Journal of Japan Society of Civil Engineers, Ser. C (Geosphere Engineering)
Online ISSN : 2185-6516
ISSN-L : 2185-6516
Paper (In Japanese)
QUASI-REAL-TIME PREDICTION OF SEEPAGE FLOW BEHAVIOR IN RIVER LEVEE DURING FLOOD BY ARTIFICIAL NEURAL NETWORK USING DEEP LEARNING
Yuji TAKESHITAYusuke TORIGOE
Author information
JOURNALS RESTRICTED ACCESS

2020 Volume 76 Issue 4 Pages 340-349

Details
Abstract

 It is important to measure and predict seepage flow behavior in river levees in order to estimate the state of seepage failure in river levees. In this paper, a quasi-real-time prediction method of the water level of the foundation of the levee is proposed using a learned artificial neural network model based on the water level changes of the river and the foundation of the levee in the event of a flood. For this purpose, the changes in the measured water levels of the river and the foundation of the levee were trained by an artificial neural network model using deep learning method during past flood events. The usefulness and validity of the proposed water level prediction method were verified by using actual water levels measured at two first-class river levees at four flood events.

Information related to the author
© 2020 by Japan Society of Civil Engineers
Previous article Next article
feedback
Top