Utilização das Técnicas de Impedância Eletromecânica e Ondas de Lamb para Identificação de Dano em Estruturas com Rebites

Abstract

The present work is devoted to studying the performance of two well-known structural health monitoring (SHM) techniques to detect damage in typical metallic riveted structures, such as beams and panels found in various types of industry. Both techniques use piezoelectric materials for mechanical excitation (actuator effect) and dynamic response analysis (sensor effect) of the structure. The first approach used is known as electromechanical impedance technique, in which the electrical impedance of the piezoelectric patch is directly associated with the mechanical impedance of the structure. By this way, the detection of damage can be made by observing the variation of this parameter and comparing the signals patterns (signatures) of the healthy and damaged states of the structure. There are several damage metrics aiming at quantifying the damage for SHM analysis. Another approach uses the socalled Lamb waves, which propagate along solid materials. Information is collected from the structure under healthy conditions by using wave propagation phenomena (such reflection and attenuation) along the structure. The evolution of the damage can be observed by using a specially defined Damage Metrics that is able to quantify damage in this case the Damage Index (D.I.). The first studied structure was an aluminium beam containing a single rivet, which was used to test both techniques for damage (rivet loss) identification purposes. After this first validation procedure, the identification of a systematic loss of rivets on an aeronautic aluminium panel containing various rivets was studied. This configuration presents a more complex structure as compared with the aluminium beam, from the structural view point. Statistical techniques based on hypothesis tests were applied, aiming at identifying damage with a certain confidence interval. Finally, this work presents an alternative approach using artificial neural networks applied to pattern recognition associated to different damage states of the structure. All techniques presented in this dissertation showed to be useful for damage detection, as far as rivet losses are concerned.