Optimization of the surface quality of brittle-hard materials in CNC grinding processes based on vibration and topography analyses and the use of machine learning

CNC-controlled machining processes have become essential in modern optics production, driven by enhanced precision and reproducibility. However, escalating demands for component quality necessitate ongoing optimization of process stability and efficiency. The selection of parameters crucial for high-quality outcomes still relies heavily on the expertise of machine operators. This study focuses on the real-time investigation and recording of process vibrations during CNC grinding, combined with an objective analysis and control of their influence on the surface quality of optical components. Using Polytec's high-resolution optical measurement technology, inline vibrations were measured with Laser Doppler Vibrometry, while two coherence scanning interferometers were used for areal non-contact characterization of the surface topography. The primary objective was to detect process vibrations and their dependence on different grinding parameters to draw conclusions about resulting surface qualities. Extensive process and component data were collected, incorporating surface metrology parameters (Ra, Rq) and applying the power spectral density (PSD) function for surface quality characterization. Insights gained into vibration development within the CNC processing machine revealed direct correlations with resulting component qualities. The machine's capability for ultrasonic-supported machining exposed critical correlations between the set US frequency and spindle speed. Investigations also covered mid-spatial frequency analysis and periodic surface errors. At the same time, a machine learning model was developed, which enables a prediction of surface qualities depending on the grinding parameter selection even on the basis of a small database. By analyzing the frequencies recorded through vibrometry in the process, additional correlations with the formation of sub-surface damage could be assumed.