Share Email Print
cover

Proceedings Paper

Interpretation of nondestructive tests on unknown bridge foundations using artificial neural networks
Author(s): Glenn J. Rix
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

To aid in determining the depth of unknown bridge foundations, two artificial neural networks were trained to predict the length of piles from sonic mobility test data. Synthetic mobility data generated using a one-dimensional model of the foundation system was used to train the networks. To simulate bridge bents and pile caps, a massive element atop the pile was included in the model. The source and receiver positions were on top of the simulated bent or cap. The first network was trained using `raw' mobility data; the second network was trained using `enhanced' mobility data in which the response of the bent or cap was subtracted from the response of the whole foundation system to emphasize resonances associated with the pile tip. Both networks yielded accurate predictions of pile length when tested with additional synthetic data. The networks also yielded reasonably accurate predictions of pile length when tested with experimental mobility data from two bridges. improved artificial neural networks would likely result if they were trained using experimental, rather than synthetic, data.

Paper Details

Date Published: 12 May 1995
PDF: 11 pages
Proc. SPIE 2457, Nondestructive Evaluation of Aging Structures and Dams, (12 May 1995); doi: 10.1117/12.209384
Show Author Affiliations
Glenn J. Rix, Georgia Institute of Technology (United States)


Published in SPIE Proceedings Vol. 2457:
Nondestructive Evaluation of Aging Structures and Dams
Soheil Nazarian; Larry D. Olson, Editor(s)

© SPIE. Terms of Use
Back to Top