exploring automatically generated design...

1 Copyright © ASME 2010

Proceedings of the ASME 2010 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference

IDETC/CIE 2010August 15-18, 2010, Montreal, Quebec, Canada

DETC2010-28765

EXPLORING AUTOMATICALLY GENERATED DESIGN CONCEPTS USING CYBERINFRASTRUCTURE

Aziz M. NaimDepartment of Mechanical and

Aerospace EngineeringUniversity at Buffalo – SUNY

Buffalo, NY [email protected]

Kenneth W. EnglishNYS Center for Engineering Design

and Industrial InnovationUniversity at Buffalo – SUNY

Buffalo, NY 14260Corresponding Author

[email protected]

Kemper E. LewisDepartment of Mechanical and

Aerospace EngineeringUniversity at Buffalo – SUNY

Buffalo, NY [email protected]

ABSTRACTOur society has witnessed major growth and innovation in

technology. However, one of innovation’s fundamental aspects, designer creativity, is still largely unsupported by information technology (IT) tools. This paper establishes a bridge between efforts to use cyberinfrastructure to generate design alternatives and a designer’s ability to explore and evaluate concepts.. The research presented provides a method of synthesizing a web-based representation of a product concept using information from a digital design repository, an automatic concept generator, and results from a clustering analysis. A visualization interface is developed to provide designers with multiple representations of product concepts. The development of this approach establishes a computing infrastructure to support an investigation into the effect of information technology on designer creativity and provides a foundation for the development of further support technology. The architecture is implemented and the result is illustrated with an example in order to show the capabilities of the approach.

1 INTRODUCTIONThis work focuses on using the wealth of information

available from computer-based design repositories and concept generation methodologies to facilitate IT-enabled innovation. Recent research has developed automated concept generation techniques in order to support the process of design innovation. This has led to the development of various design repositories in order to accommodate the large amount of information being generated for storage and retrieval of the concepts [1-3]. Traditionally, designers would sort through the generated concepts with minimal support from computational tools in

the exploration of the concept space. Ideally, designers would benefit from IT-based approaches that assist in the identification of innovative concepts for more detailed evaluation. This work presents an approach that leverages computational resources to create a more efficient way of guiding a designer to a design that would satisfy an identified need in a novel way.

The exploration of a large number of automatically generated design concepts relies on information from multiple sources. Managing the integration of this information is cumbersome to the designer, and the natural limitations on a designers’ ability to navigate through a very large set of concepts renders the process quite ineffective. This limitation creates a need for assisting designers in synthesizing a visualization of the available concepts using data from multiple sources (e.g., automated concept generation and design repositories). This would fill the void and mitigate the intimidating effect a large set of data might have on a designer.

This paper is a first step in developing a methodology to assist designers as they explore design alternatives and select candidates of interest. The result of this research is a method for integrating information from multiple sources related to concept generation and a novel platform for visualizing and exploring computer generated concept variants.

In the following section, we review the related work that serves as a foundation for our development. In Section 3, we present the underlying architecture and inner connectivity of the various IT-based platforms. The architecture’s interface is illustrated using an example in Section 4. In Section 5, we provide some general discussion, conclusion and future work regarding the approach.


2 RELATED WORKThe presented approach occurs at the intersection of several

technology areas: cyberinfrastructure, design repositories and automated concept generation, and finally concept exploration. This section contains an overview of the relevant material from each of these areas.

2.1 Digital Libraries to Support EngineeringCyberinfrastructure is becoming more and more relevant in

the overall design process. As we progress towards a knowledge economy, a need arises for an infrastructure that is based upon distributed computing, information and communication technology. However, in order for this knowledge-based economy to become reality, improvements must be made at all levels of the supporting cyberinfrastructure, including base technologies (e.g., computer architectures, networks, communication schemas), information services (e.g., automatic concept generation and design repositories), and decision support interfaces. The NSF’s Advanced Cyberinfrastructure Program has proposed to “use cyberinfrastructure to build more ubiquitous, comprehensive digital environments that become interactive and functionally complete for research communities in term of people, data , information, tools, and instruments”[4].

One example of the application of cyberinfrastructure to engineering is the development of digital libraries to enable the “collection, storage, organization, sharing and synthesis of huge volumes of widely disparate information” [4]. Research and development of wide-scale systems began with the establishment of the Internet [5]. Examples of such libraries include the educational digital library hosted by the National Science Digital Library [6] and a digital library of learning resources for engineering education that is available through the National Engineering Education Delivery System [7]. Those libraries provide resources and tools that support innovation in teaching and learning, but provide limited support for engineering design.

