Defect Identification in Metallic Walls using ANN and FEM

N.P. De Alcantara, Jr., A.M. De Carvalho, and J.A.C. Ulson (Brazil)


Metallic walls, metallic tubes, nondestructive evaluation, pattern recognition, artificialneural networks, finite element method.


This work presents an investigation on the use of the finite element method (FEM) and artificial neural networks (ANN) for the identification of defects on metallic walls (pipelines, metallic vessels, large metallic structures, etc.), due to the aggressive actions of the fluids contained by them, and/or atmospheric agents. The methodology used in this study consists in the simulation of a large number of defects in a metallic wall, considering its geometry and magnetic characteristics, by the finite element method. Both variations in the width and height of the defects are considered. Then, the obtained results are used to generate a set of vectors for the training of a perceptron multilayer artificial neural network. Finally, the obtained neural network is used to classify a group of new defects, simulated by the finite element method, but not belonging to the original dataset. The results on the simulated defects seem to support the proposed method, and encourage future works on this subject.

Important Links:

Go Back