An elastic-plastic finite element analysis using ANSYS is performed to examine closure behaviour of fatigue cracks under various loadings and the numerical results are then compared with experimental results. Several FORTRAN and UNIX shell programs are written to perform fatigue crack growth analysis with 4-node iso-parametric element using this commercial program.
The effects of several parameters such as crack advance schemes, stress states etc. are investigated and then compared with the results published by others. The results by ANSYS are shown to give similar trends and assumed to be reliable.
The ratio of the element size to the monotonic plastic zone size, $Δa/ω_p$, is used as a criterion to determine the most appropriate mesh size to provide good numerical results. It is shown that the mesh size has to be changed continuously as the maximum stress intensity factor $K_max$ increases both under plane stress and under plane strain states. The opening levels under variable and random loadings can also be predicted well using the same element size determined under constant amplitude loadings only under plane stress, where the nodes behind the crack-tip node are used to determine the opening level, but can not be predicted under plane strain, where the opening level is so low that the crack tip node is used.
For random loading, it turns out that the nearly constant crack opening points observed in the experiments are due to the nearly constant crack length during a random block loading. The numerical results using simplified random loadings, which are rewritten for narrow and wide band random loading with considering the shape of the original load history respectively, agree well with experimental data.
Also a new method so-called fixed crack length method, where the applied load is changed to increase $K_max$ while the crack length is held constant, is introduced. The value of $Δa/ω_p$ is a function of $K_max$ in the load constant method, but it is constant in the fixed crack length method.
