A task decomposition neural network approach to non-destructive testing problems

A task decomposition neural network approach to non-
destructive testing problems
World Congress on neural networks, San Diego, California (United
States), Vol. 1, pp. 1,566-1.571. Lawrence Erlbaum Association
(1994). ISBN 080581745X
An artificial neural network technique to treat inverse problems,
typically encountered in Non-Destructive Testing applications, is
introduced. The basic architecture is proposed for an electrostatic test
problem. We also address the case of plural defect recognition, and we also
show how the method here described could be useful for more realistic NDT
problems, commonly known as eddy current testing. Simulations show that
our method can be very effective, particularly when a high accuracy of the
identification procedure is required.