Variability Mining in IEC 61499

Managing variability is essential for reuse and maintenance in IEC 61499 systems. However, variability of existing systems can only be managed systematically by identifying, localizing, and presenting commonalities and differences between variants.
Despite their relevance, IEC 61499 systems lack tooling for such variability mining. The central question addressed in this work is how to effectively extract variabilities from existing IEC 61499 variants to support engineering tasks.
Prior research has explored variability mining in software systems. However, most approaches focus on textual code and overlook the domain-specific characteristics of IEC 61499. Existing methods tailored to IEC 61499 either focus only on clone detection or do not allow for localization of found variabilities.
This thesis aims to fill this gap by proposing a variability mining method tailored to IEC 61499 and then enhancing an existing approach using Natural Language Processing (NLP). Additionally, we explore the integration of Large Language Models (LLMs) to support interpretation tasks.
The development of our approach was informed by 14 interviews with IEC 61499 practitioners from academia and industry. Based on these findings, we present a pipeline consisting of (1) preprocessing IEC 61499-related elements, (2) matrix-based mining, (3) grouping through string similarity metrics, and (4) enhancement with LLM-based feature descriptions.
Our evaluation shows that matrix-based mining is effective for detecting reused elements, particularly in cloned applications. Grouping similar names through NLP techniques improves clarity, and initial experiments with LLMs show potential for assisting in understanding features.