Structural health monitoring using continuous sensors and neural network analysis

A method for damage detection in a plate structure is presented based on strain waves that are generated by foreign object impact on the structure, or by damage that is propagating in the structure. The response characteristics of continuous sensors, which are long ribbon-like sensors, were studied by simulation of wave propagation in a panel. The advantage of the continuous sensor is to improve damage detection by having a large coverage of sensors on the structure using a small number of channels of data acquisition. Strain responses from the continuous sensors were used to estimate the damage location using a neural network technique. Eight hundred numerical wave propagation simulation runs for a plate were carried out to train the neural network and verify the proposed method for damage localization. The identified damage locations agreed reasonably well with the exact damage locations. Overall, the approach presented is meant to simplify the instrumentation needed for damage detection by using continuous sensors, a small number of channels of data acquisition, and training a neural network to do the work of locating the damage source.