MACHINE LEARNING-AUGMENTED UNIVERSAL WEIGHT FUNCTION METHOD FOR STRESS INTENSITY FACTOR DETERMINATION

The accurate calculation of stress intensity factors (SIFs) constitutes a critical yet challenging task within linear elastic fracture mechanics. While the universal weight function method (WFM) has emerged as a prominent approach due to its high computational efficiency, its predictive accuracy is often constrained. This limitation arises from the difficulty in characterizing the nonlinear mapping relationships between the geometric dimensions of cracked bodies and the requisite weight function parameters. To address these challenges, this study introduces an innovative machine learning-augmented universal WFM. This method leverages Gaussian process regression (GPR) models to characterize the nonlinear mapping relationships between the geometric dimensions of cracked bodies and the weight function parameters, thereby enhancing the computational accuracy of the universal WFM. Validation cases demonstrate that the proposed method achieves superior accuracy compared to the traditional universal WFM, with the maximum relative error not exceeding 5.09%.