In design, engineers perform a multitude of tasks using various tools and resources that generally do not exhibit a high degree of information integration due to differences in data representation, storage and reference [8]. Thus a clear need arises in order to integrate design information used in evaluating design concepts. One solution was developed in [3] by providing an effective way to capture design data (i.e., geometric and functional product information) through product dissection. This work was focused on improving the quality of engineering undergraduate education [9, 10]. Although very promising, this work outlined a need for archival and long-term storage of complex engineering data. Furthermore, the ability to integrate various cyberinfrastructure tools for concept generation and exploration has not been addressed yet, and remains an open topic for research.

2.2 Design Repositories and Concept GenerationBy definition, a design repository is a digital collection

of knowledge that describes, as completely as possible, a set

of products. The objective of developing design repositories is to support engineering design by creating, populating and retrieving design knowledge at various levels of abstraction, including form, architecture, and function. In 1999, researchers at various universities and the National Institute of Standards and Technology (NIST) began gathering artifact information for a national design repository[11]. Recently Oregon State University has partnered with the University of Texas [12, 13], Penn State, Virginia Tech, University at Buffalo, Texas A&M University and Bucknell University [14] to expand the types of design information and design tools within the repository. Researchers at Drexel University have been performing research on the importance of roles and semantic structure of information contained in design repositories[15]. Design repositories offer significant potential in the support of designers in the concept generation phase of product development.

In parallel to the development of design repositories, concept generation based on computational reasoning has been developed using formal definitions for describing function or purpose in engineering design [16, 17]. The ability provide engineers with a large number of automatically generated design concepts offers a way to more broadly explore the space of possible design alternatives. One method has been inspired by the collection of empirical knowledge and relating components to functions which led to the development of relational matrices [12] and graph grammar rules [18]. Combined with a search mechanism, the matrices and rules can be used to automatically generate conceptual designs. This approach assumes that a designer has developed a functional model of the proposed design. The matrices and rules enable the automatic generation of design concepts using one or more components to achieve a required function. As a result of the open-endedness of conceptual design and the combinatorial nature of automatic concept generation, a very large number of solutions (i.e., more than a million) can be created through the generation process. Consequently, the volume of data can overwhelm a designer’s ability to explore the various conceptual solutions effectively [19].

In order to remedy the cumbersome searching procedure the Designer Preference Modeler (DPM) [20] was created. The DPM searches a large set of results to find concepts that could be most meaningful to a designer. By ranking selected concepts, the search mechanism learns what aspects of the concept the user prefers, and seeks solutions that maximize the predicted preference. Clustering techniques also provide a range of approaches to handle the solution information overload issue. Clustering is a general term for a set of exploratory data analysis techniques used to solve grouping or classification problems [21]. Typically heuristic in nature, clustering methods are most applicable when there is little information about the underlying structure of the data [22]. The objective of clustering analysis is to sort a set of data into groups such that each member of a cluster has a high degree of similarity with any other member of the cluster and a low degree of similarity with any non-member of the cluster. By clustering large numbers of candidate design


solutions into groups, more effective representation methods, including visualization tools can be used to support concept evaluation and exploration.

2.2 Concept Exploration Visualization is considered as an essential support tool for

designers. From the designer’s sketchpad to a fully immersive, simulation of a plant, visualization allows designers to be more intuitively engaged in the design process. From CAD systems, where designers can translate their ideas into production drawings, to the use of virtual reality which has provided designers the ability to actually see and experience their design before a single part is constructed [23-26]. Visualization has had a significant impact on the design process, resulting in savings by eliminating redesign and improving the fit and performance of parts [27].

Furthermore, visualization has been extended into the realm of concept exploration. Approaches like Graph Morphing [28], Cloud Visualization [29], Physical Programming Visualization [30], and Multidimensional Visualization [31] provide more functional information to the designer during the early stages of the design process in the form of design attribute or optimization information. Some additional help in concept selection has been incorporated using a “design by shopping” [32] perspective. The Advanced Trade Space Visualizer (ATSV) system [33] guides designers to a solution in an efficient manner. Unfortunately, this method may lead to erroneous results if the designer is not aware of his preferences. These techniques are often used for solution validation and selection rather than efficient exploration

tools, especially for large problems. The inability to visualize the actual form or functional structure of a design concept limits the designer’s ability to fully explore the space of available design concepts.

