Statistical monitoring of rotating machinery by cumulant spectral analysis

Abstract

A higher-order statistical (HOS) signature analysis methodology was applied to accelerometer data from a drilling machine collected in a controlled experiment drilling holes through composite circuit panels. Background on the drill wear monitoring problem including approaches using a combination of sensors and signal features are briefly summarized. Experiment results reveal that statistics from the second order cumulant spectrum not constrained to the periodic component support set had increased discrimination power when studying early or incipient drill wear. Detailed classification results show that feature sets composed of second order cumulant components had greater sensitivity than bispectrum and power spectrum features for indicating incipient drill wear.<>