Feature extraction from energy distribution of stamping processes using wavelet transform

As one of the most commonly used manufacturing processes, stamping operations are applied to an extended range from centimeter-class parts to meter-class parts. The demand for the quality and productivity of stamping products is ever increasing. Hence, the extraction of the appropriate feature to implement online monitoring has been attempted. The vibration signals provide rich information for finding the dynamic behavior at high frequency vibration of the press. In this paper, an accelerometer has been employed in place of tonnage sensors or strain gauges to monitor the stamping process. In order to characterize the transient signal of the process, this work extracts the energy densities and frequency band energy (FBE) from the vibration signal. Owing to its inherent properties, the wavelet packet transformation decomposes the original signal into basis functions, with different energy distributions dominating in different time-frequency bands. Based on the experimental results, this suggests that extracting FBE from a vibration signal using a wavelet packet as a feature for fault diagnosis can, in practice be an effective approach for stamping process monitoring.