Relatively simple fatigue tests have been performed on two common engineering materials, cast ductile iron and low‐carbon steel, using two stress states, cyclic torsion and cyclic torsion with static axial and hoop stresses. Tests were designed to discriminate between normal stress and hydrostatic stress as the most suitable mean stress correction term for high cycle fatigue analysis. Microscopy shows that cracks in low‐carbon steel nucleate and grow on maximum shear planes, while for cast iron pre‐existing flaws grow on maximum normal stress planes. The data illustrate that tensile normal stress acting on a shear plane significantly reduced fatigue life and is an appropriate input for fatigue analysis of ductile materials. Static normal stresses did not significantly affect the fatigue life for the cast iron because the net mean stress on the maximum normal stress plane was zero. Mean torsion significantly reduced the fatigue strength of the cast iron. A critical plane long‐life parameter for nodular iron which accounts for both stress state and mean stress is proposed, and is found to accurately correlate experimental data.

Historically, engineers have relied on macroscopic properties, e.g. hardness and tensile strength to predict fatigue limits as analytical tools to model the process did not exist. Consequently, many empirical modifications to the fatigue limit have been made to account for variables, e.g. surface roughness, state of stress, inclusion content, environmental effects, etc. A mechanistic model is proposed to quantify the effects of these parameters on the fatigue limit of metals, specifically steels. Fatigue resistance, i.e. the threshold condition of a non‐propagating crack, is determined by two parameters: non‐propagating defect or crack size; and the strength of the barrier to crack propagation. The concept of three defect types associated with three different flaw‐dominated fatigue regimes is introduced. Furthermore, application of the model to fatigue mechanisms in high‐strength steels, synergistic effects of surface finish and intergranular cracks, competition between surface and subsurface fatigue nucleation, tempering, and scatter in fatigue behaviour is demonstrated. The model can be implemented in material screening, selection and processing, as well as a guide for future material research and design. Overall, the model is proven as a simple and robust tool for qualifying and statistically quantifying material behaviour.