Markus Sause »  Tobias Borsdorff
Markus Sause
Identification of failure mechanisms in hybrid materials utilizing pattern recognition techniques applied to acoustic emission signals
Supervisor: Prof. Dr. Siegfried Horn [Experimental physics II]
Date of oral examination: 12/16/2010
310 pages, english , ISBN: 9783866648890
Acoustic emission signals recorded during failure of hybrid materials are analyzed by pattern recognition techniques and are compared to results of finite element simulations. A new feature based pattern recognition approach is introduced. The method uses an exhaustive screening of all feature combinations in combination with a voting scheme based on cluster validity indices. This allows identification of the number of natural clusters of signals and an automated feature selection process, which yields numerical optimal separation of the signals without a-priori assumptions. Utilizing microscopy and finite element simulations these natural clusters are correlated with distinct acoustic emission sources. In addition an analysis of the mean time-frequency behaviour of the distinct signal types is presented. The finite element simulation introduces a new microscopic source model, which takes into account the finite dimensions and the inhomogeneous elastic properties of the source. The influence of the parameters describing the source excitation and the relative source-sensor distance is investigated. To compare with experimental signals a realistic model of the experimentally used sensor is established. A systematic investigation of the acoustic emission signals recorded during failure of carbon fiber reinforced plastics is presented for various loading conditions and stacking sequences. The application of the pattern recognition method yields three distinguishable types of signals, which are correlated to the occurrence of fiber breakage, matrix cracking and interface failure. Simulated signals of the respective failure types show characteristics similar to those of the experimental signals. A quantitative evaluation of the acoustic emission signal amplitudes shows strict correlation to the mechanical properties of the specimen investigated. In addition, the acoustic emission signals originating from failure of metallic coatings applied to carbon fiber reinforced plastics is investigated by pattern recognition. In this case the distinct types of signals are correlated to the occurrence of mode-I cracking within the coating and interfacial delamination between coating and substrate. The respective measured acoustic emission energies are compared to micromechanical calculations in a fracture mechanics approach.