With the globalization of the economy, a need for better integration of the large amount of data available to designers around the world has emerged. Advances in information technology have driven the development of visualization for product designs [34]. Industry uses them for collaboration between different components using solid modeling [35, 36]. While these prove useful in identifying production and assembly issues, the collaborative platforms focus on geometry, rather than functionality, thus not providing any concept exploration capabilities.

To address the above mentioned issues, we present our IT-based platform to support concept exploration in the following section. The approach synthesizes information, including clustering, connectivity, and functional behavior into a visual format using a combination of web scripting languages (e.g., PHP [37]) a common schema and databases.

3 UNDERLYING ARCHITECTUREThe developed approach interconnects existing IT-based

resources such as functional modeling interfaces, design repositories and concept generators into a newly developed exploration interface to support the designer’s ability to efficiently explore the available concepts. Figure 1 presents its main components of the method.

Figure 1. Overview of the Concept Exploration Framework.


As previously discussed, design repositories [11] archive and make available information on how components fulfill functional requirements in existing products. Concept generation techniques [12, 13] can leverage this information to automatically create large numbers of potential designs that meet a given functional model. Clustering approaches [21, 22] provide a means of grouping similar items together in a way to reduce cognitive load on a designer. The framework developed in this work focuses on the development of the necessary interconnecting cyberinfrastructure and interfaces to help designers to navigate large sets of computationally generated designs. Rather than develop an approach that is tied to a specific implementation of computer-based concept generation and archival tools, the architecture described in this section is flexible and can scale with increasing computational power.

The approach integrates data from multiple sources in an attempt to present a more intuitive representation of design alternatives. The process begins with a functional model for a proposed design. Using the functional model, a set of design alternatives are created by a concept generation engine. For each concept, details about each component are retrieved from a design repository. Rather than present a long list of alternatives, the set of concepts are also passed to a clustering routine in the interest of identifying similarities amongst the generated alternatives. [38, 39]. These activities result in a wealth of data that may be useful to a designer, without an intuitive means of exploring the space of design alternatives.

As shown in Figure 1, the core of the framework is separated into two components, the Data Integration and Validation Module and the Web Server Interface. This separation of data from presentation was purposely implemented to allow for flexibility in the development of future clients (e.g., using a iPhone to connect to a web service interface). The approach also enables a scalable data processing approach, with the potential for implementation in a cloud-based computing environment if the number of candidates to be considered grows to be very large.

The web server interface provides a prototype platform for evaluating the work flow for exploring design alternatives and a flexible platform for initial efficacy trials. The actual representations and modes of interaction are explained in Section 4 and this section focuses on establishing the underlying architecture enabling the efficient representation of the generated concepts.

Methodologies developed to support the exploration of automatically generated design concepts face two primary challenges. First, any method must represent all the available information to the designer in an efficient and useful way. Second, the underlying architecture needs to be developed in a flexible and scalable manner to support efficient transfer of information between all the modules without losing any necessary information. One of the main issues that must be addressed is the development of a common taxonomy that will

be used in order to pass the information between components of the framework concept generators and design repositories.

3.1 Developing a Common Schema In Figure 1, information flows between every component

of the framework. Efficient functionality is thus based on an efficient information flow between the modules. In order to ensure proper information flow, a foundational naming taxonomy needs to be established. The methodology developed in this paper relies on the component naming taxonomy presented in [40]. This hierarchical naming system for engineering components supports the functional modeling approach used to describe the system of interest and by the concept generator to identify components.

The schema selected in this work is based on the Extensible Markup Language (XML) [41] and was originally developed to capture information from a given tool and pass it to another one for a representation of a function [42]. The advantage of using a consistent communication schema is that no external translation services are needed when different implementations of subcomponents are used, providing a flexible, federated infrastructure. The schema used in the presented work to capture the functional behavior of a given concept is based on Graph Synth [43]. An example is shown in Figure 2. The schema for each concept is structured using nodes and arcs. Each node entry corresponds to a function present in the overall functional model. Arcs represent the connectivity between the functions and hold information about the type of connection present and its source and sink node. Another schema is presented in Figure 3 and is used to capture the concept clustering. Each cluster is identified by a unique id with and a binning approach is used to classify the candidates.

