A surrogate artificial neural network (ANN) model trained on the data generated from a computational finite element-based (FE-based) model is developed. The developed ANN model enables the estimation of the fatigue life (number of load cycles to failure) of component-like specimens with stress concentrators. Using the developed model, the component-specific S-N curves can be generated with an accuracy comparable to that of the computational FE-based model. The investigation covered through- and surface-hardened steel components with different numbers and types of stress concentrators. The basis for data generation is the parametrized computational FE-based model, which enables the determination of the stress–strain response and the calculation of the fatigue life of examined components under cyclic loading conditions. The computational FE-based model can be adjusted to include components with different geometries and heat treatment conditions. The computational FE-based model incorporates nonlinear material behavior to provide a more accurate representation of the component’s behavior, which results in higher computational costs. In contrast, the developed ANN model provides a quicker and more efficient way to assess the fatigue life of both through- and surface-hardened components, overcoming these limitations.

The maximal vertical distance (MVD) recursive algorithm, a novel approach for the optimal discretization of stress–strain material curves, is proposed. The algorithm simplifies the process of defining multilinear curves from material stress–strain curves when conducting a finite element analysis (FEA) of components. By directly selecting points on the material curve, the MVD algorithm eliminates the requirement for initial discretization, thereby minimizing information loss. As the measure of goodness of fit of the simplified polyline to the original curve, the percentage of stress deviation (SD) is proposed. The algorithm can generate multiple multilinear curves while keeping the stress deviation of each curve within a predefined limit. This feature is particularly beneficial during the finite element analysis of components exhibiting complex and position-dependent material properties, such as surface-hardened components, ensuring consistent modelling accuracy of material properties across the components’ geometry. Consistent accuracy also proves advantageous when exploring multiple differing material states of quenched and tempered steel, ensuring fair and reliable comparisons. The MVD algorithm was compared with existing algorithms from the literature, consistently maintaining the accuracy of the multilinear curves within predetermined limits using the fewest possible points.

This paper introduces a novel method for estimating the cyclic stress–strain curves of steels based on their monotonic properties and plastic strain amplitudes, utilizing artificial neural networks (ANNs). ANNs were trained on a substantial number of experimental data for steels, collected from relevant literature, and divided into subgroups according to alloying elements content (unalloyed, low-alloy, and high-alloy steels). Only monotonic properties that were proven to be relevant for the estimation of points on the stress–strain curve were used. The performance of the developed ANNs was assessed using an independent set of data, and the results were compared to experimental values, values obtained by existing empirical estimation methods, and by previously developed ANNs. The results showed that the new approach which combines relevant monotonic properties and plastic strain amplitudes as inputs to ANNs for cyclic stress–strain curve estimation is better than the previously used approach where ANNs estimate the parameters of the Ramberg–Osgood material model separately. This shows that a more favorable approach to the estimation of cyclic stress–strain behavior would be to directly estimate corresponding material curves using monotonic properties. Additionally, this may also reduce inaccuracies resulting from simplified representations of the actual material behavior inherent in the material model.