Fatigue behavior prediction by Finite-Element-trained Data Mining models

During the design process and when a mechanical component fails, fatigue analysis theories are applied. The existing theories are based on the material properties and the applied amount of load. The intrinsic inaccuracy of the theories in addition to the precision when properties are determined, makes it difficult to obtain an exact life value. Another problem when using fatigue theories is the exact fracture point determination. Usually the applied load and the material properties data are known, but not with precision enough to give good results. Data Mining models can help to find the proper values to be used during the component improvement. This article sets out the methodology to be followed in the training of Data Mining models. A comparison between the selected techniques is done in order to obtain the most suitable to this application.