Automating the Path from Data to Decision: Generating and Adapting Simulations with Gen-AI

Sammanfattning

The application of material flow simulations to optimize production lines can significantlyimprove the efficiency of manufacturing companies, but it can be time and resource consuming.Generative artificial intelligence has gained much relevance in recent years and can improvethe efficiency of these simulations and assist them in exploring a wider range of solutions. This thesis explores the potential of applying generative artificial intelligence to automate andimprove the process of optimizing a production line using material flow simulations. Relevantartificial intelligence methods are studied and potential improvement areas in the developmentprocess are identified. Potential applications in the process include: natural language-drivenSQL query generation, simulation adaptation and XML code generation. The generative artificialintelligence enhanced implementations are applied on real production data from amanufacturing firm in the Swedish automotive sector. A holistic approach that creates, analyzes, adapts, and evaluates a simulation model of aproduction line is created, thereby automating this process, and allowing scenario generationand testing. This approach provides a framework for future extensions in detail and input asmore capable technologies become available.