Figure 2. Example Segment of Concept Description Message

<node> <name>Input_External</name> <localLabels> <string>Input External</string> </localLabels> <shapekey>68256A86-B456-4555-B6C4-33EE0B87CE71</shapekey> <screenX>16</screenX> <screenY>210</screenY></node> <arc> <toConnector>0</toConnector> <fromConnector>3</fromConnector> <curveStyle>Default</curveStyle> <name>RME</name> <localLabels> <string>RME</string> </localLabels> <From>electric_motor</From> <To>Output_External</To> <directed>true</directed> <doublyDirected>false</doublyDirected></arc>


This approach provides a flexible infrastructure that can accommodate different tools for each component (e.g., concept generation, clustering, and visualization). The development of a common schema also provides a consistent information flow between the automated concept generator, design repository and clustering analysis, reducing the likelihood of data loss due to translation between different data formats. Although the incorporation of a schema is crucial to managing the information flow, it is necessary to setup a module for integrating and validating data before it is presented to the designer.

3.2 Data Integration and ValidationThe ability to integrate results from automated concept

generators with information available in design repositories and other relevant projects would provide a synergetic impact on the designer’s decision making. However, in order to accomplish that, a robust structure needs to be in place. The architecture developed in this work bridges available components and translates the data into a meaningful representation to the designer. The process is divided into two stages: data validation and data integration.

Data validation ensures that the framework operates on correct, clean and useful data. Each component (e.g., the concept generation engine, the clustering routines) was tested for compliance with the specified schema on a unit basis before being incorporated into the framework. This approach would also be used for incorporating any future component of the framework. The first stage of data validation is a comparison between data stream and the schema definition document. This comparison is performed by the SimpleXML parser available in the PHP web-scripting language [37] used in the framework. The second stage of validation ensures that the data contained within the XML stream is consistent with the lexicon developed by Kurtoglu et al. [40]. This step is completed using a set of PHP algorithms that compare data contained within the XML with tables stored in a MySQL database [44]. The third level of data validation is a check for completeness (i. e., Does each element in the functional model have an assigned component or components). A permissive approach is used, where any data validation problems are highlighted for presentation to the designer because complete designs may not be necessary to stimulate innovative design concepts. Another advantage of this approach is that it identifies opportunities for innovation

in a product where no prior solution has been developed for a particular functional requirement.

Once the information is accepted, the data integration stage parses and integrates the data into a database that establishes the linkage between them. This bridging not only connects independent sources of information but allows data storage for rapid retrieval by multiple designers. An independent database has been setup in order to accommodate the information as well as ease the further use of data in order to generate meaningful representation to the designer.

3.3 Generation of Concept Exploration InterfaceAs shown in Figure 1, the framework relies on a web-

server approach in order to provide the designer with the ability to explore the concept space. The interface takes advantage of increasing computational and graphical resources to provide a better understanding of the topology of the generated solutions by enabling various visualization representations to be displayed to a designer. The designer is guided through a representation of the functional model, candidates are grouped together using the results of the clustering algorithms, and detailed representations of candidates of interest are made available to a designer. The underlying assumption is that designers can interact more intuitively with graphical, rather than textual, representations of design candidates. This can be seen as an analogy to browsing through music albums using the cover art on an iPod, rather than using only the textual description of the album. Instead of recalling to mind familiar albums, the representations are constructed to intuitively represent the topology and flow of material and energy through a product.

Creating the visual representations uses the integrated information from the automatically generated concepts and the digital design repository data to retrieve archetypal images of a component’s type (e.g., an image of a flat gear is used to represent any gear). These images do not change frequently over time and are retrieved from the digital design repository and cached on the web server used for the visualization interface.

Transforming the graph theory description of a concept into a human readable representation is a non trivial task when the placement of nodes and routing of arcs is considered. A computer-based graph layout toolkit, GraphViz [45], has been incorporated into the representation creation process. This layout tool is used to generate multiple representations of each concept, each using different cues for conveying semantic information for nodes and arcs. After retrieving the graph representation of the functional model or the conceptual model, a text-based input file is generated and GraphViz is executed using a system command in PHP. The result of the GraphViz code is an image that is subsequently cached on the server to speed future requests. Each graph can take two to three seconds to generate. With the potential for thousands or tens of thousands of design candidates, images are cached on the web server for rapid access. This approach will limit the impact of latency on the exploration of the concept space as discussed in [46]. A separate algorithm

Figure 3. Example Segment of Clustering Message.

<cluster clusterid=”4” parentclusterid=”2”> <member>candidate001.xml</member> <member>candidate006.xml</member> <member>candidate011.xml</member></cluster><cluster clusterid=”5” parentclusterid=”2”> <member>candidate002.xml</member> <member>candidate007.xml</member> <member>candidate012.xml</member></cluster>


was developed to generate Design Structure Matrix (DSM) [47] representations of the concept’s topology.

