Data Modeling for System Design

Abstract

Software systems are often composed of components and subsystems from diverse, independently developed sources. Integration and interoperability require correct interpretation of data exchanged between those components, which can result in increased development efforts and a risk of system failure due to data misinterpretation, especially as systems scale. Many have reached for data modeling as a tool to support systems integration by providing a common semantics and structure for documenting exchanged data, which can be reasoned over to determine the interoperability of software components. The Future Airborne Capability Environment (FACE) and Fast Healthcare Interoperability Resources (FHIR) are standards in military avionics and healthcare that exemplify this approach. Each relies on data modeling to aid in the integration task, but key limitations prevent automation, ultimately placing the burden of data model reasoning on a human, which is error-prone and does not scale. This dissertation examines these limitations and builds on existing research in model-driven engineering, conceptual data modeling languages, ontologies, and healthcare interoperability to propose solutions for automating the integration task. First, an OWL-based data modeling language (“DMSDL”) is presented, providing well-defined data compatibility semantics. Second, analysis techniques are proposed to automate incompatibility detection at scale. Finally, model transformations from FACE and FHIR are developed to demonstrate automation in those standards. These contributions are implemented in a proof-of-concept software tool (“pydmsd”), which provides the foundation for both a FACE Model Compatibility Tool and a FHIR Profile Interoperability Tool.