Material Detection by Stack Drilling Monitoring and Reservoir Computing

The multi-material stack drilling process is subjected to numerous possible defects such as tool wear, tool breakage, burr or delamination. It has been proposed to develop a monitoring system based on a digital twin of the process to validate the hole quality. This digital twin will first need to robustly identify the steps of the drilling process in order to monitor relevant process features despite signals distortions induced mainly by tool wear. In this study, two approaches based on the Gaussian mixture model and reservoir computing are compared to a state-of-the-art analytical method. These data-driven models are improved with the integration of expert knowledges. These models are tested on a series of CFRP/Al drilling which encompass the tool wear. The presented methods achieved an accuracy score over 95 % despite signals variability.