The next section illustrates the presented methodology using the newly developed web-based interface along with an existing concept generator and design repository for a given design problem.

4 IMPLEMENTATION AND EXAMPLEThe interface is implemented for a sample design problem

demonstrating the capabilities of the approach by creating visual representations of a design candidate. The sample peanut sheller problem, described in Figure 4, was used as part of an ongoing research investigation into the impact that information technology tools have on designer creativity in creating a novel design. The design problem starts with a functional model, a pools of automatically generated design alternatives, and the allocation of design concepts into clusters.

For the purposes of this paper, a web-enabled platform is used to present the results. The overall designer work process is presented using the flowchart in Figure 5. The designer is guided in a manner that mimics the progression from conceptual to detailed design using the interface. The user first authenticates into the interface and then chooses a design problem of interest. The concept exploration interface shown in Figure 6 is divided into three distinct regions on the webpage:

• The left menu allowing the designer to make selections • The top tabs allowing the designer to change between

different display formats when available, and• The main frame where information are displayed.

Before navigating a set of candidate designs, the user first selects a problem of interest in the interface, and a set of possible functional models are provided for review. Currently, a set of functional models are already populated, but ideally, the user would upload their own. In Figure 6, the selection of

Figure 6. Web Interface for Concept Exploration Framework.

Figure 5. Designer Workflow for Exploring Automatically Generated Design Concepts.

Design Problem

In places like Haiti and certain West African countries, peanuts are a significant crop. Most peanut farmers shell their peanuts by hand, an inefficient and labor-intensive process. The goal of this project is to design and build a low-cost, easy to manufacture peanut shelling machine that will increase the productivity of the African peanut farmers. The target throughput is approximately 50 kg (110 lb) per hour.

Goals:• Must remove the shell with minimal damage to the

peanuts.• Electrical outlets are not available as a power source.• A large quantity of peanuts must be quickly shelled.• Low cost• Easy to manufacture

Figure 4. Peanut Sheller Design Problem.


(a) Textual Version (b) Graphical Version

Figure 8. Design Structure Matrix Representations of the Functional Model.

a functional model brings up the preliminary exploration tabs allowing the visualization of the proposed functional model. A graphical representation of the functional model is automatically generated using the DOT automatic graph layout program [45] and is shown with tabs allowing the user to change the presentation of the information between the default graph view and both textual (i.e., matrix of 0’s and 1’s) and graphical Design Structure Matrix (DSM) representations.

The default view of the functional model is illustrated in Figure 7. The automatically generated graph provides basic information about the functional model including the different functions between the input and output together with the different flows linking them together. Each graph starts with an input at the top and ends with an output at the bottom of the graph. The different flows are color and arrow coded in order to ease the differentiation between flow types. Function and flow names are also provided in order to identify them accordingly. The designer is also able to switch the representation to a different format. Figure 8 shows the remaining two formats available through the interface. The user has the ability to choose different representations, including a matrix-based Design Structure Matrix (DSM) (shown in Figure 8a) or a graphical DSM illustration (shown in Figure 8b). The graphical DSM representation provides additional help for the designer by quickly showing the connectivity’s between the different functions.

Selecting a clustering scheme allows the interface to present the candidates grouped into different clusters to the designer. Two clustering approaches were used in evaluating the architecture. In the first, the number of concepts is reduced prior to clustering. In the second, the number of dimensions used to describe the concepts is reduced by identifying the principal components before clustering. The first option, discussed in detail in [38] Figure 7. Functional Model of a Peanut Sheller.


Figure 9. Text-Based Version of the Configuration Flow Graph.

uses a variant K-means clustering or Lloyd’s algorithm as an iterative clustering algorithm.[21, 48] A second clustering option, discussed in detail in [39] uses Principal Component Analysis (PCA) to reduce the dimensionality of the problem prior to applying the clustering algorithm[49] .

After inspecting the functional model, the designer then proceeds to select a clustering scheme (e.g., k-means) in order to explore the automatically generated candidates that would satisfy the functional model. The example problem has seven clusters, each containing seven candidates, for a total of 49 generated candidates. The user is then presented with a set of design candidates that belong to the active cluster. For illustration purposes, a single candidate (i.e., candidate 6) is used to further demonstrate the concept exploration interface.

