Prediction model for fatigue life considering microstructures of steel

In fatigue life of structure, crack initiation and crack propagation is considered separately. Behavior of large crack propagation is explained on the Paris equation. However, there is no model to simulate the behavior from the crack initiation to the large crack propagation. One of this cause is that fatigue life can varies greatly thanks to material microstructures and manufacturing and change of stress. Especially to think the effect of material microstructures is important in materials development. However, there is no model considering quantitative effect of material microstructures. Another cause is that there is no clear difference between crack initiation and crack propagation. So we made the first prediction model for fatigue life considering material microstructure. This model is for the  ferrite-pearlite, most popular steel for the structure. This model needs distribution of particle size and stress, Poisson`s ratio, elastic modulus to estimate fatigue life. First, using FEM analysis, this model gets how much stress is on the test piece. Second, this model divides surface of the test piece into small squares, and fills the squares with particles using Monte Carlo method based on distribution of the particle size. Third, the model gives each particles crystal orientation randomly. In each squares, from the stress and the crystal orientation, this model judges the initiation of the crack on particle. This particle is crack nucleation. Then, this model simulates the propagation of small crack from the crack nucleation. From the stress and the crystal orientation and the interaction between the slip band and the grain boundary, this model simulates the propagation of the small crack thinking the interaction with grain boundaries. This model continues this simulation till crack will be 1mm long. To estimate exact fatigue life, this model needs the parameter of the small crack propagation. This parameter is inherent to each material. We did some fatigue experiments and observed the propagation of the crack in detail to estimate the parameter. And we did fatigue experiments with another distribution of particle size to validate our model, and we got good agreement with the experiments and the model.