Condition monitoring (CM) is an effective way to improve the tool life of a cutting tool. However, CM techniques have not been applied to monitor tool wear in an industrial gear shaving application. Therefore, this paper introduces a novel, sensor-based, data-driven, tool wear estimation method for monitoring gear shaver tool condition. The method is applied on an industrial gear shaving machine and used to differentiate between four different tool wear conditions (new, slightly worn, significantly worn, and broken). This research focuses on combining, expanding, and implementing CM techniques in an application where no previous work has been done. In order to realize CM, this paper discusses each aspect of CM, beginning with data collection and pre-processing. Feature extraction (in the time, frequency, and time-frequency domains) is then explained. Furthermore, feature dimension reduction using principal component analysis (PCA) is described. Finally, feature fusion using a multi-layer perceptron (MLP) type of artificial neural network (ANN) is presented.

This content is only available via PDF.
You do not currently have access to this content.