Machine learning-based prediction of surface quality and tool performance in the grinding of inconel 800

Abstract

Abstract This study investigates the surface grinding behavior of Inconel 800, a nickel-based superalloy widely used in high-temperature applications. Grinding tests were performed using green silicon carbide and aluminium oxide wheels under constant parameters: 2800 RPM spindle speed, 0.1 mm depth of cut, and 3.5 mm feed rate, with bio-based coolant. Surface roughness was monitored after every two passes, along with corresponding thermal imaging and wheel surface analysis. Results showed that the green silicon carbide wheel maintained better thermal stability and wear resistance, with surface roughness rising from <0.3 µm to >0.85 µm by the 22 nd pass. In contrast, the aluminium oxide wheel delivered a finer initial finish but wore more rapidly due to heat buildup. Manually annotated particle accumulation data enabled the development of machine learning models for tool wear prediction, with Random Forest Regression achieving the highest accuracy (R 2 >0.9). The findings highlight the effectiveness of combining thermal and surface data with predictive modeling to optimize grinding performance and tool life in machining Inconel 800.