Experimental Modeling and optimization of surface quality and energy consumption in the roller burnishing process of AISI 1045 steel using intelligent models

Roller burnishing is being extensively employed for improving surface quality and saving manufacturing cost, while its parameters' optimization, especially the trade-off between surface roughness and power consumption, still faces difficulties. The purpose of this research is to predict and optimize the process on AISI 1045 steel through the RSM approach and a hybrid PSO-ANN model. The results show that the burnishing force (F) has the maximum influence on surface roughness (50.9%), followed by spindle speed (N), while the latter has its maximum influence on power consumption (92.37%). When the F increases, the surface finishes better with increasing power consumption. The PSO-ANN model has high prediction accuracy (R² = 0.9911 for Ra, 0.9987 for Pb) and outperforms the response surface method. MOALO and MOGWO algorithms also produce close optimum parameters, where MOALO promotes power efficiency. The optimal parameters yielded by MOALO are N = 300 rpm, f = 0.11 mm/rev, and F = 129.84 N. These parameters yield a burnishing power of 0.5041kW and roughness of 1.42 µm.