A neural network approach to describing the fretting fatigue in aluminium-steel couplings

Abstract Fatigue data for specimens from two aluminium alloys, two surface-finishing conditions, and different surface pressures, subjected to constant dynamic loading with different stress amplitudes, were used to evaluate possible artificial neural network (ANN) architectures for a description of the fretting-fatigue phenomena. A neural network approach shows that ANNs can be trained to model fretting-fatigue phenomena. The ANN we used provided an accurate prediction of the occurrence of fretting fatigue. The main benefit of the trained ANN was that it precisely described the effects of different factors on the occurrence of fretting fatigue. After the ANN has been trained it represents a robust tool for a description of fretting-fatigue phenomenon in aluminium–steel couplings. The ANN was able to acquire the same knowledge that it took many researchers to acquire. The robustness of the field applications is only restricted by the range of known (measured) data used for the ANN training.