A statistical and neural network approach to investigate the tribological behavior of ZA-27 alloy

This study presents a comparative analysis of the tribological behavior of ZA-27 alloy using both statistical and neural network methodologies. The research aims to develop predictive models for the tribological behavior of ZA-27 alloy, with a focus on wear rate and coefficient of friction (COF). Response Surface Methodology (RSM) and Artificial Neural Network (ANN) approaches are employed to analyze and predict the tribological properties of the alloy. The Central Composite Design (CCD) method of RSM is utilized to conduct wear studies and evaluate the impact of key parameters such as manganese (Mn) content, normal load, sliding speed, and sliding distance. Optimization based on a desirability function is performed, and the ANN model is subsequently employed to validate the optimized parameters. Results from both methodologies demonstrate a high level of agreement between observed and predicted values for wear rate and COF, highlighting the effectiveness of both statistical and neural network approaches in modeling the tribological behavior of ZA-27 alloy. The performance of the ANN prediction model surpasses that of the RSM model with a correlation coefficient (R²) of 99.76% for wear rate and 99.37% for COF, while the RSM model achieves 92.06% and 92.02% correspondingly. Oxidative wear, laminative wear, and adhesive wear mechanisms were observed for Mn contents of ZA-27/0.2%, ZA-27/0.5%, and ZA-27/1.0%, respectively. Therefore, integrating ANN for predicting the wear rate and COF of ZA-27 alloy in manufacturing processes could lead to substantial reductions in manufacturing time, effort, and costs.