Prediction of surface residual stress and hardness induced by ball burnishing through neural networks

Abstract

Ball burnishing is a mechanical surface treatment used for surface finish improvement, surface work hardening and inducement of compressive residual stresses, nevertheless, a great level of interaction is observed among the most relevant factors. Within this scenario, artificial neural networks can be employed to determine the most recommended input parameters in order to achieve the required outcome. In this work, burnishing tests were performed using annealed and hardened AISI 1060 steel specimens and the obtained surface residual stress and hardness values were used to train an artificial neural network. The experimental results showed a nonlinear relationship between the input and output parameters for annealed AISI 1060 steel and support the applicability of artificial neural networks for the burnishing process, whereas a more linear relationship between the input and output parameters was observed for hardened AISI 1060 steel, though burnishing pressure seems to be the most relevant factor affecting residual stress. The artificial neural network and optimisation procedure provided consistent input parameters, thus leading to the inducement of compressive residual stress of higher intensity. [Submitted 29 November 2017; Accepted 26 May 2018]