The aim of this thesis is to develop a model for probability-based event detection and analysis in maintenance management using Bayesian Networks (BN). To ensure a smooth start to the work, the basics of maintenance are summarized by means of a literature research. Subsequently, the topic of Key Performance Indicator is treated by not only explaining it theoretically, but also explaining the calculation in detail. This is needed for building the model later. Afterwards, the topic BN is highlighted. In the beginning, the basics are described to explain the current state of the art of BN for showing the current limits of the application of these networks. The practical applications of BN in terms of production and maintenance, as well as the state of the art, are reflected in a SWOT analysis, which summarizes the current possibilities in this field. For the suitability check of a BN on the topic of predictive maintenance, data from a production process is used, which are prepared with the help of data analysis and evaluation. Based on the data, a BN is modeled in two different processes. A manual modeling by the user creates a BN, which is subsequently extended to a dynamic BN. In addition, software helps to create a BN automatically. Finally, a comparison between manual and automatic modeling highlights the advantages and disadvantages of each approach.