Journal of Vibration Engineering and Technologies, vol.11, no.7, pp.3185-3196, 2023 (SCI-Expanded)
Background: Due to excessive heat generated during grinding, temperature-related failures such as burning, temper damage and re-hardening damage may occur. Grinding burn is the most common type of thermal damage, and its detection requires special attention as it shortens the service life of a machine part, affects its hardness and stress condition, and ultimately leads to catastrophic failures. Although there are methods available for feature extraction and estimation in grinding burn detection, there is still controversy among researchers as to which feature (or features) reflects the presence of grinding burn. Purpose: This study aims to provide a new approach to grinding burn detection using low-order frequency moments of a spectrogram for feature extraction. Methods: Due to its sensitivity to grinding conditions and suitability for on-site operations, the grinding process is monitored using acoustic emission (AE) for different operating conditions, including regular grinding, grinding at higher cutting speed and larger infeed, and smaller dressing depth of cut. The state of the grinding process is evaluated by the time-dependent parameters of a spectrogram (e.g., instantaneous energy, mean frequency, and centralized second moment). Results and Conclusions: AE is very sensitive to any change in the grinding condition, and most of the condition-indicating information is in the frequency range up to 180 kHz. When grinding burn occurs, the strength of the AE reduces, and both the instantaneous energy and the centralized second moment of a spectrogram tend to decrease compared to the healthy state for all grinding operations. Low-order frequency moments are effective in grinding burn detection, serve as robust and reliable indicators for all grinding conditions, and can be a good reason for applying the AE technique in grinding burn identification in a production environment.