ABSTRACT The fatigue limit and fatigue life vary depending on the loading mode and are also greatly affected by stress concentration. The effect of mean stress has been mathematically expressed using several models, but based on experimental data, the Walker model is the most appropriate. Although the model constants differ depending on the material, stress concentration state, and loading mode, it was shown that high prediction accuracy can be achieved by selecting appropriate constants for both smooth and notched specimens. The same was true for the effect of assembling stress due to welding distortion on the fatigue limit and the threshold stress intensity range of welded details. Furthermore, it was shown that this model can also be applied to predicting multiaxial fatigue and to estimating statistical failure probability. By using the concept of equivalent alternating stress, it became possible to apply the Walker model to predict both the fatigue limit and the fatigue life under arbitrary stresses in the high‐cycle regime.
ABSTRACT Ferritic and ferritic/martensitic steels exhibit a ductile‐to‐brittle transition (DBT) whose large scatter in cleavage toughness makes structural integrity assessment statistical. This review focuses on the experimental, data‐driven and event‐based evidence bearing on the probabilistic description of transition toughness, and on the shape of the toughness–temperature curve; the extensive numerical constraint‐modelling literature is deferred to a companion review. Treating the transition micromechanism critically, we show that the weakest‐link Master Curve survives recent fractographic findings, and that the Weibull exponent is derived from small‐scale‐yielding crack‐tip scaling and confirmed experimentally, not assumed. Using model‐free machine‐learning analyses of raw toughness data, we argue that the exponential Master Curve and the saturating Unified Curve differ decisively only beyond the transition window, that saturation is real rather than imposed, and that prior ductile crack growth is the deciding variable. Constraint, active volume, loading rate, and warm pre‐stressing are reviewed accordingly.
ABSTRACT Fatigue failure in engineering components predominantly occurs under multiaxial stress conditions. The complexity and high cost of multiaxial fatigue tests necessitate a robust equivalent stress principle capable of predicting multiaxial fatigue performance from readily available uniaxial data. However, existing principles lack generality for high‐strength steels across various conditions. This study compiles 44 stress conditions and evaluates prevailing approaches. The critical plane data indicate that a quadratic combination of shear and normal stresses with a sign correction distinguishing tension from compression is essential, and the fatigue strength data reveal that the normal stress parameter must be redefined to incorporate both amplitude and mean contributions through a weighted linear sum, rather than relying on its maximum alone. These findings yield a generalized equivalent stress principle serving dually as a critical plane identifier and a fatigue strength predictor for various loading scenarios, offering significant practical utility for the design of high‐strength steel components.
ABSTRACT Welding residual stress alters crack‐tip driving forces, complicating fatigue crack‐growth life assessment. Static corrections may leave systematic life deviations that are difficult to interpret. This study develops a physics‐constrained framework for diagnosing and decomposing these deviations in welded compact‐tension specimens. Two as‐welded specimens are assessed against a low‐residual‐stress reference from two post‐weld heat‐treated specimens. A logarithmic ratio of predicted to measured accumulated cycle increments quantifies deviations for hierarchical analysis. Relative to static correction, the residual‐stress‐pattern term reduces mean absolute deviations by 68.9% and 69.5% for A1–1 and A2–1, respectively. A closure/shielding‐related auxiliary term further reduces the deviations to 0.0538 and 0.0346 but exhibits greater boundary sensitivity. The prescribed functional form accommodates both specimens with different amplitudes, whereas direct parameter transfer increases error. The framework provides a diagnostic basis for evaluating the limitations of static residual‐stress correction.
To improve fatigue life prediction accuracy for rubber composites under different strain ratios, a physics-informed convolutional neural network (PI-CNN) model is developed. Based on uniaxial fatigue tests, an equivalent strain amplitude model incorporating a strain ratio adjustment factor is established. This model serves as a damage indicator to construct a physical model. Subsequently, a CNN model is constructed using the strain ratio and engineering strain amplitude as inputs, with the physical model predicted life as the learning target. The weights of the fully connected layer are further calibrated using experimentally measured fatigue life data, yielding the final PI-CNN model. Comparison with experimental data under different strain ratios shows that the predicted fatigue lives fall within a scatter factor of 1.5 times the measured values, demonstrating superior accuracy compared with the physical model (scatter factor of 3) or the CNN model (scatter factor of 2).
