ETD

• artificial intelligence
• EN ISO 13849-1 standard
• industrial machinery safety
• large language models
• retrieval-augmented generation

In the field of industrial machinery safety, software tools for reliability assessment and compliance with functional safety standards (e.g., EN ISO 13849-1) play a key role in the design and validation of safety functions. Among these tools, SISTEMA, developed by IFA, supports designers in the evaluation of safety-related control systems. However, the outputs generated by such tools, often provided in structured formats such as XML, are typically complex and not directly suitable for advanced analysis or automated processing.

This thesis presents the design and implementation of the Safety Function Design Platform (SFDP), a modular web-based application developed to complement SISTEMA by transforming its project and library data into structured, interpretable, and traceable representations suitable for interactive analysis. The application supports the import and visualisation of SISTEMA projects, the deterministic identification of safety-related design issues, the recommendation of potential design alternatives, and component matching and cross-library querying.

Artificial intelligence (AI) capabilities support contextual explanations, technical report generation, design recommendations, and natural-language component search. The thesis also investigates language models of different sizes, with particular attention to compact models and knowledge augmentation strategies. The results highlight the importance of external domain knowledge in improving the reliability of AI-assisted engineering tasks.

Overall, the proposed approach aims to improve the readability, consistency, and traceability of safety-related information while preserving the semantic content of the original models and supporting engineering analysis and decision-making.