype with three active panels: Context and de- mensions of 4.00m x 6.50m x 3.00m for the width, sign parameters, Explore and Integrated Dashboard. depth, and height, respectively. The parametric model of the south façade provided different set- CASE STUDY tings for the glazing area, glazing system, shading, The visualization tool prototype was used to visualize daylighting system, and insulation system. The glaz- the results of a building performance optimization, ing area was divided into upper and lower parts, in which a parametric model was optimized for the where the upper part acted as a clerestory window. integrated performance of the following parameters: Both window parts were introduced to the shad- Energy consumption, Daylighting, Thermal Comfort, ing devices separately. Seven Window-to-Wall Ratios View and Glare and Renewable Energy. The paramet- were studied together with four glazing systems, four ric model was built using Grasshopper. EnergyPlus shading systems, and four light-shelf settings. Addi- [1] and Radiance simulation engines, used through tionally, the building insulation was increased gradu- the HoneyBee (Roudsari, M. S. & Pak, M., 2013) and ally with 2.5 cm steps to a maximum of 25 cm. Over- Diva (Jakubiec, J. A., & Reinhart, C. F. 2011) plugins, all, nearly 20,000 design alternatives can be gener- were utilized for the energy and daylighting assess- ated. ments. A multi-objective optimization was carried out using the optimization tool Octopus (Vierlinger, Multi-objective optimization Workflow R., & Hofmann, A. 2013). A multi-objective optimization was performed with the evolutionary algorithm SPEA2 using Octopus. A random generation was created at first, which con- tained different cases (genomes). Then for each 724 | eCAADe 35 - DESIGN TOOLS - THEORY - Volume 1 Figure 4 Scatterplot matrix chart showing the relation between energy savings, insulation and glazing type. genome, daylighting, energy and thermal comfort grated photovoltaics was calculated. The results from analysis were carried out using DIVA and Honeybee. the analysis phase were used as the fitness values for At the same time, the openness of the façade to the the optimization, namely the spatial daylight auton- view outdoors and the possible area for building inte- omy, energy savings compared to an unshaded base DESIGN TOOLS - THEORY - Volume 1 - eCAADe 35 | 725 Figure 5 Examples of the contents of the performance tabs. A- Energy use analysis. B-Daylighting performance analysis. C- Thermal comfort. D- Visual comfort (Glare and View). E- The compare tab showing a simple comparison of three selected cases. case, the percentage of comfort hours, view and PV Optimization results analysis percentages. The optimization process continued for The optimization resulted in 150 design alternatives. 10 generations with a population size of 30. The mu- The results were analyzed in several ways using the tation rate was set to 0.5, the mutation probability to visualization tool prototype. First, the results and 0.1 and the crossover to 0.8. Cases with very high so- parameters of all the 150 cases were analyzed us- lar exposure were neglected. ing the Explore section. The parallel coordinates During the analysis, the results from the simula- chart offers an interactive tool by which the results tions were post-processed into different types of vi- could be filtered for a specific range of values for any sualizations according to the results type. After the and each of the variables and objectives. Addition- optimization ends, the optimization results are also ally, the results in the data table could be sorted for visualized using parallel coordinates and scatter plot any of the variables and objectives. A radar chart charts. Figure 2 shows the workflow of the optimiza- for the objective results is also shown for the se- tion process and the corresponding data visualiza- lected design alternative. In this case study, the de- tion in the prototype. sign alternatives were sorted for highest energy sav- ings. It was found out that several design alterna- 726 | eCAADe 35 - DESIGN TOOLS - THEORY - Volume 1 tives achieved energy savings between 30-32%. Al- CONCLUSION AND DISCUSSION though these cases achieve a similar energy perfor- Energy efficient and sustainable buildings are slowly, mance, their design parameters and other objective but surely, becoming the standard in architecture performances differed greatly. Only one alternative, and building practices. As building performance for instance, achieved an acceptable daylighting per- simulation software and optimization tools become formance value (sDA more than 50%). This enables more common in the building design process, it is the designer to choose his design wisely by taking all vital to have an integrated result analysis and visual- the objectives and also the design features in mind. ization tool to support the design decisions. This pa- A trade-off between the different objectives is of per presents a prototype for a visualization tool that course necessary. Figure 3 shows the four design al- can help analyze the results of building performance ternatives with the highest energy savings and a min- multi-objective optimizations. The visualization tool imum daylighting performance of sDA= 50%. Their aids in investigating the whole design set, analyzing corresponding performance for the other objectives the relation between variables and objectives, as well is illustrated using the radar chart. as comparing and further investigating preferred de- In a second step, the relation between variables signs. As a result, the user can define areas with po- and objectives can be studied using scatter plots ma- tential enhancements, find the most effective design trix in a separate window. For instance, the scatter variables and compare the integrated performance plot between the energy savings and glazing and