Abstract
Abstract Crack assessment relies on the linear elastic or elastic-plastic fracture mechanics that requires calculation of stress intensity factor, K, in the fitness for service codes, such as API 579 and ASME BPVC Section XI. For a surface crack in a cylinder, the K calculation becomes calculating the influence coefficients G0 and G1 of K in those codes. API 579 provided accurate tabular data of G0 and G1 for selected cylinder sizes (t/Ri), crack aspect ratios (a/c), crack depths (a/t), and crack tip locations. Recently, the curve-fit solutions of G0 and G1 were obtained for surface cracks at the deepest and surface points. For an arbitrary cylinder size, however, three-parameter interpolations are still needed to estimate the G0 and G1. To avoid performing the complex interpolation, this paper adopts the state-of-the-art machine learning technology to develop data-driven K solutions based on the tabular data of G0 and G1 given in API 579 for axial outside semi-elliptical surface cracks in thick-wall cylinders at the deepest and surface points. The machine learning method utilizes an artificial neural network (ANN), activation function, and optimal learning algorithm to learn and to determine G0 and G1 as a function of the cylinder size (t/Ri), aspect ratio (a/c), and crack depth (a/t) for axial outside surface cracks at the deepest and surface points. The proposed data-driven solutions of G0 and G1 are validated by available curve-fit solutions for the axial outside surface cracks.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have