PSI - Issue 77

Martin Matušů et al. / Procedia Structural Integrity 77 (2026) 127 –134 Martin Matušů / Structural Integrity Procedia 00 (2025) 000 – 000

131

5

K&V model was chosen because it effectively captures the fatigue behaviour across the entire lifespan — from low cycle fatigue (single-digit cycles) to the transition into the fatigue limit region (more than 10 6 cycles). Unlike the conventional power-law model, the K&V approach allows us to include into the regression analysis also experiments in the transition areas from the common single-slope region without hesitating. Fig. 2 represents only a small subset of the whole dataset, as the complete data volume is too extensive for this conference paper. While the results span various specimen geometries [1] and heat treatment conditions [2; 9], as detailed in the referenced works, the mutual positions and parameters of the fatigue curves themselves are not the primary focus of this paper and they will not be analyzed in depth. Instead, the analysis is centered on identifying outliers, with the particular attention to their position on the platform, in order to assess whether specimens from certain locations consistently exhibit reduced fatigue life compared to others. 2.3. Printing position evaluation Each of the 22 datasets, corresponding to separate experimental series, were analyzed independently. For each series, the Kohout & Věchet (K&V) regression model was fitted using the least squares method. The resulting regression curve represents the reference line corresponding to a 50% probability of failure ( P 50% ). To normalize the data, each experimental point — defined by its applied load amplitude and number of cycles to failure — was scaled relative to the fatigue strength related to the fitted K&V curve at the same number of cycles: = ( 50% ) ( ) = 1 ( ) ∙ [ ∙ , + , + ] . (2) The resulting Fatigue Index FI is derived from a KV , C, B, β parameters of the fitted K&V curve, while ( ) and ( ) are representing the experimentally measured dataset. The use of FI effectively transforms the dataset into a one-dimensional representation, allowing for direct comparison across different series and conditions. The resulting Fatigue Indexes of individual experimental data points in their relevant K&V curves were fitted using a Weibull cumulative distribution function (CDF): ( ) = 1 − [− ( ) ], (3) where η is the scale parameter and m are the shape parameter of the Weibull distribution. Outliers were defined in this paper as data points gathered within the lowest 25% of the distribution in Eq. (3). The corresponding threshold value is calculated from the inverse of the Weibull CDF: 25% = ∙ [− (0.75)] 1 (4) To further refine outlier identification in Eq.(4), the Fatigue Indexes were compared against the median curve predicted by the K&V model, denoted ( 50% ) . Specimens were classified as outliers if they fulfilled the following condition: