Tom DeNucci
The complexities of ship design continue to increase. This can be attributed to the em- phasis on multi-mission vessels, the interaction between sophisticated shipboard systems and equipment and an increase in the number of design requirements. At the same time, the knowledge used to resolve these complexities is decreasing due to a loss of intellec- tual resources, e.g., the retirement of experienced workers, retention barriers with current employees and the recruitment of promising talent and skill sets. In order to remedy these losses, Naval Architects must capitalize on the knowledge available in both commercial and government design teams.
The solution to this quandary involves the capture of Design Rationale (DR). Design rationale embodies the reasoning and justification behind design decisions. This not only includes a description of the intent behind original design decisions, but can also describe designer preferences in the design. There are many potential applications of design rationale in Naval Architecture including design evaluation, design verification (design thresholds and gates) and communication of design decisions to stake holders. In its most basic form, design rationale also serves as a tangible artifact of institutional memory. In the end, design rationale can add substantial value to both the conceptual design process of complex ships as well as the resulting product (the ship); most import- antly, it must be captured before it disappears!
SUMMARY
Although design rationale appears in nearly all aspects of the design process, it would be particularly valuable during the configuration design of complex vessels. In general arrangements, the decision knowledge required to identify the relationships, i.e., interac- tions, between objects in the design is often tacit, qualitative and not explicitly available. As a result, this type of rationale is rarely captured, i.e., elicited, structured and stored for future (re)use or reference. The expression of this rationale is implicit in the design process, i.e., only the consequences are currently captured. The underlying reasoning is not expressed, which makes its formalization, storage, transfer and (re)use impossible.
The work presented in this dissertation tackles these deficiencies. The goal of this re- search was to develop a methodology to capture the rationale behind object interactions in conceptual ship design. The solution consists of four task-oriented modules integrated within a Rationale Capture Tool (RCT): elicitation of design rationale, structure of design rationale, storage of design rationale and a methodology to expand the rationale database.
A new methodology was developed to “trigger” the expression of configuration rationale: Reactive Knowledge Capturing (RKC). The premise of RKC involves the reaction of a designer to a set of feasible ship configurations. Since it has been observed that it is easier for designers to identify what they don’t like in a design, as opposed to what they do like, the approach hinges on a collection of unconventional and unorthodox designs. From this set, the Naval Architect identifies both the desirable and undesirable interactions together with the underlying rationale. For each interaction, the rationale consists of a proximity preference and its associated justification.
A proof-of-concept experiment demonstrates the validity of this approach and offers additional insight on rationale structure and classification. The RKC approach is suc- cessful because it provides the designers with a design review environment that promotes reflection and supports the expression of rationale. The computer-based implementation of the RKC approach is based on an existing program to generate feasible ship configur- ations.
To support the transfer and (re)usability of rationale expressed using the RKC meth- odology, it is structured using a modified version of Decision Representation Language (DRL). The argument-based semi-formal representation includes the following elements: decision problem, alternatives, claims, goals and assumptions. A template-based implementation of the DRL elements allow for minimal user in- terruption and automatic generation of hierarchical relationships between the elements. Natural language description, in conjunction with text-based inequalities, support both human and machine comprehension of rationale. Other key features of this represent- ation include error checking (inferencing), the capture of dependency relationships and the use of a standardized vocabulary suite, which adds a layer of uniformity and consist- ency to the expressed rationale.
Following structuring, the object interactions and associated rationale are stored in a MySQL relational database. The structure and format of this medium allow for rapid access and retrieval of information. Standard database I/O functionality also allows for easy porting of rationale to other software applications. In addition, rationales are also concurrently stored in a Design Structure Matrix (DSM) to support interactive design generation and the ientification of decision de- pendencies. Both storage formats also support the sorting and indexing of the expressed rationale.
A dedicated feedback mechanism was also developed to expand the rationale database. An optimization-based approach is used to interactively steer a NSGA-II genetic algorithm to design solutions where design rationale is absent, masked or hidden. This approach provides a continuous method for expanding and refining the database and im- proves the overall efficiency of the RCT.
A comprehensive test case was performed to evaluate the Rationale Capture Tool in the context of a realistic design environment. Sixteen Naval Architects used the tool in prac- tice to capture over 400 unique object interactions and their associated rationales. The test case showed that the approach embodied in the RCT is able to capture a wide-breadth of high quality configuration rationale. Moreover, results from a post-test questionnaire indicate the tool is highly usable, easy to learn, and generally beneficial to designers.
Lastly, rationales collected in the test case were (re)used to illustrate possible applications of rationale capturing in ship design; examples include the generation of “improved” designs, visualization of the follow-on consequences of configuration decisions and the assessment of organizational quality by means of rationale analysis.
The Rationale Capture Tool developed in this dissertation is a welcome addition to the ship design process. During design development, the tool can be used to capture the rationale behind object interactions. In addition, previously captured rationales can be applied as constraints for generating new arrangements. During the review process, the tool can be used as a substitute for formal design meetings. Post design, the tool can be used to capture lessons learned. The RKC methodology might also have potential for capturing rationale in other stages of the design process.
Ultimately, this approach developed in this research improves the quantity of know- ledge available in the early stages of the design process. This not only helps preserve the industry’s fleeting knowledge, but also provides a foundation for making improved decisions in the ship design process.