ID 112202
Author
Liao, Kuo-Wei National Taiwan University
Lin, Jhe-Yu Veterans Affairs Council
Keywords
scour
flood
complicated foundation
optimization
machine learning
Content Type
Journal Article
Description
A scour depth prediction formula for a river bridge is established using experimental data in which the effects of the pier, pile-cap and pile group are considered. More than 170 experimental data entries, including different pier structural sizes, flow depths and soil covering depths, are collected and verified by existing formulae, which failed to deliver a promising prediction. A machine learning prediction model was then developed to enhance the accuracy. For application purpose, a sequential quadratic programming optimization was adopted to construct an explicit prediction formula. The MAPE was significantly improved from 102.8 to 28.9. The results indicate that the proposed formula can simultaneously satisfy the requirements of accuracy and simplicity. The proposed formula has the advantages of being conceptually consistent with observed scour behaviors and provides a solid scour depth prediction, which is an important and critical step in the bridge safety evaluation if floods are considered.
Journal Title
KSCE Journal of Civil Engineering
ISSN
12267988
19763808
Publisher
Korean Society of Civil Engineers
Volume
22
Issue
7
Start Page
2241
End Page
2255
Published Date
2017-10-12
Remark
This is a post-peer-review, pre-copyedit version of an article published in KSCE Journal of Civil Engineering.The final publication is available at link.springer.com
Rights
©2018 Korean Society of Civil Engineers
EDB ID
DOI (Published Version)
URL ( Publisher's Version )
FullText File
language
eng
TextVersion
Author
departments
Science and Technology