Comparison of Response Surface Methodology (RSM) and Machine Learning Algorithms in Predicting Tensile Strength and Surface Roughness of AA8090/B4C Surface Composites Fabricated by Friction Stir Processing

Abstract

Friction stir processing is an innovative solid-state process, widely utilized for surface composite fabrication, material property enhancement, and microstructural modification. Rotational speed, traverse speed, groove width, and axial force are key FSP parameters that improve the characteristics of surface composites (SCs). This work makes use of FSP to fabricate AA8090/B<inf>4</inf>C SCs by altering parameters within ranges. Response variables include ultimate tensile strength (UTS) and surface roughness (SR). Central composite design (CCD) of response surface methodology (RSM) leads trials, establishing a mathematical relationship between input parameters and UTS/SR. The models’ adequacy is validated using ANOVA, which investigates the impact of input parameters on UTS and SR. This study also looks into machine learning regression methodologies for UTS and SR forecasting in AA8090/B<inf>4</inf>C SCs. The ML algorithms are evaluated by utilizing performance metrics like coefficient of determination (R2) and root mean squared error (RMSE). Predicted UTS and SR values from RSM are compared with machine learning outcomes. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.