Fault Analysis in Composed Material: A Neural Net Application Using Acoustical Signal

Abstract

Since ’80 an increasing interest in Global Faults Determination is developing. Here, we
are interested in non destructive analysis. Non destructive analysis is applied to detect and to localize structure faults by using a signal with a wavelength smaller than the detecting fault. To detect the fault, this type of analysis requires the evaluation of the object in numerous small sections. In Global Faults Determination the fault detection procedure requires only a global measurement in the structural component in operational conditions, which decreases the cost considerably since do not require very large number of measurements. In this effort a neural network as a global fault diagnosis detector for structural mechanical components will be applied. The research is applied in structures such as composite beams. The composite material is epoxy reinforced with fiber glass. Those beams have saw cuts with different deepness in order to simulate possible faults. Acoustic signals, based on signals captured by a microphone, are used as neuronal network input. A Levenberg-Marquardt backpropagation algorithm is used for training a supervised fully connected feedforward neural network.