부산대학교 도서관

A Method for Detecting Design Error in the UML Models Using OntologyJin-won SeoDepartment of Computer EngineeringGraduate School, Kongju National UniversityRepublic of Korea(Supervised by Professor Chi-Su Kim) Improving software quality continues to be a critical issue in software development. Various stakeholders view the system models from different perspectives. As a result, initial software specifications were quite often incomplete, but have evolved and progressed as software has become more developed. But as this development progresses, the cost and effort of fixing errors also increases. As a result, early detection of inconsistency is important for successful software development projects. However, few tools and techniques exist which apply to formal inconsistency detection techniques at the software design level. Two common paradigms for software development are the object-oriented and agent-oriented methodologies. The agent-oriented methodology evolved via a merging of the object-oriented methodology and artificial intelligence. The excellence of object-oriented methodology has been demonstrated in various fields such as programming, analysis and design, database, artificial intelligence and so on. But the object-oriented methodology shows its weaknesses, namely: insufficient abstraction and support for object interaction, bottom-up design methods and the creation of large and generic libraries. Presently software design performed using UML has become the real standard. The major benefits of using UML are the various modeling notations designed to encourage considerable freedom of specification. These notations enable designers to specify partially overlapping views of the system. However, this flexibility often introduces inconsistencies into a software design. Unfortunately, UML does not have precisely unified semantics to clearly specify and verify consistency. In this thesis, COADE (Conversion from Object to Agent fot the Detection of design Errors) is suggested for detecting inconsistency error due to the multiple views of software design at an early stage. To detect inconsistency error in software design, COADE utilizes a common ontology based model, OSSD (Ontology for Software Specification and Design), which integrates object-oriented and agent-oriented concepts. By extracting structure, data, association relationship from UML design and transforming these into an OSSD model following transformations into KAOS specifications using mapping tables, inconsistency error of the design can be detected. In the transformation process to an OSSD model which detects inconsistency error by combination of detection methods between views and consistency framework related rules, vocabulary and semantic analyses utilizing multiple mapping tables are included. An elevator case study is used for an explanation of the inconsistency error detecting process and evaluated from 3 different viewpoints. By comparing it with a generally agreed upon sound ontological model of BWW Model, an ontology represented by OSSD was evaluated, error detection was assessed and transformation from UML into KAOS was estimated. COADE's error detection capability did successfully detect design error and the transformation of a set of semantically significant features was successfully performed from source design to target specification. By combinations of multiple views of software design and by providing consistency error detection methods for software, this thesis has contributed to the improvement of software design quality.