Discovering and documenting potential failures and irregular user behavior that can interrupt the normal system behaviour is very important during the development of critical systems. Failure Mode and Effects Analysis (FMEA) is a bottom-up inductive analysis method that helps to identify potential failure modes based on experience with similar products and processes. Model-Driven Development (MDD) is a software development paradigm that raises the level of abstraction of software development by changing the focus from code to models and automating code generation from models. MDD also eases the derivation of analysis models for different software non-functional properties in the early stage of software development. The objective of this thesis is to develop a model transformation process that takes as input a UML software model with failure mode annotations and generates a FMEA model. The transformation is developed in Epsilon, a new family of languages specialized in model transformations.