The detection and elimination of deviations play a decisive role in shop floor management. This not only makes processes continuously more stable and efficient, but also improves employees' skills. The deviation management process generates a large amount of data (deviation descriptions, causes, measures) that represents the valuable knowledge of employees. By documenting this data, this knowledge can be processed and distributed, turning valuable experience into a sustainable resource. This promotes fast and efficient deviation management. Large Language Models (LLMs) have great potential in the area of production. By integrating the knowledge, questions can be asked via digital conversations with a ChatBot, which facilitates and accelerates deviation detection and the problem-solving process.

The aim of the project is to investigate how LLMs can be used to support shop floor professionals in deviation management. The following work packages are part of the thesis: research on the topics of digital shop floor management & deviation management , research on the topics of LLMs, differentiation of various models and their applications in production, development of methods to improve the performance of an LLM, testing and evaluation of these use cases with company data Documentation of the results.