Prediction of frictional behaviour through regression equations: A statistical modelling approach validated with machine learning

Abstract

Friction in machinery and transportation has played a significant energy challenge, with the reducing friction and wear can improve the performance of engines, motors and vehicles.Therefore, this study involves fabrication, characterization, mechanical and tribological analysis of metal matrix composites (MMCs). Further, this study has developed of a statistical model to predict the frictional behaviour of Zinc-Aluminium in-situ composites reinforced with Zirconium diboride (ZrB₂) particles. The alloy and composite were fabricated using liquid metallurgy, resulting in altered physical and morphological properties. Reinforcement with ZrB₂ led to significant improvements in mechanical and tribological properties, resulting in a notably low coefficient of friction (COF) with lubrication, even under higher loads, signifying reduced energy losses. Response Surface Methodology (RSM) using Design Expert 13 software was used to analyze the relation between input parameters which included applied normal load, sliding distance, and ZrB₂ percentage and output parameter (COF). A quadratic model was recommended as the best fit for the present material system and a regression equation was developed which predicts the frictional behaviour with more than 94% accuracy at given input variables. Moreover, ANOVA highlighted load as the primary influencer on COF characteristics. An Artificial Neural Network (ANN) in PYTHON validated RSM-recommended optimized COF parameters, enabling reliable COF predictions. A comparison of experimental and statistical outcomes from RSM and ANN confirms the model's strength in anticipating the tribological traits of alloys and composites.