Adaptive prognosis of fatigue damage based on the combination of particle filters and neural networks

A method for the prediction of the residual life of a component subject to structural degradation which stems from the combination of a particle filter with an artificial neural network is proposed in this study. The artificial neural network is adaptively trained on-line, i.e., its parameters are identified in real time by the particle filter as new observations of the structural degradation become available from a generic diagnostic system. The adaptive network is then used to perform a multiple-step ahead prediction, thus estimating the probability density function of the residual life. The advantage of the method is that it can potentially adapt to different trends in the damage evolution, even in presence of anomalous behaviors due to failures or unforeseen operating conditions.