Research on Smart Structures for Detecting Cracks Using Neural Network.

Abstract

Smart structures which have the ability to detect defects by themselves and which warn us of such danger have received wide attention. Smart structures with strain sensors such as piezo elastic ceramic films are especially attractive because the sensors are easily applicable to the structures in operution. Studies on the brain of smart structures with the strain sensors, however, have not been conducted. Recently, a back propagation model, a kind of neural network, has been the reported to be effective for an inverse problem. In this study, therefore, the back propagation model was examined for use it as the brain of the smart structures with strain sensors. For the smart structures which detect fatigue cracks, of a B747 airplane several parameters such as the learning method, neuron number at the hidden layer and the position of the strain sensors were surveyed by numerical simulations.