Neural Networks Model for Settlement Prediction of Embankment

Abstract:

Based on the theory of artificial neural networks and back propagation algorithm, a model for predicting the settlement of embankment was proposed. Neural networks was proceeded with Matlab program. Combining with the real settlement data of Jieyang highway, the model was trained again and again, and then the parameters of model was calibrated. Comparing the prediction value of model with the observational data based on field measurement, it shows that the accordance of the predicted settlements by the proposed model with the measured data is better. As a branch of nonlinear science, neural network theory has strong practical value and advantages in embankment settlement engineering analysis and prediction, because of its strong ability to learn and its high accuracy to approach any nonlinear function.

• Periodical: Applied Mechanics and Materials (Volume 197)
• Pages: 722-726
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.197.722
• Online since: September 2012
• Authors: Qing Tao Bi, Shu Yun Ding
• Keywords: Back Propagation Algorithm (BPA), MATLAB Program, Settlement of Embankment

References

[1] X. L. Gong, Adcance soil mechanics, Zhejiang: Zhejiang University Press, April 1994, p.21–25.

[2] M. T. Hagan, H. B. Demuth, M. H. Beale, Design of neural networks, Beijing: Mechanical Industry Press, September 2002, p.89–91.

[3] W. Wang, Principles of artificial neural network, Beijing: Beijing Aeronautics and Astronautics Press, October 1995, p.12–14.

[4] Highway Embankment Design and Construction Specifications, Beijing: China Communications Press, (1997).

[5] L. Cui, Y.M. Dou, Application of artificial neural networks in displacement prediction of soft ground, Subgrade Engineering, Vol. 4, 2006, pp.11-14.