Uncertainty Quantification of the Crack Propagation Behavior in Welded Stiffened Panels Using a Hybrid Framework Integrating Artificial Neural Networks and Finite Element Analysis

Abstract

Predicting the crack propagation behavior in welded stiffened panels often used in marine, civil, and aerospace structures, has been challenging. The presence of welded stiffeners creates complex stress conditions that need to be properly considered while predicting the crack growth. Another challenge is the presence of significant uncertainties related to the variability in the geometric parameters. In this context, the proper consideration of the uncertainties associated with these parameters is crucial for the accurate prediction of the fatigue service life and for ensuring structural integrity and operational reliability throughout the service life. This paper conducts a sensitivity analysis to evaluate the effect of relevant input parameters, covering the characteristics of the main panel and stiffener characteristics, on the crack propagation behavior. A 3D finite element analysis, an artificial neural network, and an elastic-plastic crack growth model are integrated to predict crack propagation under cyclic loading. The sensitivity assessment approach is illustrated on stiffened panels with T- and L-shape stiffeners subjected to axial tensile fatigue loading. The proposed approach is validated using experimental test data reported in literature.