Despite the fact that deep rolling is being performed in crankshafts for the past sixty years, there is no consolidated understanding on its residual stress generation process. Researchers have treated the subject in a purely experimental way, without an attempt for a phenomenological explanation for the observations. Little information was found on the evolution throughout the processes inside the crankshaft manufacturing chain. The few authors that tried to model the deep rolling process through a numerical tool were still not able to consolidate a methodology for results assessment integrated to crankshaft fatigue performance. With that stated, the hypothesis of this research is that the residual stress distribution due to deep rolling parameters variation added to the operational loads perceptibly alters crankshaft fatigue performance. The objective, therefore, is the understanding and prediction of the influence of the deep rolling process on the endurance behavior of crankshafts. In order to accomplish that, the research followed an approach based on numerical modeling of the process together with the execution of fatigue experiments. That way, a comprehension based on the fundamental phenomenon can be addressed. Residual stresses were measured in order to verify the model and enable its use for a wide range of parameter alteration. The correlation between fatigue and numerical results is also imperative to push the knowledge boundaries on the topic. The developed model was satisfactorily verified through comparison with residual stress measurements. Models with diverse deep rolling parameters were generated and clear tendencies regarding residual stress profile shape and magnitude were outlined. Specimen batches with the same evaluated deep rolling parameters were produced for further residual stress measurements and fatigue experimentation. Their results' interconnection shed light over crankshaft crack nucleation and propagation. A correlation between numerical model and fatigue results could be drawn. The comparison of simulation contour graphs with the crack site fractography gave encouraging indication that the developed model went through the right path. The expectation was to develop a model capable of being implemented in an industrial application environment. To accomplish that, accurate residual stress generation prediction, regardless of the process conditions, and understanding its correlation with the fatigue behavior of crankshafts were paramount.