Title: Bidirectional Text-to-Model Element Requirement Transformation
Abstract: Elicitation, representation, and analysis of requirements are important tasks performed early in the systems engineering process. This remains true with the adoption of Model-Based Systems Engineering (MBSE) methodologies. Existing SysML-based methodologies often choose between (i) using external requirements documents and/or databases as the authoritative source for requirements truth versus (ii) generating requirements directly, as elements in the system model. In either case, there is often need for the systems engineer to manually develop a model-based requirements representation, as this faculty is not automatic in the commonly-used SysML feature set. Additionally, once the system model has been completed, systems engineers typically must prepare traditional “shall-statement” requirements for external review purposes, as not all stakeholders can be expected to be trained in system model interpretation. This paper details a novel effort to address both problems, by automatically transforming text-based requirements (TBR) into SysML model-based requirement (MBR) representations, and vice versa. The text-to-model based transformation direction uses requirement templates and natural language processing techniques, expanding on work from the field of requirements engineering. This paper also presents an aerospace-domain case study application of the developed tool. In the case study, a selected set of requirements were analyzed, and a system model was constructed. Then, the intermediate output system model was updated with additional elements, to represent the progression of the project's systems engineering process. The modified system model was then analyzed, constructing text-based requirements from the structure. The resulting text-based requirements were compared to the initial set of input requirements to assess consistency in both directions of analysis. The methodology developed in this paper improves the systems engineering process by saving the systems engineer time constructing potentially repetitive model elements, and by enabling model-based requirement analyses to methodologies previously only capable of processing text-based requirements. Further, the methodology eases the responsibility of the systems engineer to maintain a copy of the model-based requirements in text-based format.