Process Integration Platform: Applied to Multi-Disciplinary Building Design Reid R. Senescu (rsenescu@stanford.edu) Center for Integrated Facility Engineering, Stanford University, Y2E2 Building, 473 Via Ortega, Room 292 Stanford, CA 94305-4020 USA John Haymaker (haymaker@stanford.edu) Center for Integrated Facility Engineering, Stanford University, Y2E2 Building, 473 Via Ortega, Room 292 Stanford, CA 94305-4020 USA Abstract Despite the influx of building information modeling and advanced building simulations, design processes remain similar to the pre-computer era. We propose that interactive process models improve design process effectiveness and efficiency through integrating and sharing of successful design processes. We present examples of shortcomings in practice, summarize a survey of practicing engineers, and review points of departure. Based on this research we develop a theoretical framework for a web-based collaborative and interactive process modeling platform. We use a hypothetical narrative to explain its use and establish a method for evaluating the platform. Keywords: design process; interoperability; knowledge sharing; BIM Introduction To some, Building Information Modeling (BIM) is a process (Eastman, 2007). To many others, BIM at the very least changes the design process (Autodesk; Government Services Agency, 2007). The association of BIM with process change pervades the Architecture, Engineering and Construction (AEC) industry. Yet, little commercial development of BIM is process-centric, and none of the major BIM tools provide explicit support for defining and sharing processes. Despite the rhetoric that BIM is a fundamental shift in the design process, the industry generally still plugs in the new tools into a design process that predates computers. While the capabilities of simulation and modeling tools expand, order of magnitude increases in design process efficiency and effectiveness remain elusive. This paper uses two examples from current practice to demonstrate how current limitations in interoperability limit BIM’s promise to drastically improve design process efficiency and effectiveness.1 The paper then summarizes results from a survey of engineers. The survey and a literature review lay the foundation for the theoretical framework of the Process Integration Platform (P.I.P.). The paper uses a Narrative (Haymaker, 2006) to describe the characteristics of P.I.P. While the potential impact of P.I.P. on current design 1 This paper uses the International Alliance for Interoperability definition of interoperability: “the exchange of information among project participants throughout the lifecycle of a facility by direct communication between software application” (Barrett and Grobler 2000). practices is broad, the paper focuses on how the proposed platform catalyses continuous improvements to the design process through incentivizing the development and sharing of improved design processes. The paper closes with a method to evaluate the platform’s success. BIM Opportunities Lost Designing Louvers for Optimal Day Lighting With the goal of informing the design team’s decision regarding the quantity and size of louvers on the south façade of the new Stanford Graduate School of Business Library, engineers at Arup’s San Francisco office created video simulations of sunlight moving across a space. Looking to improve the realism of the output, the lighting engineers used the architect’s Revit model as the basis for their geometry. The engineers were confident in the science behind their analysis tool, Radiance, and did not have the time to learn and validate other programs. Through trial and error, they discovered they could use Revit Architecture, AutoCAD, 3D Studio Max and Radiance to create an improved day lighting simulation. The process required much manual manipulation of geometry and materials to reformat the information so that each new software package could interpret the data representing the building. Though not explicitly defined prior to an interview with the first author, the process is shown in Table 1. The engineers run these processes several times per month. The process described in Table 1 takes about 24 man-hours to provide feedback to the architects. About 16 man-hours are spent on non-value added tasks, such as cleaning up geometry or reassigning materials in AutoCAD layers that were already differentiated in Revit. The engineers felt that elimination of the 16 hours of non-value added time would either increase the number of options that the architect could consider or increase the company’s profit. The engineers believed that with improved efficiency, the design space could be further explored, and a better solution discovered. At some point, the return from expanding the options considered would diminish, and engineers would just spend less time producing the same quality design, improving company profit. Despite the obvious benefit of an improved process with improved interoperability between Revit and Radiance, no solution was being developed to meet the needs of the engineers. Table 1: The steps taken to produce day lighting videos. The steps in the process and time for each step were not documented until the authors intervened Time (hrs) Task Software 1. Create Revit Model Revit Architecture N/A 2. Import 3d .dwg with 3d shapes AutoCAD 2007 0.3 3. Clean up geometry