Verification of predictive models for the fatigue life of alloys based on nickel, titanium, aluminium and cast irons

Article deals with the verification of the accuracy and limitations of analytical relations for predicting the fatigue life of selected structural materials. The objective is to compare experimentally obtained data with theoretical values calculated using the Basquin equation for the stress-based approach. The basic mechanical properties as input data (ultimate tensile strength – UTS and elongation – A) were also employed for fatigue life prediction. Four distinct material groups with differing micro-structures were selected for the application of these prediction models: a nickel-based superalloy (Inconel IN718), a titanium alloy (Ti6Al4V), an aluminium alloy (AlSi9Cu3), and selected types of nodular cast irons (with a ferritic-pearlitic matrix EN-GJS-X300SiCu4-1.5 and an austenitic matrix ENGJSA-XNiMn13-7). The predictive fatigue life models were verified using an experimental dataset compiled from conventional fatigue testing. The testing frequencies ranged from 85 Hz to 150 Hz, specifically adjusted according to the dynamic characteristics of the testing machines and the evaluated material groups (150 Hz for IN718, 100 Hz for Ti6Al4V, 85 Hz for AlSi9Cu3, and 120 Hz for the cast irons). The results exhibit varying degrees of deviation in prediction accuracy. While the empirically calibrated Basquin-Morrow model restricts relative stress errors below 4.0% for the alloys tested under symmetrical loading (R= -1), asymmetrical testing (1 < R < ∞) yields higher prediction errors up to 28.7% due to structural inhomogeneity and bending stress complexity. Globally, replacing generic literature data with experimentally determined fatigue parameters reduced the Mean Absolute Error (MAE) from 46.0% down to 5.8%.