Abstract:
Appropriate computational techniques to model laser shock peening to retrieve residual stress results using finite element modeling are largely under investigation. Hence, this original study investigated a numerical modeling technique employing a quasi-static analysis (using ABAQUS CAE) to deliberately introduce approximation into the analysis by ignoring the inertia normally present in a conventional dynamic explicit laser shock peening analysis (which has good prediction capability of residual stresses) in order to reduce computational time. To compensate for the modeling approximation, the input pressure load spatial profile parameters were varied in an attempt to obtain similar surface residual stress results to those of a dynamic explicit analysis. Such a study has not previously been conducted. The outcome of the study showed that a quasi-static analysis accelerates the simulation time up to 29 times. In addition, when machine learning with a partial least squares algorithm (employing the PyCharm Python integrated development environment) was used on the quasi-static input spatial profile and output surface residual stress data sets (obtained from performing trial and error in order to match the dynamic explicit results closely), the quasi-static input spatial results were obtained in sub one second, aiding with computational efficiency. Applying the predicted profile to the quasi-static model resulted in the surface residual stress profiles correlating well for the first 0.5 mm, with a 7.88% difference to the dynamic explicit model and a 3.27% difference to the experimental result. However, this explorative study demonstrated that the quasi-static modeling technique produced unrealistic results when the other principal stress components are compared to the dynamic explicit model. It becomes clear that inertial effects are essential in modeling laser shock peening realistically. Therefore, when stress components other than the surface residual stress profile in one plane are required, the conventional dynamic explicit analysis technique should be employed.
