First you need to understand how fatigue cracks in real materials advance through the microstructure, and that is complex. See http://youtu.be/yiUrhwTsbd4.
i am working on fatigue crack propagtion too. i use ansys and i have generated o code but in montonic loading by using X-FEM so i trie do to cyclic loading.
i can send to you the code. and i am pleased to work together.
It sounds like you expect ANSYS to show you the crack progression through the material as a sequential parting of the geometry. I dont think the standard elements will allow that to happen. Where it may be useful is to predict the point of crack instigation, the circumstances for which you would need to extablish through empirical tests if the data is not there already. The exact crack path will be a funciton of the grain boundary position in the material so can only be generalised by high stress areas and cross section reduciton.