PSI - Issue 57

Mohammad F. Tamimi et al. / Procedia Structural Integrity 57 (2024) 121–132 123 Mohammad F. Tamimi & Mohamed Soliman/ Structural Integrity Procedia 00 (2023) 000 – 000 3

Nomenclature LEFM Linear Elastic Fracture Mechanics the crack size. the number of cycles. 0 Paris power law coefficient. 0 Paris exponent. ∆ the range of the stress intensity factor. E the modulus of elasticity. Poisson’s ratio the material yield strength. the ANN output. O neuron number m layer number. the activation function. the number of connections to the previous layer. −1 the weights of each connection. the bias factor. 2. Sensitivity Analysis

Sensitivity analysis is a vital tool for assessing the impact of random input variables on the response of a model. This analysis is crucial for problems involving probabilistic modeling and reliability assessment (Gaspar et al., 2016). It aids in determining influential parameters that significantly contribute to model output variability (Saltelli et al., 2010), which can assist in structural design optimization and cost-benefit analysis (Leheta & Mansour, 1997). Two main approaches exist for conducting the sensitivity analysis: local and global approaches (Saltelli et al., 2010). Local methods, such as first-order second-moments (FOSM), are typically used for assessing model response sensitivity around a specific point (Gaspar et al., 2016), but are limited by their dependence on this point. Global methods, on the other hand, provide a comprehensive sensitivity measure by considering the entire input space of model variables. They can account for interactions between input random variables and are not limited to a specific point, which makes them especially suited for complex problems with large number of random variables (Velarde et al., 2019). The variance-based global sensitivity analysis adopted herein provides a quantitative measure that represents the various input parameters to the output variance (Saltelli et al., 2010). Among the common methods for performing this type of analysis, Sobol’s approach (Sobol, 2001) was chosen for its ability to quantify variance -based sensitivity measures in complex systems where nonlinearity and/or interactions can significantly impact the results. The Sobol’s approach for quantifying the sensitivity measures is discussed in more detail in Sobol, (2001). 3. Fatigue Crack Propagation Prediction Approach Several models have been proposed to predict fatigue crack propagation in steel components. These include the Paris and Erdogan (1963) model based on linear elastic fracture mechanics (LEFM). The model calculates the crack propagation rate using specific material regression parameters and the stress intensity factor (SIF). In this model, the crack propagation rate under constant amplitude loading is (Paris and Erdogan 1963): = 0 (∆ ) 0 (1) where is the crack size, is the number of cycles, 0 and 0 are material regression parameters denoted, respectively, as the power law coefficient and Paris exponent, and ∆ is the range of the SIF. This model is applicable for predicting fatigue crack growth under LEFM condition, which assumes that the plastic zone size around the crack tip is negligible compared to the component size, representing a small-scale yielding (SSY) condition. The accurate estimation of the Stress Intensity Factor range, denoted as ∆K, is essential for accurately predicting the fatigue crack growth rate. While closed-form solutions can determine the SIF in simple structural elements (Tada et al., 2000), the intricacies of stiffened panel details necessitate a different approach. In this paper, the J -integral (Rice, 1968) is