We addressed a methodology for detecting and locating defects or discontinuities on the outside covering of metal underground pipelines. The pipelines were either cathodically protected or non-cathodically protected. By applying a wide range of AC Impedance signals for different frequencies to a steel coated-pipeline and by measuring its corresponding transfer fimction under laboratory-simulated real
conditions, we can design an algorithm capable of studying the pipeline system and determine a specific pattern for monitoring under simulated
"real" conditions. Due to the nature of the system, AC response can be responsible for an incorrect interpretation of data. This work shows AC Impedance data for the possible different scenarios in a 3-D physical model laboratory test simulating an underground cathodic protected coated-pipeline. Level of cathodic protection, location of discontinuities (holidays) and severity of corrosion can be classified and predicted by training an Artificial Neural Network (ANN). An ANN was built and designed to train different Impedance data for experimental results
(transfer function) and was used to predict the exact location of the active holidays and defects on the buried pipelines.
Key Words: AC Impedance, underground pipelines, Artificial Neural Networks, training.