Why Fatigue Life Predictions Do Not Agree with Experiments:

Quantifying the prediction scatter through the FABEST benchmark competition

F​ree Webinar

Tuesday, 29 September 2026

15:00 (London) | 16:00 (Berlin)
07:00 (Los Angeles) | 10:00 (New York)

T​his webinar is brought to you by the NAFEMS Central Eastern Europe Region

Fatigue-life calculations can be remarkably sophisticated, yet their predictions still frequently disagree with experimental results. Why? Is the problem in the material data, the stress analysis, the fatigue model, or in effects that the model simply does not capture?

The webinar will introduce this problem from an engineering perspective, starting with the basic principles of S–N based fatigue-life prediction and gradually addressing the factors that complicate the apparently simple transition from material fatigue data to the life of a real component. Mean stress, size and notch effects, surface condition, residual stresses, loading frequency, material orientation, multiaxial loading and other factors can substantially affect fatigue behaviour. Different methods exist to account for them – but how much does the analyst's choice of method ultimately affect the predicted life?

A simple calculation example will demonstrate how seemingly moderate differences in the treatment of such effects can propagate into very substantial differences in predicted fatigue life.

This question is at the heart of FABEST, a benchmark competition organized within FABER – Fatigue Benchmark Repository, COST Action CA23109. FABEST provides participating analysts with a common set of carefully generated experimental input data on a single heat of 42CrMo4 +QT and asks them to predict fatigue behaviour for test cases initially withheld from them. The subsequent comparison with experimental results provides a unique opportunity not only to assess individual prediction methods, but also to quantify the scatter of predictions produced by the fatigue community when solving the same engineering problem from the same starting information.

The experimental campaign and its resulting dataset are therefore an essential foundation of FABEST, but not its final objective. The broader aim is to understand how reliably fatigue life can presently be predicted, which modelling choices contribute most strongly to prediction scatter, and where further methodological development and standardisation are needed.

The webinar is intended both for engineers working directly with fatigue assessment and for simulation engineers who use fatigue-life prediction tools without necessarily being specialists in fatigue testing or fatigue methodology.

W​elcome and Introduction
Jo Potts, NAFEMS & Witold Krakol, NAFEMS Central Eastern Europe Chair

Why Fatigue Life Predictions Do Not Agree with Experiments:
Quantifying the prediction scatter through the FABEST benchmark competition
Jan Papuga, Czech Technical University

Q&A D​iscussion

U​se the "book" button on the right to register your free place.

Our Speaker

Jan Papuga, PhD, is a fatigue analyst and R&D Visionary at Evektor and a researcher at the Czech Technical University in Prague. His research focuses on fatigue-life prediction, multiaxial fatigue, notch effects and fatigue assessment of engineering structures. He is the developer of the PragTic fatigue-analysis freeware (www.pragtic.com/program.php), Chair of the Workshop on Computational Fatigue Analysis (www.pragtic.com/wcfa.php), and Chair of the FABER COST Action CA23109 (www.faber-cost.eu).