Cluster analysis of acoustic emission activity within wood material: Towards a real-time monitoring of crack tip propagation

International audienceNon-destructive monitoring of crack growth within material becomes an essential tool to better understand cracking mechanisms of materials, and therefore to reduce failure risks. This paper presents results from laboratory static tests that were conducted to investigate the monitoring of crack tip growth within wood material. Double Cantilever Beam specimens with variable inertia were tested in opening mode configuration. Acoustic emission (AE) technique was used to monitor the acoustic activity within wood material. In a first step, the reliability of the crack tip monitoring is evaluated in terms of Probability of Detection (PoD). Results show that for all specimens the PoD is larger than 70% for a detection threshold of 1 mm. In a second step, a K-means++ algorithm was used to perform a cluster analysis of AE data, and to allow the AE events that were generated by the crack tip growth to be identified. Evolutions of both strain energy and AE energy are compared, and the experimental results show that the first AE events appear when the wood material reaches its limit to store strain energy, so at the vicinity of the end of its linear elastic behaviour.Highlights DCB wood specimens with variable inertia were tested in opening mode configuration. Unsupervised classification of AE events to identify three failure mechanisms. Probability of Detection to evaluate the reliability of the crack tip monitoring. Probability of Detection is larger than 70% for a detection threshold of 1 mm.