Once a candidate has been selected by the designer, the functions that model the design problem are replaced by different component configurations for each available candidate. The interface replaces the previous functional model graph from Figure 7 with an updated one where the functions and flows are replaced by components that would satisfy the function and flow combinations. A new set of tabs are available for the designer to choose from. Again, the default view is the functional graph with textual information.

The designer can now inspect the selected candidate’s functional model (candidate 6) and quickly indentify which function (or set of functions) has been replaced by which component. Figure 9 presents the configuration flow graph for candidate 6 using textual descriptions of components. Functions from the original functional model in Figure 7 that have not been replaced by a component are highlighted using a different background color. Figure 10 shows how the representation can also be changed to replace the text nodes in the configuration flow graph with images of archetypal components. Textual and graphical DSM representations are also available for the selected candidate. The presented methodology that is translated into a highly organized database provides a novel way of linking various information sources into a powerful exploration tool that enables data integration in order to support the decision making process of a designer. The approach illustrated can be used with any other concept generators or design repositories to provide even more solutions to the designer to explore.

5 CONCLUSIONThe proposed methodology is an initial step in order to

provide a methodology for concept space exploration using IT components. By interconnecting various tools for the support of conceptual design, the architecture allows the designer to address their respective shortcomings in terms of usability and also presents a new way to explore the multitude of concepts in a novel way. By developing a web server interface, the various data from different IT-based tools are validated and integrated using a component naming taxonomy and a common schema. The information is linked and stored in a database that allows rapid access to the wealth of information that results from

concept generators and design repositories. Furthermore, the current web-server interface allows the designer to use the exploration interface through a browser, handheld device or any device that supports web data.

The presented work illustrates a single concept generator linked to a design repository. The approach methodology is easily scalable to multiple concept generators linked with multiple design repositories and additional components like clustering methods in order to enable concept exploration. This is made possible by using flexible and extensible communication schemas. Designers are provided with multiple representations of product concepts using a computational tool for automated graph layout.

6 FUTURE WORKAs of now, the approach provides a foundational methodology

for future research involving the various IT components. Future work will be focused in two parts: the technical aspect of the infrastructure and the theoretical ramifications that could


emerge from this research. The technical aspect of future work will include developing various web-service approaches for the interface in order to allow more flexibility in accessing the information available. Designers could use those services to create trade space representations of the available design candidates. Validation could also be offloaded to external algorithms to check for redundant designs or dominated solutions, further refining the set of available solutions.

From a theoretical standpoint, the platform can be used as a foundation for concept exploration impact on the designer. The inclusion of various methodologies for current or future concept clustering like mixed-initiative reasoning [50] will be used to study the impact of information density on designers’ ability to explore the concept space efficiently. In the evaluation of impact of IT tools on engineering creativity, the proposed approach in conjunction with some sort of innovation metric to reproduce innovative products available on the market, can be used as a validation tool

ACKNOWLEDGMENTSThis work is part of a larger research effort into the impact

of IT tools on designer creativity. The authors would like to express gratitude to Matthew Campbell, Julie Linsey, Daniel A. McAdams, Seth Orsborn, and Robert Stone for their efforts in supporting this research. We are grateful to the National Science Foundation (IIS-0841389 – Univ. at Buffalo - SUNY) and the New York State Office of Science, Technology, and Academic Research, Grant #M050100 for their support of this work.

REFERENCES[1] Murdock, J. W., Szykman, S., Sriram, S. D., 1997. An

Information Modeling Framework to Support Design Databases And Repositories.

[2] Bohm, M.R., R.B. Stone, T.W. Simpson, and E.D. Steva,” Introduction of a data schema to support a design repository”, Computer Aided Design.Vol. 40, No. 7, 2008, pp. 801-811.

[3] Devendorf, M., Lewis, K., Simpson, T. W., Stone, R. B. and Regli, W. C., 2007, “Evaluating the Use of Cyberinfrastructure in the Classroom to Enhance Product Dissection,” Proc. ASME 2007 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference, Las Vegas, NV, DETC2007-35549.

[4] Atkins, D. E., Droegemeier, K. K., Feldman, S. I., Garcia-Molina, H., Klein, M. L., Messerschmitt, D. G., Messina, P., Ostriker, J. P. and Wright, M. H., 2003, “Revolutionizing Science and Engineering Through Cyberinfrastructure,” Report of the National Science Foundation Blue-Ribbon Advisory Panel on Cyberinfrastructure, National Science Foundation, Arlington, VA.

