Physic-informed machine learning for centerless grinding optimization

Centerless grinding is a machining process characterized by highly nonlinear dynamics, large model uncertainty and considerable sensitivity to process parameters. Guaranteeing quality of the worked parts, basically roundness, and short processing time by an optimal process setup is typically a specialized and time-consuming task. In this work, an approach is presented for centerless grinding parameters setup and optimization based on physic-informed machine learning techniques. The real data for model training are provided both as measurements on the ground workpieces and as on-line monitoring signals. The developed functionalities concur in implementing the concept of "intelligent grinding machine", proposed by Monzesi srl, an innovative SME operating in machine tools sector.