in- between different designs in a visually-informative sulation shows how triple glazing and double low- way. By achieving these different functions, the tool E glazing have a higher potential for energy savings can help in the design decision process by shortening compare to single and conventional double glazings. the time required to analyze the vast amount of data For the insulation, it could be noted that the poten- resulting from multi-objective optimization. In future tial for energy savings increase with the increase of works, other enhancements could be investigated, the thickness of the insulation. Nevertheless, most of such as building the tool with a more sophisticated the cases with 10 cm insulation were able to achieve programming language like Python or Java, support- energy savings between 25% and 30% (Figure 4). ing dashboard customization, and validating the tool To compare the performance of the preferred de- by focus groups. Implementing a guiding system can signs, marked cases are automatically added to the also be a valuable addition to the prototype to ensure compare panel, where a simple bar chart comparison that an optimal performance is reached by showing is created. Finally, the integrative dashboard and per- possible areas of enhancement. formance tabs show detailed and alternative visual- izations for each of the optimized objectives. Other REFERENCES simulation outputs can also be investigated such as Attia, S, Beltran, L, de Herde, A and Hensen, J 2009 ’Archi- lighting and occupancy schedules; heating, cooling tect Friendly: a comparison of ten different building and equipment’s load; alternative daylighting perfor- performance simulation tools’, Building Simulation mance metrics like the daylight autonomy, daylight 2009 Proceedings of the Eleventh International IBPSA availability, ... etc.; glare analysis for various times and Conference, Glasgow, Scotland Attia, S, Hamdy, M, O’Brien, W and Carlucci, S 2013, dates; ... etc., Figure 4 shows part of the different tabs ’Assessing gaps and needs for integrating building for a single design alternative and the compare tab performance optimization tools in net zero energy with a comparison between the cases with the high- buildings design’, Energy and Buildings, 60, pp. 110- est daylighting, energy and BIPV area (Figure 5). 124 Brownlee, A and Wright, JA 2012 ’Solution Analysis in Multi-Objective Optimization’, Proceeding of BSO12 the First Building Simulation and Optimization Confer- DESIGN TOOLS - THEORY - Volume 1 - eCAADe 35 | 727 ence, Loughborough, UK, pp. 317-324 Ware, C (eds) 2012, Information Visualization: Perception Chaszar, A, von Buelow, P and Turrin, M 2016 ’Multivari- for Design, Elsevier ate Interactive Visualization of Data in Generative Witowski, K, Liebscher, M and Goel, T 2009 ’Decision Design’, 2016 Proceedings of the Symposium on Sim- making in Multi-Objective Optimization for Indus- ulation for Architecture and Urban Design trial Applications–Data mining and visualization of Cleveland, WS (eds) 1985, The elements of graphing data, Pareto data.’, Proceedings of the 7th European LS- Wadsworth advanced books and software, Mon- DYNA Conference, Salzburg, Austria terey, CA Wortmann, T 2016 ’Surveying Design Spaces with Perfor- Few, S (eds) 2006, Information dashboard design, O’Reilly mance Maps: A Multivariate Visualization Method Haeb, K, Schweitzer, S, Fernandez Prieto, D, Hagen, E, En- for Parametric Design and Architectural Design Op- gel, D, Bottinger, M and Scheler, I 2014 ’Visualization timization’, Complexity \& Simplicity - Proceedings of of Building Performance Simulation Results: State- the 34th International Conference on Education and of-the-Art and Future Directions’, 2014 IEEE Pacific Vi- Research in Computer Aided Architectural Design in sualization Symposium (PacificVis), pp. 311-315 Europe,, Oulu, Finland, pp. 239-248 Hensen, J, Clarke, JA, Hand, JW, Johnson, K, Wittchen, Yi, JS, Kang, YA, Stasko, J and Jacko, J 2007, ’Toward a K and Madsen, C 1996 ’Integrated Performance Ap- deeper understanding of the role of interaction in praisal of Daylight-Europe Case Study Buildings’, information visualization’, IEEE transactions on visu- Proceedings of the 4th European Conference of So- alization and computer graphics, 13(6), pp. 1224- lar Energy in Architecture and Urban Planning, Berlin, 1231 March, pp. 1-6. [1] http://apps1.eere.energy.gov/buildings/EnergyPl Jakubiec, JA and Reinhart, CF 2011 ’DIVA 2.0: integrating us/ daylight and thermal simulations using Rhinoceros 3D, DAYSIM and EnergyPlus’, Proceedings of Building Simulation 2011, Sydney Prazeres, L and Clarke, JA 2005 ’Qualitative Analysis on The Usefulness of Perceptualization Techniques in Communicating Building Simulation Outputs’, Build- ing Simulation 2005, the Ninth International IBPSA Conference, Montreal, Canada. Pryke, A, Mostaghim, S and Nazemi, A 2007 ’Heatmap Visualization of Population Based Multi Objective Algorithms’, K. Deb, C. Poloni, T. Hiroyasu, & T. Mu- rata (Eds.), Lecture Notes in Computer Science. Evo- lutionary Multi-Criterion Optimization, Vol. 4403, pp. 361–375. Roudsari, MS and Pak, M 2013 ’Ladybug: a paramet- ric environmental plugin for grasshopper to help designers create an environmentally-conscious de- sign.’, Proceedings of the 13th International IBPSA Con- ference Held, Lyon (France) Struck, C, Bossart, R, Menti, UP and Steimer, M 2012 ’User-Centric and Contextualised Communication of Integrated System Performance Data’, BauSIM 2012 : Fourth German-Austrian IBPSA Conference ; IBPSA Berlin University of the Arts Tufte, E (eds) 2001, The Visual Display of Quantitative In- formation (vol. 2), Graphics Press, Cheshire, CT Vierlinger, R and Hofmann, A 2013 ’A Framework for flexible search and optimization in parametric de- sign’, Proceedings of the Design Modeling Symposium, Berlin (Germany) 728 | eCAADe 35 - DESIGN TOOLS - THEORY - Volume 1