Search

• Fatigue life prediction method of 25CrMo4 steel is investigated using fracture mechanics and machine learning methods. • Residual stress relaxation process is incorporated into fatigue crack growth in combination with NASGRO equation and equivalent initial flaw size (EIFS) methodology. • Different machine learning methods areemployed to investigate theireffectiveness in fatigue life prediction. In this work, fracture mechanics and machine learning methods are separately employed to predict the high cycle fatigue (HCF) life of micro-shot peened 25CrMo4 alloy steel containing artificial notches. The influences of notch size, stress level, and surface condition on fatigue life are incorporated into the prediction model. The results shows that the fatigue life of shot-peened specimens is enhanced by the compressive residual stress (CRS) compared with that of the un-peened specimens. The prediction result using fracture mechanics method is within the error band of ±2, while that using machine learning method almost lies in the error band of ±1.5. The stress level and notch size are the most critical factors affecting the HCF fatigue life. The applicability and limitation of the two methods are finally discussed. [ABSTRACT FROM AUTHOR]