J10.4
Augmenting SADT with UML, a hybrid approach for the design of operational science algorithms

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 26 January 2011: 4:45 PM
Augmenting SADT with UML, a hybrid approach for the design of operational science algorithms
607 (Washington State Convention Center)
John L. Baldwin, AER, Lexington, MA; and A. Werbos and T. S. Zaccheo
Manuscript (616.5 kB)

Science algorithms perform calculations for modeling physical phenomena and estimating physical properties. They take input and control data to generate output data. Implemented in software, science algorithms are functional in nature and generally not user interactive. As such they can be designed as modular computational units which may be run independently, as part of batch processing, or as components within a user or service interactive environment. Two contemporary approaches to address the analysis and design of such systems are Structured Analysis/Structured Design and Object Oriented Analysis and Design. Structured Analysis and Design Technique (SADT) is a specific methodology which emerged out of the variety of different structured analysis and structured design approaches. SADT was designed to simplify and provide a consistent model to describe hierarchical systems. Unified Modeling Language (UML) is a specific object oriented design methodology, serving a similar role to object oriented systems which SADT provides for hierarchical systems. Collaborative computing and ever larger and more complex operational scientific software systems create greater demands on software and systems engineers to reduce risk and provide greater features while working to keep costs as low as possible. Augmenting UML with SADT provides more in-depth algorithm analysis and functional decomposition, while still retaining the strong object model that facilitates large-scale system design.

This paper proposes a hybrid structured analysis and object oriented design approach using SADT with UML toward developing science algorithms and documenting legacy algorithms for incorporation into operational systems and environments. A case study example is included, with specific issues and challenges discussed.