Stress Concentration Factors in Tubular T-Joint Braces Under Compressive Loads Using Artificial Neural Networks

Stress concentration factors (SCFs) are often calculated using formulas based on experimental testing and finite element analysis (FEA). While maximum SCF could occur at any location along the brace axis of the tubular T-joint’s brace, only the SCFs at the crown and saddle points can be determined from the available formulae, which can result in imprecise fatigue life determination. The current study presents a methodology to determine the SCFs in T-joints using FEA and ANN. ANNs are more effective than conventional data-fitting techniques at modelling intricate phenomena. In this work, parametric equations to estimate the SCFs of the T-joint’s brace under compressive loading were developed. Utilizing parametric equations allows for rapid estimates of SCFs, in contrast to time-consuming FEA and expensive testing. The equations are based on an artificial neural network’s training weights and biases (ANN). 625 finite element simulations were performed on tubular T-joints with various dimensions under compressive loads to determine the SCFs at the brace of the T-joint. These SCFs were then used to train an ANN. The weights and biases of the ANN were subsequently used to derive equations for calculating SCFs based on dimensionless parameters. The equations can estimate the SCF of a T-joint brace with less than 7% error and a root mean square error (RMSE) of less than 0.19.