Prediction of Tubular T/Y-Joint SIF by GA-BP Neural Network

In fracture mechanics, the fatigue evaluation depends on the precise provisions of the stress intensity factor (SIF), which is a quantitative factor established to evaluate the effect of stresses along the crack front. Stress intensity factor performs a definitive role in reflecting material fracture and estimating fatigue life of tubular joints. Some researchers have proposed a series of parametric equations to predict SIF values; however, it gave arguably inaccurate results. There is still a need to improve the accuracy of the SIF prediction’s equation. Thus, to overcome this shortage, this study introduced back-propagation (BP) neural network with conjunction to genetic algorithm (GA) (GA-BP) in predicting the SIF of cracked tubular T/Y-joint. To train and verify the results of the hybrid algorithm GA-BP SIFs’ database covers a wide variety of cracked tubular T/Y joints were simulated by ABAQUS and verified against experimental results. Meanwhile, the SIFs’ result produced by the GA-BP optimization method were compared with those SIF calculated from the accessible parametric equations. The comparison indicated that the GA-BP neural network optimization method is reliable, precise and capable tool in calculating the SIFs of cracked T/Y joints and it also provides higher accuracy than common methods.