Neural network application in fatigue damage analysis under multiaxial random loadings

A back-propagation neural network was applied to fatigue life prediction under multiaxial random loading. The proposed artificial neural network (ANN) model was demonstrated by predicting the multiaxial fatigue life and finding critical locations of an automotive sub-frame within reasonable time. While the conventional methods calculating multiaxial fatigue life with critical plane model require very long-time, this method outputs instantaneously result for a given set of input after the ANN model has been trained. The performance of the ANN model was evaluated by comparing outputs of the ANN with results of the conventional calculation method and acceptable in most fatigue design considerations. Particularly, this method gave good results in searching critical locations.