New approaches for a reliable fatigue life prediction of powder metallurgy components using machine learning
The design of highly loaded crane structures is linked to the fatigue life assessment of welded structures in the LCF regime. Constant and variable amplitude fatigue test results of butt joints, transverse attachments, sample components, base material, welded small scale and microstructure specimens are used to investigate influences on the fatigue life of the welded high-strength steels S960QL, S960M and S1100QL. Endurable nominal stresses resulting from stress-controlled tests under constant amplitude loading of butt joints and transverse attachments can be extended into the LCF regime up to the material yield strength. While evaluated Woehler lines are correlated with FAT 71 and FAT 80 of common fatigue standards, results under variable amplitude loading of a Gaussian spectrum show a significant increase in fatigue life, which can be expressed by the linear damage sum. A fatigue life increase can be achieved by the use of investigated post weld treatment methods: TIG-dressing, highfrequency hammer peening and grinding. TIG-dressing of S1100QL butt joints is especially appropriate for the generation of a significant increase in fatigue life in the LCF regime. Strain-controlled fatigue testing of base material and butt welded sheet specimens have been carried out in order to evaluate the cyclic material behaviour. The cyclic material behaviour is influenced by the local weld geometry and microstructure. Local weld features and a reference radius of r = 1 mm and r = 0.05 mm are assigned to numerical finite element models in order to evaluate the fatigue life according to the local concept. The notch stress concept using r = 1 mm is a promising method until the yield strength of the base material is reached. Elastic-plastic simulations of the notch strain concept are capable of deriving PSWT Woehler lines.