K-means clustering for feature extraction from wavelet-based flute signals

Abstract

High Speed Machines (HSM) are machines that operate at cutting speeds significantly higher than those typically utilized for a particular material and are very important near-end-line manufacturing devices. The quality of cutting highly depends on cutter specifications, type of work piece and cutting conditions.

To determine the performance, some industry-standard measurements are applied on the resulting surface. But, these tests cannot be easily carried out when the HSM center is running. Therefore, an alternative solution is to use sensory-based signals collected from the cutting process and correlate them to the quality of the resulting surface.

The objective of this project is to carry out research of Wavelet analysis effect on sensory-based signals collected. Next, K-means clustering is implemented to aid in developing methodologies to further facilitate milling tool optimization parameters as well as tool wear abnormality detection subsequently.

The project requires understanding of Wavelet analysis to analyse the data (force, acoustic, vibration) obtained from the cutter and establish a relationship between the signal and performance of the blade. From there, the relevant parameters that can be manipulated in K-means clustering are determined