Suitable methods and transferability criteria and knowledge of the cyclic material behaviour is essential for a durable design of a component. For this reason, the cyclic material parameters are determined as a function of the product's properties (level of deformation, microstructure, surface finish, residual stresses) and different loading parameters. However, since the determination of the cyclic parameters is associated with considerable experimental effort and costs, a cost-effective and easy method is sought to determine these parameters. A very promising approach for this is the application of artificial neural networks (ANN) since they have the ability to generate the influences on the fatigue strength from the manufacturing and environmental parameters using sensibly selected input parameters. They offer the possibility to access acquired knowledge and to thus construct a multidimensional map based on a few tests. By combining a few experimental tests with the ANN, the result of the estimation can be improved and the experimental effort can be reduced.

Several estimation methods have been developed to estimate the cyclic material parameters out of the static material properties. Most of these methods are based on empirical equations. Increasing numbers of input- and influencing parameters lead to an rising effort for determining these equations and the accuracy decreases. For this reason new suitable methods are sought to estimate the cyclic material behaviour. A very promising approach is the application of the artificial neural networks, which can derive self-depended a relationship between in-and output parameters. Static parameters such as yield strength, tensile strength. etc., which can rapidly be determined used as input parameters. The output parameters are the cyclic material parameters of the strain-life curve and stress-strain curve according to the Manson-Coffin-Basquinand Ramberg-Osgood curve. Many different artificial neural networks with different structures and complexity can be applied. In this paper the influence of the topology of an artificial neural network on the estimation accuracy will be investigated. Based on the results of a reference artificial neural network it will be shown, that more complex topologies in the network do not lead inevitably to better estimations.

Zur Abschätzung der Lebensdauer von zyklisch belasteten Bauteilen anhand statischer Werkstoffkennwerte existieren zahlreiche Abschätzungsverfahren, die größtenteils auf empirischen Gleichungen beruhen. Mit zunehmender Anzahl der Eingabe‐ und Einflussparameter steigt auch der Aufwand zur Ableitung dieser Gleichungen und die Abschätzungsgenauigkeit sinkt. Daher wurde nach neuartigen Methoden gesucht, um die zyklischen Kennwerte zuverlässig abschätzen zu können. Ein vielversprechender Ansatz ist mit den künstlich neuronalen Netzen gefunden worden, welche den Zusammenhang zwischen Eingangsparametern und Ausgabeparametern selbst erschließen können. Als Eingangsparameter kommen die statischen Kennwerte der zu untersuchenden Werkstoffe zum Einsatz, da diese gegenüber den zyklischen schneller und kostengünstiger zu ermitteln sind. Als Ausgabegrößen werden die zyklischen Kennwerte der Ramberg‐Osgood‐ und der Manson‐Coffin‐Basquin‐Gleichung abgeschätzt. Die künstlich neuronalen Netze können dabei sehr unterschiedlich und beliebig komplex aufgebaut sein. In dieser Arbeit wird der Einfluss der Verknüpfungen der Neuronen untereinander, der sogenannten Topologie, auf die Abschätzungsgüte der künstlich neuronalen Netze untersucht. Ausgehend von einem Referenzfall wird gezeigt, dass komplexere Topologien demgegenüber nicht die gewünschten Verbesserungen bei der Abschätzung der zyklischen Kennwerte erzielen.

The paper contains a proposal for the classification of some selected constructional materials starting from the material properties as assumed according to the models based on the Manson–Coffin–Basquin (MCB) and Ramberg–Osgood (RO) laws. These laws are commonly used for strain-based fatigue life assessments. Three methods for determination of fatigue material constants occurring in the MCB and RO models, namely, the conventional, numerical and 3D methods, were used. The compatibility between the parameters derived from the aforementioned models was checked by evaluating the fatigue results for five groups of selected constructional materials.