AutoCAD 2007 4-6 4. Set layers to match materials AutoCAD 2007 1.5 5. Import 3d .dwg with 2d surfaces and layers AutoCAD 2005 0.2 6. Identify holes Radiance 2-3 7. Import 3d .dwg 3D Studio Max 1-2 8. Remesh geometry 3D Studio Max 0.2 9. Import 3DS Rendering AutoCAD 2005 0.5 10. Import 3d .dwg Radiance 1-2 11. Set sunlight properties, material reflectivity, etc. Radiance 4 12. Create video QuickTime 8 13. Discuss with architect Discussion 2-3 The first author attempted to find an improved solution. Asking for an improved process on Arup’s forum revealed suggestions for similar processes, but no solutions that matched the project’s given input and desired output. An internet search of blogs and forums reveal that other companies had similar problems trying to go from Revit to Radiance (Toast, 2004; McGrew), but no one had posted an improved solution. An interview with Revit’s developers revealed they are interested in improving links between their object oriented 3d building software and analysis programs, but development toward improving the direct link between Revit and Radiance was not planned. One potential alternative is to go from Revit to IES to Radiance, but the lighting engineers were once disappointed with this option and did not have the time to try it again. The lighting engineers were simply too busy to invest time in improving their process. Their supervisors, software developers, and their clients had no methods for understanding how an improved process could increase profits or design quality. The first author intervened by approaching some principals in the firm to explain the inefficiency of the current process, and they committed funds to improving the process. With about 16 hours of research, the first author cut step 3 and 4 to a few minutes, eliminating six hours of tedious, non-value adding time from the process. With an investment of roughly $2400, we estimate $32,400 of added value annually, assuming 6 hours x 3 iterations/month x 12 months x $150/hour. Three other documented case studies at the firm suggest that this opportunity for improving process efficiency is prolific across disciplines. In summary, the current tools provide an opportunity to improve the efficiency and effectiveness of the lighting design process. Yet, designers do not capitalize on this opportunity. The individual engineers are not incentivized to invest their time in improving the process. Their tools do not track their process (and the resulting inefficiency), place them in a small peer community to improve the process together, nor provide transparent access to other processes that could form the basis for improvements. Also, managers lack a transparent method for understanding the inefficiency and therefore, lack a monetary justification for encouraging development of alternatives. Designing with Sustainable Materials For the design of the Stanford Graduate School of Business campus, stakeholders explicitly communicated the importance of material responsibility when choosing structural systems. The first author, while working as an engineer for Arup, created schematic Revit Structure models of steel and concrete options for one building on the project. The model was used to support a visual understanding of the two systems, as well as some structural and cost analysis. Arup San Francisco had recently purchased Athena, software that uses a database to output the environmental impact of building materials. Despite a 3d object oriented model (containing a database of structural materials and quantities), a database of the environmental impacts of those materials, and a clear desire by the stakeholders for the designers to consider material responsibility in their design recommendations, the author did not conduct a formal environmental impact analysis comparing the concrete and steel options. Given the time and budget constraints, the first author was not able to find or create a process that would allow the team to effectively use the information available to further inform the team’s structural system decision with information on material responsibility. As was the case with the lighting design, there is a clear demand for an improved process. Unlike the lighting example where an improved process was non-existent, the first author recently discovered similar processes being employed internal to Arup and externally at the time of the structural design (Tobias & Haymaker, 2007). However, there was no simple mechanism to quickly search for existing processes and build on them to develop new processes. These cases illustrate that designers need a platform that enables them to more easily collaboratively improve and share interoperable design processes. Survey Affirms Generality of Case Studies The authors conducted a survey of 74 engineers from a global engineering firm that belonged to a group interested in virtual design. Half of respondents felt they should increase the amount of virtual design information that they disseminate to others. The top three barriers to sharing information were “Lack of Time”, “Not knowing where information was kept”, and “Lack of communication.” We then asked respondents to describe desirable characteristics of an intranet site for sharing information about virtual design. We coded the responses with 43 adjectives each with its own definition. The top five responses were: 1. User-Friendly: No time to understand, easy to navigate, easy to find, intuitive, good GUI 2. Crowd-Sourcing: No time to understand, easy to navigate, easy to find, intuitive, good GUI 3. Information Referencing: Reference other sites, linking to other already existing things. 