Health monitoring: new techniques based on vibrations measurements and identification algorithms

Purpose of the paper is to present an innovative application inside the nondestructive testing field based on vibrations measurements, developed, at the Department of Aeronautical Engineering of the University of Naples "Federico II" (Italy), by the authors during the last three years, and already tested for analysing damages of many structural elements. The aim has been the development of a nondestructive test (NDT) which meet to most of the mandatory requirements for effective health monitoring systems, simultaneously reducing as much as possible the complexity of the data analysis algorithm and of the experimental acquisition instrumentation; these peculiarities may, in fact, not be neglected for an operative implementation of such a system. The proposed new method is based on the acquisition and comparison of frequency response functions (FRFs) of the monitored structure before and after an occurred damage. Structural damages modify the dynamical behaviour of the structure such as mass, stiffened and damping, and consequently the FRFs of the damaged structure in comparison with the FRFs of the sound structure, making possible to identify, to localize and quantify a structural damage. The activities, presented in the paper, mostly focused on a new FRFs processing technique based on the determining of a representative "damage index" for identifying and analysing damages both on real scale aeronautical structural components, like large-scale fuselage reinforced panels, and on aeronautical composite panels. Besides it has been carried out a dedicated neural network algorithm aiming at obtaining a "recognition-based learning"; this kind of learning methodology permits to train the neural network in order to let it recognise only "positive" examples discarding as a consequence the "negative" ones. Within the structural NDT a "positive" example means "healthy" state of the analysed structural component and, obviously, a "negative" one means a "damaged" or perturbed sta- - te. With this object in view the neural network has been trained making use of the same FRFs of the healthy structure used for the determining of the damage index, as positive examples. From an architectural point of view magnetostrictive devices have been tested as actuators, and piezoceramic patches as actuators and sensors. Besides it has been used a laser-scanning vibrometer system to validate the behaviour of the piezoceramic patches and define their technical parameters in order to lay the bases for design a light and reliability system. These techniques promise to bring a step forward for the implementation of an automatic "health monitoring" system which will be able to identify a structural damage in real time, improving the safety and reducing maintenance costs