Neural Networks Applied to the Prediction of Flexible Pipes Armor Wires Configuration Under Pure Bending

Gabriel Mattos Gonzalez¹; João Paulo Ramos Cortina¹; Luis Volnei Sudati Sagrilo¹; José Renato Mendes de Sousa¹

¹ COPPE/UFRJ

doi:10.20906/CPS/CILAMCE2017-0433

Resumo

Flexible pipes are a feasible alternative to offshore oil and gas exploitation. High flexibility combined with high axial resistance compound their main characteristics. Local bending mechanics of these pipes usually treats their tensile armors as curves on a torus. Geodesic curves are one of the most utilized curves. It is not possible, however, to solve geodesic differential equations analytically, which imposes difficulties on the mechanical analysis. This work intends to train Artificial Neural Networks with data obtained from the numerical solution of these equations thus proposing an alternative solution for the stiff nonlinear differential equations for the special case of unbounded geodesic of a torus. The results showed good correlations for the whole set of cases presented.

Palavras-chave: Flexible Pipes; Neural networks; Finite element modeling