4. Broad: Appeal to everyone at all positions, expertise, or all disciplines, comprehensive 5. Searchable: Ability to enter in text and find the right information 82% of respondents thought that “absolutely” the site should be “A repository of project information relevant to virtual design (software used, processes, people involved, etc.).” The survey reveals that designers want better ways of sharing information and to capture process knowledge learned from other projects, but that knowledge sharing methods must not take up user time. Broad, CrowdSourcing, and Information Referencing suggest that engineers do not want a system dependent on rigorous management, nor do they want details. Rather, they want a tool that allows everyone to participate in the exchange of information mostly by referencing information that already exists, but is not necessarily widely available. Process Integration Platform Characteristics The case studies demonstrate how a lack of interoperability process development and sharing impact the efficiency and effectiveness of the design process. Motivated by the cases and the survey, we intuit several characteristics for a webbased collaborative and interactive Process Integration Platform. A transparent design process is quickly and accurately understood by those not involved in the design. In the first case, the lighting engineers’ process is hidden from software companies, managers, architects, and other lighting engineers. No individual fully understands the opportunities for improving the process and so, no one works together to improve the process. For the structural engineer, even if a search revealed that others in the company wanted to use a building model for environmental impact analysis, no formal documentation of that process exists, and so, it would be very time consuming for the engineer to find the appropriate interoperability solution for his needs. Most engineers do not have the time to extensively document their design processes. They rarely have time to learn how to use new tools. The user-interface for process modeling and searching must be user-friendly. For interoperability solutions to be successful, they must be sharable. Sharing must be embedded in the design process. Users must be able to restrict sharing to within their company or to particular users. Designers must be incentivized to use process modeling and to develop interoperability solutions. These incentives may be embedded in the tool, and/or in the company organization. To be successful, the first user of P.I.P. must find value in its use even when no other processes or interoperability links exist. Once in use, P.I.P. also incentivizes the development of interoperability solutions. P.I.P. must be searchable by processes under multiple criteria, and the searching algorithms must be intelligent enough to understand multi-discipline and multi-cultural users and their context. The solution must allow users to combine modular parts of other processes into a new process. It should also allow geographically separated engineers to work on different parts of an interoperability solution and then combine their work. Designers must be able to choose how to scale their products and processes. For example, a structural engineer may perform one process on concrete beam section, a different process on the concrete beam, and an entirely different process on the entire building. The process modeling must permit interoperability solutions developed that are scalable. The platform must be computable, so designers can automate processes and drive processes from P.I.P. Points of Departure The case studies above demonstrate how the lack of interoperability solution development and sharing fundamentally limit BIM’s impact on AEC. We propose these deficiencies stem largely from the absence of a process modeling tool with all of the characteristics listed above. Before explaining such a tool, we first review related work in defining and sharing design information. Lee et al. (2007) use Process Models to improve product data models. They argue that product models must have a closer linkage with workflow and that mapping between processes and the product model data should become an explicit part of the definition activity. Lee et al. seek to use process models to develop future product data models. Information Delivery Manuals (IDMs) provide a humanreadable integrated reference identifying “best practice” design processes and the data schemas and information flows necessary to execute effective model-based design analyses (IAI). However, current process modeling approaches are formulated at an abstract level to define general data exchanges, processes, and design team requirements, and have limited value as a project-specific design guidance and management tool. That is, software developers, not designers, use these process models, and they are therefore not intended to be transparent, usable, and sharable tools used by project teams. Design Rationale research develops methods for better understanding the design process and ideas for improving its current state