Any changes in specimen size in relation to the reference dimensions involve scaling inaccuracies resulting in the variances in strength testing (monotonic, fatigue) results. It is referred to as a size effect. The size effect is described using a cross-sectional coefficient determined for various specimen sizes and test types. The analysed material is aluminium alloy EN AW-6063 T6 with a cross-sectional area of 28, 7 and 3.5 mm2.

On the basis of the results, Wöhler's curves were prepared, presenting the dependency of fatigue life to equivalent stress levels. Their analysis shows that fatigue life is strictly connected with the value of coefficient λ. The existence of its critical value can also be observed, which results in the highest fatigue life reduction. The value is different for each material. Furthermore, fractographic tests were conducted showing the influence of the level and type of load on the fracture face.

The article presents the results of fatigue life and fractographic testing of steel X2CrNiMo17-12-2 exposed to proportional and non-proportional fatigue loads. The following load types were applied: tension-compressive strength, torsion, proportional combined/complex loads produced by tension-compressive strength and torsion as well as non-proportional combined load – by tension-compressive strength and torsion by the phase shift angle φ=90°. The paper analyses the effect of the load method on the fatigue life and fractography of fatigue fractures recorded, and especially the effect of non-proportional load.

The study presents two approaches to plotting an S-N curve based on the experimental results. The first approach is commonly used by researchers and presented in detail in many studies and standard documents. The model uses a linear regression whose parameters are estimated by using the least squares method. A staircase method is used for an unlimited fatigue life criterion. The second model combines the S-N curve defined as a straight line and the record of random occurrence of the fatigue limit. A maximum likelihood method is used to estimate the S-N curve parameters. Fatigue data for C45+C steel obtained in the torsional bending test were used to compare the estimated S-N curves.For pseudo-random numbers generated by using the Mersenne Twister algorithm, the estimated S-N curve for 10 experimental results plotted by using the second model, estimates the fatigue life in the scatter band of the factor 3. The result gives good approximation, especially regarding the time required to plot the S-N curve.