This conference paper was submitted for presentation at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

In the evolving landscape of system modeling and simulation, the integration of Knowledge Engineering and Generative AI offers unprecedented opportunities for efficiency and accuracy. This presentation explores the transformative potential of Knowledge Graph technology and Generative AI for recommendation-based creation processes of system models. Case studies and practical examples are used to illustrate how integrating knowledge graphs with AI not only enhances and expedites model creation but also ensures that these models adhere to rigorous engineering standards. Knowledge graphs leverage curated engineering knowledge in formal descriptions and serve as a robust foundation for building system models. By capturing and interlinking diverse pieces of information, knowledge graphs enable users to seamlessly access relevant data and insights, facilitate the generation of precise model recommendations and streamline the modelling process and thereby reduce the time and effort needed to construct detailed and accurate simulations. Looking ahead, we envision a future where AI plays an integral role in system modeling and simulation. To harness the power of AI effectively, we must ensure that the data it consumes is derived from a formal, organized source'?Knowledge Graphs. Approaches such as Graph RAG (Graph Retrieval Augmented Generation) are particularly pertinent in this context, as they facilitate the extraction and structuring of relevant data, essential for informed AI-driven decision-making. A critical aspect of employing knowledge graphs in system modeling is the validation of this knowledge against established engineering rules. This ensures that the insights derived from the graph are reliable and conform to industry standards. In the beginning we will outline the basic concepts of knowledge graph technology and its application in system modeling. Furthermore, we will demonstrate how AI, when fed with structured data from knowledge graphs, can enhance the modeling process, offering recommendations that are both accurate and efficient. Then our presentation will delve into two main validation aspects. The first is knowledge validation during the curation phase, where rule-based techniques are used to verify the integrity and accuracy of the knowledge being incorporated into the graph, cross-checking against known engineering principles. The second is response validation within the GenAI architecture, which rigorously checks generated outputs against predefined engineering rules and standards before presenting them to users, ensuring that only validated and reliable information reaches the end-users. Finally, we will discuss the implications of this approach for future developments in system simulation, emphasizing the role of formal data structures in enabling sophisticated AI applications, supported by robust knowledge and response validation processes to maintain accuracy and reliability.