The Novel Method of Structural Health Monitoring Using FEM and Neural Networks


Article Preview

In this paper, a new method of combining computational mechanics and neural networks for prediction of composite beam delamination is proposed. One beam with delamination, as well as a ‘healthy’ beam with no delamination, had a four-ply symmetric carbon/epoxy composite design, were fabricated simultaneously. The delamination was assumed at different location of the beam, and then the finite element analysis was performed and the modal frequencies of the composite beam were obtained, which were used to train the neural network. The piezoelectric patch was attached to the top of the composite beam to measure its modal frequencies. A feedforward backpropagation neural network was designed, trained, and used to predict the delamination location using the experimental modal values as inputs. The experimental results demonstrate that the predicted delamination location and size error is small.



Materials Science Forum (Volumes 475-479)

Main Theme:

Edited by:

Z.Y. Zhong, H. Saka, T.H. Kim, E.A. Holm, Y.F. Han and X.S. Xie




S. Zheng et al., "The Novel Method of Structural Health Monitoring Using FEM and Neural Networks", Materials Science Forum, Vols. 475-479, pp. 2099-2102, 2005

Online since:

January 2005




[1] L. Zou : Journal of Sound and Vibration Vol. 230(2000), p.357.

[2] C. H. Keilers Jr., and F. -K. Chang: Journal of Intelligent Materials Systems and Structures Vol. 6(1995), p.649.

[3] A.S. Islam and K.C. Cralg: Smart Materials and Structures Vol. 3(1994), p.318.

[4] H. Luo and S. Hanagud: AIAA Journal Vol. 35(1997), p.1552.

[5] R Roopesh Kumar Reddy and Ranjan Ganguli: Smart Mater. Struct. Vol. 12 (2003) p.232.

[6] Steve E Watkins et al.: Smart Mater. Struct. Vol. 11 (2002) p.489.