Pattern Recognition for damage detection in aerospace vehicle structures

Abstract

A Pattern Recognition based damage detection scheme for aerospace vehicle structures is proposed. It involves capturing mechanical vibration signals from plate like structures using displacement sensors; removal of noise and extraction of features using Wavelet Transform based signal processing techniques, and training a Neural Network Ensemble to classify and identify the damages that appear in the structure. A few cases are studied. Encouraging results are found in classifying single damages in the structure. However success rates dropped in case of identifying multiple damages for the same structural form. A sensor placement strategy is then drawn out that improved the results significantly.