Digital modeling and AI-driven optimization of ball burnishing process parameters

Abstract. In this paper, a digital model of the ball burnishing process, developed using data from tests on bronze and aluminum alloys, to create regular micro-reliefs on a cylindrical surface." is presented. The experiments were carried out on a CNC lathe using a cylindrical stock. The input features were burnishing force, feedrate, machine condition and material type, while the results were surface roughness and hardness. A Bayesian approach was used to infer the parameters of the digital model, enabling the creation of distributions that directly account for uncertainty caused to regression coefficients and by random factors incorporated in machine condition feature. The proposed digital model was subsequently used to run simulations and check various optimization algorithms. Additionally, a large language model NexusRaven-V2 that were fine tunned for function calling is combined with the functions from the digital model.