MATHEMATICAL MODEL DESIGN TO PREDICT THE FATIGUE LIFE BEHAVIOR FOR BEARING STEELS

Abstract

Conventional and ultrasonic fatigue testing is usually conducted under axial or rotating- bending loading. In spite of the differences in shape and design of the fatigue specimens, the stressed volume plays a mutual role in assessing the fatigue life. This is due to the fact that the probability of finding voids or inclusions, which are the sources of cracks, increases as the stressed volume increases. The fatigue life of bearing steel generally extends to the Giga cycle regime so conducting such fatigue testing in the laboratory is time consuming and costly. In this work, a model based on neural network techniques has been suggested to predict the fatigue strength of bearing steel. The input data for this model includes hardness, the stressed volume, stress ratio and number of cycles. The model captures reasonable trends and is able to estimate unseen experimental results on high strength bearing steel AISI 52100. Extrapolation has been conducted for the rolling contact fatigue life and the results show good agreement with the experimental data.