Damage Identification of Cylindrical Shells Based on Mode Shapes Curvature

Currently there are just a few papers about fault detection of tubes and cylindrical shells. Recently, methods based on modal curvatures have gained great attention due to their sensitivity to defects. But the methods require dense number of data extraction points which limits their industrial application. In this research a new method for damage identification of cylindrical shells has been introduced. The method does not need intact structure's data and in contrast to other methods requires a few data extraction points. By transferring the modal information of the structure to the ANN, output of the network is exact position of the defect on the structure and the method does not need skilled technician to interpret the data. Performance of the network is validated by unfamiliar data for the network and 0.97 regression is obtained