This paper discusses the experimental results obtained by using a principal component analysis based algorithm in joint with Self Organizing Maps (SOM) Neural Networks for detection of damages in pipeline structures. Also, a differential evolutive algorithm is used for tunig the neural network parameters. A pipeline section test structure was instrumented with an active piezoelectric system in order to apply a high known frequency signal and to determine the base-line structural dynamical performance. Several piezoelectric sensors were located along the structure surface and damage features are obtained by processing the time vibrational dynamical response through principal component analysis. Q-statistic and Hotteling T2 indexes obtained from the PCA analysis are used to detect deviations of the current vibrational response respect to the undamaged one. Algorithm validation was achieved by using experimental data, obtained from a carbon steel tubing section, where damages were induced by adding masses to the structure. The obtained results indicate that is possible to identify and locate faults in pipeline structures and by using evolutive differential genetic technique improve the performance of the studied algorithm