by representing the reasons behind the creation of a particular artifact (Moran & Carroll, 1996). Conklin and Yakemovic use Issue-Based Information Systems (IBIS) and process tracking to better understand and track reasons behind design decisions (1991). Professionals rarely implement Design Rationale systems, because they are not incentivized & computable - designers thus struggle to perform the extra step to document their rationale concurrently to performing design. Several CIFE methods have advanced the design of process models for use by AEC professionals. Geometric Narrator (Haymaker, 2006), enables a designer to build a scalable, computable process from automated sub processes, but lacks an intuitive visual interface and mechanisms to easily collaborate. Narratives are formal, visual descriptions of the design process (Haymaker 2006). MACDADI (Haymaker & Chachere, 2006) further communicates the people, tools, goals, preferences, options, analyses, and decisions associated in design, and contains some mechanisms for collaborative modeling, however, MACDADI and Narratives are not well integrated into software, limiting their computable power. Knowledge Sharing research explains a “project memory system” to explicitly define informal knowledge and make it available to others. However, Conklin (1996) generally applies this system to capturing knowledge from meetings, whereas we focus on capturing design process information. Software, media, internet companies, and academia open up their domains and turn to the crowd to tackle huge challenges in their industries. These Crowd-Sourcing Open Source movements transcend traditional organizational science explanation for why people innovate. Hippel and Krogh (2003) proposes that to contribute, participants must learn from the experience, should profit from networking effects of sharing or at least, not lose any money by sharing, and must receive “selective incentives,” such as a sense of ownership or enjoyment of the process. A Narrative for a Solution We have given two examples of current challenges faced by engineers and validated the generality of these examples with a survey. We proposed eight characteristics for a successful P.I.P. and explained previous attempts at improving design processes. Now, we use a narrative to further describe the eight characteristics of the P.I.P.: Transparent, Usable, Sharable, Incentivized, Searchable, Modular, Scalable, and Computable. It is early Scheme Design for a new library on the Stanford Graduate School of Business. Through the use of MACDADI, the stakeholders have communicated their relative preference for, among other goals, good lighting and reduced energy consumption. The design team is considering two geometric options for the library atrium. A meeting between the architect and the engineering team produced some sketches for an atrium option that extends in part to the basement and one that stops at the ground floor. The architect creates a Revit Architecture model and has just saved the model as a node in the webbased process model shown in Figure 1 (a). The engineers begin using that model as the basis for their individual analysis. The lighting engineer wants to create a video of Figure 1: (a) Lighting consultant adds Day Light node to analysis process to the library project. (b) Designer searches for appropriate Day Light analysis processes. Figure 2. (a) Designer browses search results; (b) looks at process info; finds two appropriate processes. (c) Working with the user community, (d) he automates one process. the daylight moving across the space, so the architect can see the different aesthetics of the two atrium options. He also wants to calculate the total energy required to light the basement for the two options, so that the mechanical engineer can use this information for an energy analysis. The lighting engineer opens P.I.P. in his internet browser. He searches for “Input: Revit Architecture,” “Output: Lighting Energy Use”. The search results display various projects that meet these criteria. He sees in the Process Info window (Figure 2 (b) ) that the process for the E Cubed Laboratory project has just been used twice, but that various links in the process have been copied into other projects 31 times. On average the process takes 22 hours to complete. Thinking that the process could easily be more efficient, the lighting engineer asks the community of users (Figure 2 (c) ) whether they want to pool resources and invest in automating the process (indicated by the green arrows). The lighting engineer selects the process from a previous project (Figure 2) and copies just the process (not the information) into the Stanford GSB project (not shown). The lighting engineer finds another process used on a different project to produce day lighting movies, but does not attempt to automate it. He selects the applicable modules of the process and again copies it into the Stanford GSB project (Figure 2). He then connects the new processes to the existing Revit Architecture model. The P.I.P. automatically runs software on the desktop computer to produce lighting energy output, which is used by the mechanical engineer for energy analysis simulations. The video making process is not automated and so, the engineer goes through each step in the process using the appropriate software. While not automated, the information created is saved to the process map, so it is easily accessible (if security privileges are granted) and understood by the entire project team. Results (not shown) are directly linked to the processes that produced them, so users can quickly compare options and understand the processes producing the results. Method for Evaluating P.I.P. This section explains our method for evaluating whether designers use more efficient and effective processes with P.I.P. In Phase 1, we observe current design processes. In Phase 2, we introduce P.I.P. to the same group of students or designers and again observe design processes. Based on observations of several student and professional design processes we categorized design processes into the following tasks: 1. Find the process to follow 2. 