Prediction of coefficient of friction in friction stir welding and its implementation in a thermo-mechanical model

Friction stir welding (FSW) faces a critical challenge in numerical modelling due to the widespread assumption of a constant coefficient of friction ( µ ), which leads to inaccuracies in predicting thermal profiles and heat generation dynamics. This study addresses this gap by developing a temperature-dependent µ model to enhance the fidelity of thermo-mechanical simulations—a vital step for optimizing weld quality and process efficiency in industries like aerospace and automotive. Experiments employed an AA6061-T6 aluminium alloy and 100 Cr steel ball in a pin-on-disc tribometer under controlled conditions (load: 2–8 N, sliding speed: 30–120 mm/s, temperature: 50–300 °C) using an L16 orthogonal array. Regression analysis via Minitab 17 revealed that µ ranged from 0.415 to 0.682, with temperature contributing 49% of the variability, followed by load (30%) and sliding speed (7%). A second-order polynomial equation ( R 2 = 0.9886) quantified µ as a function of temperature, load, and speed, which was integrated into a modified thermo-mechanical model. Compared to constant- µ models, the revised model improved heating rate predictions by 23%, achieving <5% error in peak temperature estimates during experimental validation. Cooling rate disparities (up to 15%) highlighted the need for refined boundary condition modelling. The work's novelty lies in its first-of-its-kind empirical regression framework for µ in FSW, bridging a critical gap between tribological experimentation and thermo-mechanical simulations. By replacing static µ assumptions with dynamic, parameter-driven values, this approach advances predictive accuracy in FSW modelling beyond prior literature, which predominantly relied on fixed µ ranges (e.g., 0.25–0.6) without systemic parameter interactions.