Development and validation of a probabilistic model for notch fatigue strength prediction of tool steels based on surface defects

A new model for the high cycle notch fatigue strength prediction of tool steels subjected to axial loading is proposed, based on previous literatures studies and experimental tests carried out on six different tool steels, including rotating bending fatigue tests on notched specimens, fractographic analyses, hardness, residual stress, and roughness measurements. The novelty is the assumption that surface defects are the main cause of notch fatigue failures of such steels. A probabilistic approach was implemented by modeling size distributions of defects, resulting in the prediction of normal distributions of fatigue strength. Like to other previous models, the effect of steel hardness, surface residual stress, notch severity, and specimen size was also taken into account. Model calibration and validation were performed using the data collected by the experimental activity. Model behavior was investigated by performing a sensitivity analysis, aiming to verify the response to variations of the considered input variables. Prediction errors of only 1.3% (on average) and 3.1% (maximum) resulted from the comparison between model-predicted and experimental notch fatigue strength.