Part of the Advances in Cryogenic Engineering book series (volume 28)
A Neural Network Approach for Solving Inverse Problems in NDE
Solution to inverse problems is of interest in many fields of science and engineering. In nondestructive evaluation , for example, inverse techniques are used to obtain quantitative estimates of the size, shape and nature of defects in materials. Inv.:rse scattering problems in electromagnetics deal with estimation of scatterer information from knowledge of incident and scattered fields. Inverse problems are frequently described by Fredholm integral equations in the form
where u(x) represents the measured data, z(y) represents the source function or the system states or parameters, and k(x,y) represents the kernel of the transformation. The objective of inverse problem is then to solve for the source or state function from known measurements. This problem is sensitive to the system parameters z, to the shape of the kernel k, and to the accuracy of the measurements u.
$$ \smallint _a^bk(x,y)z(y)dy = u(x) (c \leqslant x \leqslant d)$$
KeywordsSine Hunt Eosine Geophysics Remote Sensing
Unable to display preview. Download preview PDF.
- 1.L. Udpa and W. Lord, “A Discussion of the Inverse problem in Electromagnetic Nondestructive Testing,” Review of Progress in Quantitative NDE, edited by D. O. Thompson and D. E. Chimenti (Plenum Press, New York, 1986), Vol. 5A, pp.375–382.Google Scholar
- 2.I. Elshafiey, L. Udpa and S. S. Udpa, “A Neural Network Approach for Solving Integral Equations,” Proceedings of IEEE Int. Symposium on Circuits and Systems, Singapore, pp. 1416–1419 (1991).Google Scholar
- 9.D. W. Marquardt, “Solution of Nonlinear Chemical Engineering Models.” Chemical Engineering Progress, Vol. 55, No. 6, p. 65–70 (1959).Google Scholar
© Springer Science+Business Media New York 1992