Grinding Vibration Detection Using a Neural Network

Abstract
The amplitude of grinding vibration increases gradually throughout the grinding wheel wear process. In the meantime the predominant vibration frequency shifts in a region close to a natural frequency of the system. The complex time-varying pattern of vibrations makes it a problem to objectively identify when the grinding vibration becomes unacceptable and when the wheel should be redressed. A neural network approach method was proposed in this paper to identify the wheel life. The signal data were pre-treated by eight-band-pass filters, which covered the whole frequency range of the grinding chatter. These pre-treated data were used as the inputs to the neural network. By training the neural network, an objective criterion can be determined for the wheel redress life.