[5] Fox, E., Akscyn, R., Furuta, R. and Legget, J., 1995, “Digital Libraries,” Communication of the ACM, 38(4), pp. 23-28.

[6] National Science Digital Library http://www.nsdl.org[7] National Engineering Education Delivery System http:// Figure 10. Image-Based Version of the Configuration Flow Graph.


www.needs.org[8] Zhang, S., Shen, W., and Ghenniwa, H., 2004, “A Review

of Internet-based Product Information Sharing and Visualization,” Computers in Industry, 54(1), pp. 1-15.

[9] Fincher, C., 1986, “Trends and Issues in Curricular Development in Higher Education,” Handbook of Theory and Research, J. Smart, ed., Agathon, New York, pp. 275-308.

[10] Stone, R. B., and Wood, K. L., 2000, “Development of a Functional Basis for Design,” Journal of Mechanical Design, 122(4), pp. 359-370.

[11] Bohm, M. R., Stone, R. B., and Szykman, S., 2005, “Enhancing Virtual Product Representations for Advanced Design Repository Systems,” ASME Journal of Computer and Information Science in Engineering, 5(4), pp. 360-372.

[12] Bryant, C., Mcadams, D., Stone, R., Kurtoglu, T., and Campbell, M.I., 2005, “A Computational Technique for Concept Generation,” Proc. ASME 2005 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference, Long Beach, CA, DETC2005/DTM-85323

[13] Bryant, C., Stone, R., Mcadams, D., Kurtoglu, T., and Campbell, M., 2005, “Concept Generation from the Functional Basis of Design,” Proc. International Conference on Engineering Design, ICED ‘05, Melbourne, Australia

[14] Nicolai, L. M., 1995, “Design a Better Engineer,” Aerospace America, 30(4), pp. 30-33.

[15] Kopena, J. B., Regli, W. C., 2003, “Design Repositories on The Semantic Web with Description-Logic Enabled Services” First International Semantic Web and Databases Workshop, Berlin, Germany.

[16] Hirtz, J., Stone, R. B., Mcadams, D. A., Szykman, S., and Wood, K. L., 2001, “A Functional Basis for Engineering Design: Reconciling and Evolving Previous Efforts,” Research in engineering design, 13(2), pp. 65-65..

[17] Shooter, S. B., Simpson, T. W., Kumaray, S. R. T., Stone, R. B., and Terpenny, J. P., 2005, “Toward a Multi-Agent Information Infrastructure for Product Family Planning and Mass Customization,” International Journal of Mass Customization, 1(1), pp. 134-155.

[18] Kurtoglu, T., and Campbell, M. I., 2009, “Automated Synthesis of Electromechanical Design Configurations from Empirical Analysis of Function to Form Mapping,” Journal of Engineering Design, 20(1), pp. 83-104.

[19] Bohm, M., 2009, “Information Archival and Reuse: Drawing Conclusions from the Past,” Missouri University of Science and Technology.

[20] Kurtoglu, T., and Campbell, M. I., 2007, “Exploring the Worth of Automatically Generated Design Alternatives Based on Designer Preferences,” Proc. International Conference on Engineering Design, ICED’07, Paris, France.

[21] Backer, E., 1995, Computer-Assisted Reasoning in Cluster Analysis, Prentice Hall, New York.

[22] Anderberg, M. R., 1982, Cluster Analysis for Applications,

Academic Press, New York.[23] Ames, A. L., Nadeau, D. R., and Moreland, J. L., 1997, The

Vrml 2.0 Sourcebook, Wiley, New York; Chichester, Eng.[24] Bryson, S., and Levit, C., 1992, “The Virtual Wind Tunnel,”

IEEE Computer Graphics and Applications, 12(4), pp. 25-34.

[25] Cruz-Neira, C., Sandin, D. J., and Defanti, T. A., 1993, “Surround-Screen Projection-Based Virtual Reality: The Design and Implementation of the Cave,” Computer Graphics, 27(SPI), pp. 135-135.

[26] Geist, G. A., Kohl, J. A., and Papadopoulos, P. M., 1997, “Cumulvs: Providing Fault Tolerance, Visualization, and Steering of Parallel Applications,” The International Journal of Supercomputer Applications and High Performance Computing, 11(3), pp. 224-235.

[27] Price, A. M., 1998, “Virtual Product Development ‘Case Study of the T-45 Horizontal Stabilator’,” Proc. 39th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference and Exhibit, Long Beach, CA, A98-25247.

