Predição da vida em fadiga de dutos danificados através de rede neural artificial

Abstract

This work develops a methodology to evaluate the fatigue life of pipelines with plain dents under cyclic internal pressure loading. The methodology uses the results of innite element analysis (FEA) to train an artificial neural network (ANN) that can be used to assess fatigue life. The study also analyzes the efect of introducing plain dents on the fatigue life of pipelines under cyclic loads during service and shutdown stages for diferent pipeline geometries, dent depths, indenter types, and steels of diferent API 5L grades. A 3D numerical finite element model of a pipe with a plain dent was validated through full-scale experimental tests with API 5L Gr B steel specimens. After validation, a comprehensive parametric study was performed on the main geometric parameters of the pipe and dent, comprising 240 FEM models. These parametric datasets were then used to train an artificial neural network (ANN) that can predict the stress concentration factor (SCF), the maximum plastic strain, and the residual dimensional parameters of the dent in the pipeline. The results show that ANNs can be used to predict these values with accuracy comparable to that achieved using FEM. The use of ANNs has the added benefit of achieving results in much less time than using MEF alone, with immediate results and feasible application across a whole system, including pipeline diagnostics and maintenance.