Surrogate-assisted multi-objective robust optimization of helical gear pair considering manufacturing uncertainty due to profile error

Durable, efficient, and low-noise gears are critical to meet real-world performance requirements. In this context, the variations in performance indices due to manufacturing errors must be accounted for to successfully optimize real-world gear systems. This study presents a robust multi-objective optimization methodology for helical gear pairs that accounts for profile errors. The study focused on analyzing key performance indicators, specifically peak-to-peak static transmission error and efficiency, using both an analytical model and a machine learning-based elastohydrodynamic lubrication solver, respectively. Surrogate models were employed to expedite Monte Carlo simulations over the uncertainty space. Notably, the Kriging-based Monte Carlo method demonstrated a significant reduction in sensitivity to profile errors, resulting in a 50% reduction in standard deviations for PPTE and efficiency compared with deterministic optimization. The application of surrogate models effectively accelerated simulations while maintaining high predictive accuracy. Overall, the proposed optimization approach proved effective in developing robust gear designs that mitigate the impact of profile errors.