The structural durability design of components requires the knowledge of cyclic material properties. These parameters are strongly dependent on environmental conditions and manufacturing processes, and require many experimental tests to be correctly determined. Considering time and costs, it is not possible to include in the tests all the variables that influence the material behaviour. For this reason, the computational method of the Artificial Neural Network (ANN) can be implemented to support these investigations. This method allows an estimation of the cyclic material properties starting from the static parameters deducted through tensile tests. The results permit a very good approximation of cyclic material properties using just a few specimens in tests, so that the experimental effort can be deeply reduced. The ANN has been implemented in the software called Artificial Neural Strain Life Curves (ANSLC), and has been tested on a large database of steels. In this paper the method of the ANN and the program ANSLC will be presented. © 2011 Published by Elsevier Ltd. Selection and peer-review under responsibility of ICM11

A concept to evaluate the fatigue strength of linear flow split profile’s sections using a local strain approach based on the hardness distribution is presented. For this purpose established correlations between hardness and cyclic material properties are adapted to fit experimental results derived by fatigue tests on smooth, non-homogeneous specimens extracted from such profiles. For validation a numerical fatigue strength evaluation of a four-point-bending fatigue test of linear flow split profiles is presented using an elastic–plastic FE model with the material property distribution derived. The developed approach allows an improved estimation of the fatigue strength of the component analysed.

Load controlled fatigue tests were performed up to 107 cycles on flat notched specimens (Kt = 2.5) under constant amplitude and variable amplitude loadings with and without periodical overloads. Two materials are studied: a ferritic‐bainitic steel and a cast aluminium alloy. These materials have a very different cyclic behaviour: the steel exhibits cyclic strain softening whereas the Al alloy shows cyclic strain hardening. The fatigue tests show that, for the steel, periodical overload applications reduce significantly the fatigue life for fully reversed load ratio (Rσ = –1), while they have no influence under pulsating loading (Rσ = 0). For the Al alloy overloads have an effect (fatigue life decreasing) only for variable amplitude loadings. The detrimental effect of overloads on the steel is due to ratcheting at the notch root which evolution is overload's dependent.

Die Ergebnisse von Schwingfestigkeitsversuchen von Proben, die aus dem stark umgeformten Bereich von Spaltprofilen entnommen wurden, werden dargestellt. Im Fokus steht die Wirkung von Mittelspannungen und Kerben im Zusammenhang mit der vorhergegangenen Umformung. Der Eigenspannungszustand im Bauteil und in den untersuchten Proben wird beschrieben. Numerische Analysen der Schwingfestigkeitsversuche an den Proben unter Berücksichtigung des Eigenspannungszustands ermöglichen es, den Einfluss von Eigenspannungen auf die Versuchsergebnisse rechnerisch zu analysieren. Zur Übertragung der an ungekerbten Proben ermittelten Kennwerte auf gekerbte, bauteilähnlichere Proben wird das Weibullsche Fehlstellen‐Modell angewendet.

The innovative sheet metal forming technology "Linear Flow Splitting" offers various new options for designing profile-like components. The forming process leads to severe changes in local material properties, inhomogeneities and residual stresses within the manufactured component. These effects influence the mechanical properties of the manufactured components. If the components are designed to endure cyclic mechanical loads, it is especially important to know the components fatigue properties. This paper focuses on a method to derive the fatigue properties of Linear Flow Split Profiles by nonlinear numerical FE analysis, including durability analysis and forming simulations. This numerical approach offers the possibility to estimate the fatigue properties of components before manufacturing physical prototypes, only based on material parameters derived from tests on smooth samples. The Finite-Element analysis of the Linear Flow Splitting Process provides distributions of local material deformation and residual stresses. These results are mapped by an appropriate interface on FE models, which allow simulating the component behavior under external loads. Thus, the inhomogeneous elastic-plastic material behavior and residual stresses are considered in the computed stresses and strains. Further on, a post-processing tool was implemented to interpret the FE results considering the inhomogeneous distribution of materials fatigue properties, the mean stress distribution and the statistical size effect.