The goal of this thesis is to develop a software framework for injecting various fault patterns during the operation of neural networks, enabling the study of their behavior in fault-inducing environments, such as the impact of space radiation on satellites. The project will compare existing approaches from the literature and create a dynamic framework for future investigations in different applications.