[28] Winer, E. H., and Bloebaum, C. L., 2002, “Development of Visual Design Steering as an Aid in Large-Scale Multidisciplinary Design Optimization. Part I: Method Development,” Structural and Multidisciplinary Optimization, 23(6), pp. 412-412.

[29] Eddy, J., and Lewis, K., 2002, “Visualization of Multi-Dimensional Design and Optimization Data Using Cloud Visualization,” Proc. 2002 ASME Design Technical Conferences, Design Automation Conference, Montreal, Canada, DETC02/DAC-02006.

[30] Mattson, C., Mullur, A., and Messac, A., 2004, “Smart Pareto Filter: Obtaining a Minimal Representation of Multiobjective Design Space,” Engineering Optimization, 36(6), pp. 721-740.

[31] Ribarsky, W., Ayers, E., Eble, J., and Mukherjea, S., 1994, “Glyphmaker: Creating Customized Visualizations of Complex Data,” IEEE Transactions on Computing, 27(7), pp. 57-64.

[32] Balling, R., 1999, “Design by Shopping: A New Paradigm.,” Proc. Third World Congress of Structural and Multidisciplinary Optimization, Buffalo, NY, pp. 295-297.

[33] Stump, G., Lego, S., Yukish, M., Simpson, T. W., and Donndelinger, J. A., 2009, “Visual Steering Commands for Trade Space Exploration: User-Guided Sampling with Example,” ASME Journal of Computing and Information Science in Engineering.

[34] Price, A. M., 1998, “Virtual Product Development: Case Study of the T-45 Horizontal Stabilator,” 39th AIAA/ASME/ASCE/AHS/ASC Structures, Strucutral Dynamics, Materials Conference, Long Beach, CA, A98-25247

[35] PTC Pro/ENGINEER http://www.ptc.com/products/proengineer/

[36] SolidWorks http://www.solidworks.com/ [37] Lerdorf, R., and Tatroe, K., 2002, Programming Php,


O’Reilly, Beijing; Sebastopol, CA.[38] Kumar, M. and Campbell, MI, 2010, “Organizing a Design

Space of Disparate Component Topologies”. Design Cognition and Computing Conference, Stuttgart Germany, Paper #157

[39] Poppa, Kerry, Stone, Robert and Orsborn, Seth, “Exploring Automated Concept Generator Output Through Principal Component Analysis”

[40] Kurtoglu, T., M. Campbell, C. Bryant, R. Stone and D. McAdams, 2008, “A Component Taxonomy as a Framework for Computational Design Synthesis,” Journal of Computers and Information Science in Engineering, Accepted, to appear.

[41] World Wide Web Consortium, (2006) Extensible Markup Language, http://www.w3.org/TR/REC-xml/

[42] Szykman, S., Racz, J. W., and Sriram, R. D., 1999, “The Representation of Function in Computer-Based Design,” Proc. 1999 ASME Design Technical Conferences, Design Automation Conference, Las Vegas, NV, DETC99/DTM-8742.

[43] GraphSynth http://graphsynth.com/ [44] MySQL AB, 2002, MySQL reference manual. MySQL AB.

http://www.mysql.com/doc/en/.[45] Gansner, E. R. and North, S. C., 1999, “An Open Graph

Visualization System and its Applications to Software Engineering,” Software Practice and Experience, 00(S1), pp. 1-5.

[46] Simpson, T. W., Barron, K., Rothrock, L., Frecker, M., Barton, R. R. and Ligetti, C., 2007, “Graphical User Interfaces: for Engineering Design: Impact of Response Delay and Training on User Performance”, Research in Engineering Design, 18(2) pp. 49-65.

[47] Steward, Donald V. 1981. “The Design Structure System: A Method for Managing the Design of Complex Systems.” IEEE Transactions on Engineering Management 28: 71-74.

[48] Arthur, D., and Vassilvitskii, S., 2007, “K-Means++: The Advantages of Careful Seeding,” Proc. Eighteenth Annual ACM-SIAM Symposium on Discrete Algorithms, pp. 1027-1035.

[49] Orsborn, S., Boatwright, P., and Cagan, J., 2008, “Identifying Product Shape Relationships Using Principal Component Analysis,” Research in Engineering Design, 18(4), pp. 163-180

[50] Tecuci, G., Boicu, M., & Cox, M. 2007 Jun 15. Seven Aspects of Mixed-Initiative Reasoning: An Introduction to this Special Issue on Mixed-Initiative Assistants. AI Magazine. 28:2.

exploring automatically generated design...

Documents