Investigation of a Vibratory Stress Relief System for Hard-Turned EN 31 Components

This study examines the use of artificial neural networks (ANNs) to forecast and optimize residual stress and Brinell hardness in EN 31 components subjected to vibratory stress relief (VSR). The influence of important process parameters – amplitude, frequency, and time – was determined through comprehensive ANOVA analyses. According to the findings, residual stress and Brinell hardness are substantially influenced by amplitude, while frequency plays a crucial role in managing stress and hardness before VSR. The significance of time varied across different processes. The ANN model consistently demonstrated high predictive accuracy, achieving 99.82% for Brinell hardness after VSR, 98.27% for residual stress after VSR, 99.98% for Brinell hardness before VSR, and 98.20% for residual stress before VSR. Model performance was further improved through data transformation and normalization. A robust framework for optimizing VSR process parameters was established by integrating ANOVA and ANN, which enabled precise control over mechanical properties. This research emphasizes the potential of ANN in predictive modeling and process optimization in materials engineering, providing valuable insights for enhancing the performance and reliability of mechanical components through customized VSR processes. Keywords: vibratory stress relief (VSR); residual stress; Brinell hardness; artificial neural network