This research addresses the complexities inherent in grinding operations, aiming to identify the most effective process parameters and predict the behavior of the workpiece surface condition. This task is particularly challenging due to the difficulty of achieving smooth surfaces and complex, nonlinear interactions between input factors such as grinding conditions and cooling system type, and output factors such as surface roughness. These challenges are further intensified when considering additional elements, such as grinding wheel wear and advanced cryogenic lubricants or coolants. To address these issues, this study advances beyond traditional modeling methods, such as General Linear Regression or Random Forest models, to explore novel distributional modeling techniques, including General Additive Models for Shape and Scale and Distributional Random Forest. These advanced models are designed to elucidate the intricate connections between input factors and their corresponding outputs, mainly focusing on predicting the distribution of surface roughness profiles. The enhanced accuracy of these models (predictive error decreasing at around 7% and 24%) is instrumental in determining the most effective process parameters. These models offer deeper insights into the interdependencies in grinding operations, enabling more precise process control. Additionally, these models shed light on the potential improvements in surface profile quality achieved by implementing cryogenic techniques, opening new paths for optimization in grinding operations.
Abedrabbo et al. (Tue,) studied this question.