Prediction of Micro-scale Forces in Dry Grinding Process Through a FEM—ML Hybrid Approach

Grinding process modeling represents a great challenge due to its stochastic nature. The uncertainty factor of grinding technology is mainly attributable to the undefined grain morphology, with the influence of this aspect becoming more pronounced in a dry configuration. Even though grinding has always used lubricants, nowadays the reduction or complete elimination of this element could mean a significant reduction in environmental pollution. Many modeling approaches have been used in literature to investigate phenomena related to grinding but each exhibits some disadvantages. In this paper a hybrid FEM—ML approach is proposed to forecast forces generated by the action of a single grain in dry conditions, overcoming the main modeling limitations observed to date. Experiments and force measurements were performed on a CNC surface grinding machine using sintered aluminum oxide grains of size 60. FEM simulations were developed in DEFORM 3D to predict grinding forces and increase the data set. ML algorithms were proposed to increase model prediction productivity and optimize the control of process parameters.