Developing artificial neural networks to estimate the fatigue strength of structural steel details using the new European database

Abstract This article outlines a method for utilizing machine learning, particularly artificial neural networks, to estimate the fatigue strength of structural steel details. Data have been taken out of a structured database of fatigue tests, depicting the background of EN 1993‐1‐9. The artificial neural network has been trained and verified on the basis of experimental fatigue test results on the example of the transverse stiffener. Results show that the neural network is capable of predicting the fatigue strength of random transverse stiffener details. Comparisons have been made to a numerical approach applying the effective notch stress approach, showing also limits. This study helps paving the way for a thorough investigation into the complex relationship between different influencing factors and fatigue strength, highlighting the benefits and limitations of using machine learning tools.