A neural network approach to damage detection in Euler-Bernoulli beams subjected to external forces

The aim of this contribution is to present two methods for online damage detection in Euler-Bernoulli beams subjected to external forces. Both methods detect damage by tracking changes in the beam parameters. Here, this change is assumed to occur in time, but not in space; that is, it occurs at a certain time instant, being the same along the beam. The input to the methods consists of the beam vibration data collected at different points. The first method is based on the use of a single Hopfield neural network. At each time instant, this network produces an estimate of the beam parameters and this estimate is the same for all beam points. In turn, the second method combines several Hopfield neural networks. At each time instant, each network produces an initial estimate of the parameters at a certain beam point and the estimates of neighbouring points are then combined to produce a final estimate at each point.