Improving predictions made by ANN model using data quality assessment

An application to local scour around bridge piers

Sung-Uk Choi, Byungwoong Choi, Seongwook Choi

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

The scour depth that develops around bridge piers is known to be related to flow intensity, particle size of bed sediment, and pier dimensions. Earlier approaches to this issue have mainly relied on empirical formulas. Even numerical simulations have not been so successful due to problems associated with interactions between water flow and streambed morphology. This necessitates the application of an artificial intelligence (AI)-type approach to understanding the effects of local scour around bridge piers. Although previous studies reported that AI-based models are better predictors, they do not predict field-scale local scour well. Motivated by this, the present study reports on the use of data quality assessment with an artificial neural network (ANN) model for predicting field-scale scour depth around bridge piers. Both univariate and multivariate methods were applied and the predicted results are compared. For the multivariate method, the Euclidean distance method and Mahalanobis distance method were used and the predicted results are compared. The ANN model was first trained and validated using laboratory data and the model was applied to data obtained in laboratory experiments. The model was then applied to field data. Quantitative descriptions are given on how much the data quality assessment improves predictions based on the use of the ANN model.

Original languageEnglish
Pages (from-to)977-989
Number of pages13
JournalJournal of Hydroinformatics
Volume17
Issue number6
DOIs
Publication statusPublished - 2015 Nov 1

Fingerprint

Bridge piers
Scour
pier
scour
data quality
artificial neural network
Neural networks
prediction
artificial intelligence
Artificial intelligence
Piers
water flow
Sediments
Particle size
particle size
method
Computer simulation
sediment
simulation
Water

All Science Journal Classification (ASJC) codes

  • Geotechnical Engineering and Engineering Geology
  • Atmospheric Science

Cite this

@article{74a4a77bf46949f693e54d317b08ff2d,
title = "Improving predictions made by ANN model using data quality assessment: An application to local scour around bridge piers",
abstract = "The scour depth that develops around bridge piers is known to be related to flow intensity, particle size of bed sediment, and pier dimensions. Earlier approaches to this issue have mainly relied on empirical formulas. Even numerical simulations have not been so successful due to problems associated with interactions between water flow and streambed morphology. This necessitates the application of an artificial intelligence (AI)-type approach to understanding the effects of local scour around bridge piers. Although previous studies reported that AI-based models are better predictors, they do not predict field-scale local scour well. Motivated by this, the present study reports on the use of data quality assessment with an artificial neural network (ANN) model for predicting field-scale scour depth around bridge piers. Both univariate and multivariate methods were applied and the predicted results are compared. For the multivariate method, the Euclidean distance method and Mahalanobis distance method were used and the predicted results are compared. The ANN model was first trained and validated using laboratory data and the model was applied to data obtained in laboratory experiments. The model was then applied to field data. Quantitative descriptions are given on how much the data quality assessment improves predictions based on the use of the ANN model.",
author = "Sung-Uk Choi and Byungwoong Choi and Seongwook Choi",
year = "2015",
month = "11",
day = "1",
doi = "10.2166/hydro.2015.097",
language = "English",
volume = "17",
pages = "977--989",
journal = "Journal of Hydroinformatics",
issn = "1464-7141",
publisher = "IWA Publishing",
number = "6",

}

Improving predictions made by ANN model using data quality assessment : An application to local scour around bridge piers. / Choi, Sung-Uk; Choi, Byungwoong; Choi, Seongwook.

In: Journal of Hydroinformatics, Vol. 17, No. 6, 01.11.2015, p. 977-989.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Improving predictions made by ANN model using data quality assessment

T2 - An application to local scour around bridge piers

AU - Choi, Sung-Uk

AU - Choi, Byungwoong

AU - Choi, Seongwook

PY - 2015/11/1

Y1 - 2015/11/1

N2 - The scour depth that develops around bridge piers is known to be related to flow intensity, particle size of bed sediment, and pier dimensions. Earlier approaches to this issue have mainly relied on empirical formulas. Even numerical simulations have not been so successful due to problems associated with interactions between water flow and streambed morphology. This necessitates the application of an artificial intelligence (AI)-type approach to understanding the effects of local scour around bridge piers. Although previous studies reported that AI-based models are better predictors, they do not predict field-scale local scour well. Motivated by this, the present study reports on the use of data quality assessment with an artificial neural network (ANN) model for predicting field-scale scour depth around bridge piers. Both univariate and multivariate methods were applied and the predicted results are compared. For the multivariate method, the Euclidean distance method and Mahalanobis distance method were used and the predicted results are compared. The ANN model was first trained and validated using laboratory data and the model was applied to data obtained in laboratory experiments. The model was then applied to field data. Quantitative descriptions are given on how much the data quality assessment improves predictions based on the use of the ANN model.

AB - The scour depth that develops around bridge piers is known to be related to flow intensity, particle size of bed sediment, and pier dimensions. Earlier approaches to this issue have mainly relied on empirical formulas. Even numerical simulations have not been so successful due to problems associated with interactions between water flow and streambed morphology. This necessitates the application of an artificial intelligence (AI)-type approach to understanding the effects of local scour around bridge piers. Although previous studies reported that AI-based models are better predictors, they do not predict field-scale local scour well. Motivated by this, the present study reports on the use of data quality assessment with an artificial neural network (ANN) model for predicting field-scale scour depth around bridge piers. Both univariate and multivariate methods were applied and the predicted results are compared. For the multivariate method, the Euclidean distance method and Mahalanobis distance method were used and the predicted results are compared. The ANN model was first trained and validated using laboratory data and the model was applied to data obtained in laboratory experiments. The model was then applied to field data. Quantitative descriptions are given on how much the data quality assessment improves predictions based on the use of the ANN model.

UR - http://www.scopus.com/inward/record.url?scp=85000461281&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85000461281&partnerID=8YFLogxK

U2 - 10.2166/hydro.2015.097

DO - 10.2166/hydro.2015.097

M3 - Article

VL - 17

SP - 977

EP - 989

JO - Journal of Hydroinformatics

JF - Journal of Hydroinformatics

SN - 1464-7141

IS - 6

ER -