Implantes

Composites: Part A 43 (2012) 254–260

Contents lists available at SciVerse ScienceDirect

Composites: Part A
journal homepage: www.elsevier.com/locate/compositesa

Unsupervised and supervised classification of AE data collected during fatigue test on CMC at high temperature
S. Momon, N. Godin ⇑, P. Reynaud, M. R’Mili, G. Fantozzi
Université de Lyon, MATEIS, INSA-Lyon, 7 Avenue JeanCapelle, 69621 Villeurbanne, France

a r t i c l e

i n f o

a b s t r a c t
This paper aims at giving a better understanding of damage mechanisms that control lifetime of Cf/SiC composites at high temperatures (700–1200 °C) under static and cyclic fatigue. Acoustic emission (AE) signals were analysed with a view to identify classes corresponding to a specific damage mode. An unsupervisedclassification method allowed differentiating signals resulting from the following damage mechanisms: collective or individual fibre breaks, matrix cracking, fibre/matrix debonding, yarn/yarn debonding and sliding at fibre/matrix interfaces or matrix cracks closing after unloading. Then, a supervised classification method was developed. It allows real-time identification of damage mechanisms regardless of testingconditions (temperature, applied load and loading mode). Ó 2011 Elsevier Ltd. All rights reserved.

Article history: Received 10 June 2011 Received in revised form 14 October 2011 Accepted 17 October 2011 Available online 3 November 2011 Keywords: A. Ceramic-matrix composites C. Damage mechanics D. Data clustering D. Acoustic emission

1. Introduction Ceramic matrix composites (CMCs) areinteresting structural materials for high temperature applications, owing to their good mechanical properties at elevated temperatures. Cf/SiC composites have been developed for aeronautic and aerospace applications. Several authors have studied their mechanical behaviour and damage mechanisms under static and cyclic loading [1–6]. Recently, composites with self-healing matrix have been introduced toimprove resistance to oxidation. Detection and analysis of acoustic emission (AE) are powerful means for identification of damage phenomena and monitoring of their evolution. Acoustic emission [7] is a transient wave resulting from the sudden release of stored energy during damage. In the case of composite materials, various mechanisms act as AE sources including matrix cracking, fibre/matrixinterface debonding, fibre failure and delamination [8–13]. A major issue in the use of AE technique is to associate every AE signature to a specific damage mechanism. Consequently, AE signals recorded during tests must be segmented into clusters based on similarity measures. However, this analysis is a non-trivial task for two main reasons. First, AE signals are complex objects that must be characterisedby multiple pertinent descriptors. Second, there is no a priori knowledge of the acoustic signatures of damage events and these are assumed rather scattered. A single parameter analysis usually allows differentiating two mechanisms with very different energies. Some authors applied this
⇑ Corresponding author. Tel.: +33 472 438 073; fax: +33 472 437 930.
E-mail address:nathalie.godin@insa-lyon.fr (N. Godin). 1359-835X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.compositesa.2011.10.016

type of analysis to CMCs [14–16]. Also in the case of CMCs, Morscher et al. used waveforms energy to monitor the occurrence of through thickness matrix cracks, considering that signals caused by fibre failure, interfacial debonding and sliding had a negligiblecontribution to the overall energy release [17,18]. However, a conventional analysis based on only one parameter (such as amplitude) is often inadequate for composite materials. An improvement is the use of clustering algorithms [19–22] able to group together into clusters signals exhibiting similarities. AE signals are represented in an ndimensional space by a vector of signal features, often called...