The versatility of gas turbines makes them the key to the energy transition, as they are able to meet various requirements. They are ideal for dealing with peak loads and effectively supplement renewable energy sources at times of low production.However, due to the many start and stop processes in this area of application, the components are exposed to high mechanical and thermal loads. Accurate modelling of the probability of failure is essential for safe operation and optimization of maintenance intervals.Nickel-based superalloys, which are commonly used in turbine components, have complex grain structures that influence fatigue life. In this work, we use the grain structure to create a percolation model. The crack initiation times of a single grain are based on the random orientations of the crystal lattice. The influence of a cracked grain on the surrounding material is modeled with finite element simulations. These simulations are used to create an infection function for the percolation model.The resulting epidemiological crack percolation model is calibrated and tested with experimental LCF data. For the transfer to large components, the crack initiation process is modeled as a spatio-temporal point process on the surface. This results in a post-processor for FE simulations with local cumulative hazard rates.