Position Paper: Decision-Based Design

Venkat N. Rajan
Industrial and Manufacturing Engineering Department
Wichita State University
Wichita, KS 67260-0035

Decision-Based Design (DBD) brings a new perspective to the product design and manufacture activities. It represents the concept oviewing the design process as a series of decisions. The decisions may be made by an individual design engineer, or collaboratively by a group of design engineers or an integrated product development team. The decision-makers may be co-located or distributed. DBD removes the overwhelming emphasis on the ARTIFACT and transfers some of it to the DECISION. The ARTIFACT is still the deliverable, the result of the series of DECISIONS. For certain types of analyses, the focus on the artifact is appropriate. Examples include stress and thermal analyses. But, how does one analyze the risk and uncertainty associated with an artifact? The risk and uncertainty are associated with the decision that resulted in the selection of the particular design artifact.

Risk and uncertainty are some of the issues that can be better analyzed and handled by focusing on the decision. Other advantages include the promise of reducing the number of design iterations, and the ability to integrate manufacturing analysis issues throughout the design process. If we accept that the design process involves a series of decisions, then we see that analyzing the ARTIFACT will result in modifying or reversing a previously made decision. Thus, the design engineer makes a decision, analyzes the artifact, and revises or reverses the decision. Why not focus on the DECISION up front? By focusing on the decision, it seems possible to concurrently design and analyze and therefore, reduce the number of iterations. An example, is the reduction of part count. Large number of parts result in increased effort during the design process but also impact many downstream functions adversely. Reduced part count is a desirable objective provided we can keep the cost down. If we view the product decomposition activity as a decision process, then by questioning the decomposition decision, we can effectively control the part count. Unless the design engineer or team is able to provide sufficient justification for the decomposition, it is not accepted. Acceptable justifications stem from functional requirements, manufacturing cost concerns, or assembly problems. Other issues that can be included are company culture, vendor capabilities, etc.

Design engineers find it difficult to reuse designs. Some reasons include lack of history including alternate designs that have been explored, non-parametric designs, and lack of understanding of the design rationale. These issues have resulted in the development of design history/rationale/intent capturing systems. Most of these systems are passive. They do not evaluate the justifications provided by the design engineer. However, the history/intent/rationale lies with the DECISION not with the ARTIFACT. Why did the design engineer make the particular choice among the various available alternatives? The true value of a design history/rationale/intent system lies in its ability to provide the basis for a DECISION made by the design engineer. By doing so, it can not only improve design reuse, but can also guide the design engineers towards better designs.

How do we handle the many personal preferences and conflicts that arise during the design process? However objective we may believe the design process to be, personal preferences of the design engineers do get reflected in the decisions they make and, ultimately, in the artifact they design. Conflicts in preferences, resource requirements, etc., do play a role in determining the final result. Taking a decision-based design view of the process can potentially allow us to model these preferences and conflicts in an effective manner and provide insight into the reasons why and how certain designs result.

I believe that Decision-Based Design is a new paradigm of design. It definitely affects all aspects and phases of design and manufacturing. It has the potential to provide a new direction for the development of tools and techniques, and promises the ability to effectively integrate risk and uncertainty, manufacturing, and other concerns with the design process. There are many open issues related to representations and methodologies. What is an appropriate representation for the decisions and how is it linked with the product representation scheme? What is an acceptable justification for a design decision? What are the various approaches to decision-making under risk and uncertainty and how can they be used during the various phases of design? How can we model the preferences and conflicts that arise during the design process?

Background

In my previous research, I have proposed new concepts related to DBD [1]. I have used Game Theory concepts in manufacturing problems [2] in which local and global objectives of the players (preferences and system objectives) are considered. I believe that Game Theory has a lot to offer for modeling preferences and conflicts in the DBD process [3]. I have also developed representations and methodologies for integrated assembly design, analysis, and planning [4, 5]. I believe that DBD provides a new way to address conceptual design where limited geometry is available. It is also during this stage that the assembly design gets defined and therefore, the representations of the decision and the assembly are closely related, as are the methodologies used in DBD and assembly modeling.

References

1. Rajan, V. N., A Framework for Concurrent Assembly Design and Planning, to appear in the Proceedings of the First Annual International Conference on Industrial Engineering: Applications and Practice, Houston, TX, December 4-7, 1996.

2. Rajan, V. N., Cooperation Requirement Planning for Multirobot Assembly Cells, Ph.D. Dissertation, Purdue University, W. Lafayette, IN, May 1993.

3. Rajan, V. N., An Agent-Based Fractal Model of Agile Manufacturing Enterprises: Modeling and Decision-Making Issues, Proceedings of the AAAI Research Planning Workshop on AI and Manufacturing, Albuquerque, NM, June 24-26, 1996.

4. Lyons, K. W., V. N. Rajan, and R. Sreerangam, Representations and Methodologies for Assembly Modeling, Internal Report (first draft), National Institute of Standards and Technology, Gaithersburg, MD, August 1996.

5. Rajan, V. N., K. W. Lyons, and R. Sreerangam, Requirements and Representations for Assembly Modeling, Submitted to the Fourth ACM Symposium on Solid Modeling and Applications, Atlanta, GA, May 14-16, 1997.