Domain knowledge-enhanced bottom-up extraction of feature models from system structure models in SysML

Model-based Product Line Engineering (MBPLE) is the combination of Model-based Systems Engineering (MBSE) and Product Line Engineering (PLE) to maximise the reusability of existing system design solutions by differentiating their commonalities and variabilities. The Feature Model (FM) is the core asset in MBPLE to characterise the variants from externally visible properties. However, with the increasing variability of products and growing system complexity, the construction of FM that heavily relies on manual efforts has become a labor-intensive and expertise-intensive task for practitioners. To address this issue, a bottom-up approach for the automatic extraction of FM from SysML structure models is proposed in this study. First, the traceability between features and their implementation model elements in the input SysML structure models is established through element partitioning and feature traceability detection. Subsequently, Formal Concept Analysis (FCA) is applied to identify dependencies among features. Finally, logical reasoning combined with semantic knowledge enables the automatic identification of feature types, feature groups, and hierarchical relationships, resulting in a well-defined and semantically consistent FM. A case study of a vehicle system including 20 distinct variants is used to illustrate the effectiveness of the proposed approach.