Building of constant life diagrams of fatigue using artificialneural networks

Abstract The objective of this paper is to demonstrate the applicability of artificial neural networks on building constant life diagrams of fatigue, making it possible to be obtained from a reduced number of S–N curves. In order to demonstrate this applicability, fatigue data at constant stress for 12 different S–N curves was obtained from the literature. Training three S–N curves, for R=10, 2 and 0.1, produces satisfactory results, however when the number of S–N curves is increased, more reliable solutions are obtained.