In this study, a predictive approach was developed to estimate the hardness profile and hardened depths in 4340 steel spur gears subjected to laser surface heat treatment. First, a three-dimensional numerical model was developed in COMSOL Multiphysics to simulate the thermal and metallurgical behavior during laser heating. The model incorporated heat transfer equations and phase transformation kinetics to compute the resulting hardness profiles. Experimental validation was performed using a 3 kW Nd:YAG laser and a Taguchi-designed test matrix, confirming strong agreement between simulated and measured hardened depths, with errors below 6%. Statistical analysis using analysis of variance (ANOVA) was then applied to identify the most influential process parameters and their interactions. This analysis guided the design of a simulation campaign that systematically varied laser power, scanning speed, and rotational speed to build a robust data foundation. Based on this hybrid dataset—combining simulated and experimental results—a multilayer perceptron artificial neural network was trained to predict the hardened depths at the tooth tip and root. The resulting model achieved high predictive accuracy, with R2 values above 95% and prediction errors under 10%. The surface hardness values obtained after laser treatment ranged from 32 HRC to 55 HRC, and the maximum hardened depth reached 2.45 mm at the tooth tip, confirming the model’s robustness. This integrated modeling strategy offers an efficient and reliable tool for optimizing laser surface hardening in components with complex geometries.
Seghier et al. (Sat,) studied this question.