Prediction Accuracy of Fatigue Life under Random Loading

Abstract

The measured fatigue life of mild steel under random loading is compared with the fatigue lives predicted in the time and frequency domains, and the factors that reduce the accuracy of fatigue life prediction in the frequency domain is identified in this study. Several random stress and strain histories for the comparison of the fatigue lives were generated by conducting the numerical analysis of an artificial structure. To obtain the measured fatigue life, the smooth and notched round bar specimens made of mild steel were tested so that the generated random stress or strain histories could occur in the cross section of the specimens. In addition to the fatigue tests, the fatigue lives of the specimens were predicted in the time domain using the rainflow cycle counting method and the Palmgren-Miner cumulative damage rule given the SN curve of mild steel and the generated random histories. The fatigue lives were also predicted in the frequency domain using Dirlik’s method with the SN curve and the random PSDs that were generated using fast Fourier transformation of the random histories or the numerical analysis. As the result of the comparison, the relation between the measured fatigue life and the fatigue life predicted in the frequency domain was changed depending on the crest factor of the random stress and strain histories. Since this tendency of the fatigue life prediction in the frequency domain is explained by considering the feature of the PSD and the cycle counting method used in the frequency domain, the fatigue life under random loading can be accurately predicted in the frequency domain by adjusting the stress or strain range distribution obtained using Dirlik’s method.