Use of neural networks for the identification of damage in ship structures

• Main Author: Zubaydi, Achmad, 1959-
• Other Authors: Memorial University of Newfoundland. Faculty of Engineering and Applied Science.
• Format: Thesis
• Language: English
• Published: 2001
• Subjects: Shells (Engineering) > Mathematical models; Plates (Engineering) > Mathematical models; Neural networks (Computer science); Marine engineering; Newfoundland studies; University of Newfoundland
• Online Access: http://collections.mun.ca/cdm/ref/collection/theses4/id/173586

Thesis (Ph.D.)--Memorial University of Newfoundland, 2001. Engineering and Applied Science Includes bibliographical references : leaves 166-174. The occurrence of damage in a ship's structure especially at the connection between a longitudinal and a heavy transverse members of the side shell is unavoidable under all operating conditions. The damage does not generally result in the loss of ships, nevertheless, it is often the cause of costly repairs and replacements of hull structures. Therefore, damage should be identified at an early stage in order to prevent the development of a more significant damage. This study presents a procedure for the identification of damage occurrence in the side shell of a ship's structure using a neural network technique. The structure is modeled as a stiffened plate. -- An experimental study using modal testing techniques was carried out for measuring the time history of the random response of undamaged and damaged models. The damage was made using a hacksaw at several locations on the longitudinal faceplate near the transverse member. The random decrement signatures, and the auto and cross-correlation functions were obtained from the random response. -- A finite element model was developed to generate numerical acceleration frequency response functions for the model. Excellent agreement was obtained between the numerical and the experimental acceleration frequency response functions. The numerical and the experimental data were used for validating an identification technique using neural networks. The results of the present study show that one can use the random signature or the autocorrelation function for the random response to identify the extent and location of damage.