Hybrid machine learning model-based approach for Intelligent Grinding

Current manufacturing evolution places a growing demand on autonomous control and optimization of manufacturing processes, especially for unattended machines. In grinding, operator surveillance is still fundamental to cope with strong process variability, introduced by modifications in tool characteristics, machine dynamics and workpiece material. Surface quality of the machined workpiece can be affected by different mechanisms: in cylindrical grinding system resonances are typically involved, while in center-less grinding the geometrical configuration plays a paramount role. The goal of this study is to develop SW tools to support production, both off-line during the planning and setup phases, and on-line, by monitoring signals related to the performed grinding process. Looking for industrial applications, the proposed solutions must assure an adequate robustness against system variability and disturbances: to cope with such demanding objectives, a proper mixture of physics-based and black-box models are proposed. Examples are shown in roll grinding and in center-less grinding.