Clustering of acoustic emission signals collected during drilling process of composite materials using unsupervised classifiers

Conventional methods of analysis for drilling of composite materials usually study the amount of damaged area, thrust force, and effective parameters. However, these methods do not provide the investigator with sufficient information about drilling mechanisms. In the current investigation, a procedure for diagnosing different drilling mechanisms based on the analysis of the signals of acoustic emission is presented. According to the number of time domain acoustic emission parameters, using multi-variable methods of analysis is unavoidable. In this work, unsupervised pattern recognition analyses (fuzzy C-means clustering) associated with a principal component analysis are the tools that are used for the classification of the recorded acoustic emission data. After classification of acoustic emission events, the resulting classes are correlated with the different drilling stages and mechanisms. Acoustic emission signal analysis provides a better discrimination of drilling stages than mechanic-based analyses.