This study aims to develop a Gauss–Newton iterative model that predicts the wear depth of AISI 52100 steel by linking surface roughness evolution to the lubricant film parameter ∧ across different lubricant viscosities.
The model exploits the systematic dependence of the roughness decay rate and steady-state roughness on time, which integrates load, lubricant rheology and contact geometry. Nonlinear exponential functions describing disk surface roughness during running-in are fitted using the Gauss–Newton algorithm. Ball-on-disk tests are performed under 30 N, 100 N and 500 N with PAO8, PAO25 and PAO100 lubricants, and a 3D optical profiler measures the roughness of both balls and disks.
Disk roughness decreases exponentially with time. Higher viscosity yields a larger film parameter, leading to lower steady-state roughness and a slower decay rate. The model achieves excellent agreement with PAO25 and PAO100 data (MSE < 0.01). Calibrated exclusively on PAO25 and PAO100, it successfully predicts the entire wear evolution of PAO8 – a viscosity grade not used in training – yielding a predicted wear depth of 11.48 µm, which agrees with the measured 12.17 µm within a 6.6% relative error.
By integrating roughness with film parameter, the unified approach predicts wear performance across lubrication regimes without retraining, offering a reliable theoretical and methodological basis for service life assessment of mechanical systems.
The peer review history for this article is available at: https://publons.com/publon/10.1108/ILT-04-2026-0187/