3. 4. 5. 6. Attain the tools or references required Strategize the modeling approach Translate information from other sources Simplify or filter translated information Generate new information (geometric, material, loads, etc.). 7. Verify that the results are reasonable or correct. 8. Troubleshoot the model. 9. Fix errors in the model. 10. Track process or results with no intent of communicating information to others. 11. Report or document results with the intent of communicating information to others. 12. Decide on a design option We ask designers to track their design processes and categorize their tasks according to this list. We then determine the percentage of time spent on non-value added tasks (e.g. Finding, Attaining, and Translating), which permits the measurement of process efficiency, defined as: Efficiency = Non-Value Added Time / Total Process Time. From this equation, we compare the relative efficiency of various processes currently in practice. The effectiveness of the process is more difficult to measure. We take three approaches. 1. Design Options Considered: We presume wider exploration of the design space increases the likelihood of finding a better design solution. 2. Designer Review: We interview designers about the quality of the results produced by the processes they follow. 3. MACDADI: Effective processes result in designs that better meet the goals of the designers. We will use MACDADI to score the quality of the designs. By combining these three measurements, we can determine the relative effectiveness of processes with similar input and output, but different information flows. Conclusion We explained case studies where engineers are not drastically improving the efficiency and effectiveness of their design process despite the availability of a 3d objectoriented model. The lighting engineer did not develop an improved interoperability solution and the structural engineer did not find an interoperability solution. To increase the development and sharing of interoperability solutions, we explain a Process Integration Platform (P.I.P.). Because P.I.P. is Transparent, Usable, Sharable, Incentivized, Searchable, Modular, Scalable, and Computable, AEC can better develop and share design process solutions. References Autodesk, Revit Building Information Modeling: Transitioning to BIM. Available Online: http://images.autodesk.com/adsk/files/transitioning-tobim_jan07_1_.pdf. Barrett, D., & Grobler, F. (2000). Understanding the different purposes of ifcs and aecxml in achieving interoperability. Available Online: http://www.iaina.org/technical/faqs.php Conklin, E. J., & Yakemovic, K. C. B. (1991). A processoriented approach to design rationale. HumanComputer Interaction, 6, 357-391. Conklin, E. J. (1996). Designing organizational memory: preserving intellectual assets in a knowledge economy. CogNexus Institute, Available Online: http://cognexus.org/dom.pdf Eastman, C. (2007). Building information modeling, What is BIM? Georgia Tech, AEC Integration Laboratory. Available Online: http://bim.arch.gatech.edu/content_view.asp?id=402 Government Services Agency (2007). GSA BIM guide overview. Available Online: http://www.gsa.gov Haymaker, J. (2006). Communicating, integrating and improving multidisciplinary design narratives. Second International Conference on Design Computing and Cognition (pp. 635-653). Technical University of Eindhoven, The Netherlands. Haymaker, J., & Chachere, J. (2006). MACDADI: A methodology for systematically & transparently achieving breakthrough quality (Tech. Rep. 10-1-2006). Stanford, CA. Stanford University, Center for Integrated Facility Engineering. Hippel, E., & Krogh, G. (2003). Open source software and the “private-collective” innovation model: issues for organization science. Organization Science, 14(2), 209223. International Alliance for Interoperability. Available Online: http://www.iai-international.org/ Lee, G., Eastman, C. M., & Sacks, R. (2007). Eliciting information for product modeling using process modeling. Data & Knowledge Engineering, 62(2), 292307. McGrew, J. Revit to Radiance (via an expensive translator...), because we can. Available Online: http://www.becausewecan.org/node/190 Moran, T. P., & Carroll, J. M. (1996). Design Rationale: Concepts, Techniques, and Use. Lawrence Erlbaum Associates Toast (2004). Questions about dxf import code for a Revitto-Blender tool. Available Online: http://blenderartists.org/forum/showthread.php?t=30227 Tobias, J., & Haymaker, J. (2007). Model-based LCA on Stanford's green dorm. International Life Cycle Assessment Conference. Portland, Oregon.