Using and Improving Neural Network Models for Ground Settlement Prediction

Motohei Kanayama, Alexander Rohe, Leon van Paassen

Research output: Contribution to journalArticle

9 Citations (Scopus)

Abstract

Earth-fill structures such as embankments, which are constructed for the preservation of land and infrastructure, show significant amount of settlement during and after construction in lowland areas with soft grounds. Settlements are often still predicted with large uncertainty and frequently observational methods are applied using settlement monitoring results in the early stage after construction to predict the long term settlement. Most of these methods require a significant amount of measurements to enable accurate predictions. In this paper, an artificial neural network model for settlement prediction is evaluated and improved using measurement records from a test embankment in The Netherlands. Based on a learning pattern that focuses on convergence of the settlement rate, the basic model predicted settlements which were in good agreement with the measurements, when the amount of measured data used as teach data for the model exceeded a degree of consolidation of 69 %. For lower amounts of teach data the accuracy of settlement prediction was limited. To improve the accuracy of settlement prediction, it is proposed to add short-term predicted values that satisfy predefined statistical criteria of low coefficient of variance or low standard deviation to the teach data, after which the model is allowed to relearn and repredict the settlement. This procedure is repeated until all predicted values satisfy the criterion. Using the improved network model resulted in significantly better predictions. Predicted settlements were in good agreement with the measurements, even when only the measurements up to a consolidation stage of 35 % were used as initial teach data.

Original languageEnglish (US)
Pages (from-to)687-697
Number of pages11
JournalGeotechnical and Geological Engineering
Volume32
Issue number3
DOIs
StatePublished - Jan 1 2014
Externally publishedYes

Fingerprint

ground settlement
neural networks
Neural networks
prediction
Embankments
embankment
Consolidation
consolidation
observational method
infrastructure
artificial neural network
Netherlands
lowlands
fill
uncertainty
learning
Earth (planet)
Monitoring
monitoring
methodology

Keywords

  • Measurement record
  • Neural network
  • Observational method
  • Settlement prediction
  • Soft ground

ASJC Scopus subject areas

  • Architecture
  • Geology
  • Soil Science
  • Geotechnical Engineering and Engineering Geology

Cite this

Using and Improving Neural Network Models for Ground Settlement Prediction. / Kanayama, Motohei; Rohe, Alexander; van Paassen, Leon.

In: Geotechnical and Geological Engineering, Vol. 32, No. 3, 01.01.2014, p. 687-697.

Research output: Contribution to journalArticle

@article{299be02e600c44248caeb8b8a4a1ccdb,
title = "Using and Improving Neural Network Models for Ground Settlement Prediction",
abstract = "Earth-fill structures such as embankments, which are constructed for the preservation of land and infrastructure, show significant amount of settlement during and after construction in lowland areas with soft grounds. Settlements are often still predicted with large uncertainty and frequently observational methods are applied using settlement monitoring results in the early stage after construction to predict the long term settlement. Most of these methods require a significant amount of measurements to enable accurate predictions. In this paper, an artificial neural network model for settlement prediction is evaluated and improved using measurement records from a test embankment in The Netherlands. Based on a learning pattern that focuses on convergence of the settlement rate, the basic model predicted settlements which were in good agreement with the measurements, when the amount of measured data used as teach data for the model exceeded a degree of consolidation of 69 {\%}. For lower amounts of teach data the accuracy of settlement prediction was limited. To improve the accuracy of settlement prediction, it is proposed to add short-term predicted values that satisfy predefined statistical criteria of low coefficient of variance or low standard deviation to the teach data, after which the model is allowed to relearn and repredict the settlement. This procedure is repeated until all predicted values satisfy the criterion. Using the improved network model resulted in significantly better predictions. Predicted settlements were in good agreement with the measurements, even when only the measurements up to a consolidation stage of 35 {\%} were used as initial teach data.",
keywords = "Measurement record, Neural network, Observational method, Settlement prediction, Soft ground",
author = "Motohei Kanayama and Alexander Rohe and {van Paassen}, Leon",
year = "2014",
month = "1",
day = "1",
doi = "10.1007/s10706-014-9745-8",
language = "English (US)",
volume = "32",
pages = "687--697",
journal = "International Journal of Mining Engineering",
issn = "0263-4546",
publisher = "Springer Netherlands",
number = "3",

}

TY - JOUR

T1 - Using and Improving Neural Network Models for Ground Settlement Prediction

AU - Kanayama, Motohei

AU - Rohe, Alexander

AU - van Paassen, Leon

PY - 2014/1/1

Y1 - 2014/1/1

N2 - Earth-fill structures such as embankments, which are constructed for the preservation of land and infrastructure, show significant amount of settlement during and after construction in lowland areas with soft grounds. Settlements are often still predicted with large uncertainty and frequently observational methods are applied using settlement monitoring results in the early stage after construction to predict the long term settlement. Most of these methods require a significant amount of measurements to enable accurate predictions. In this paper, an artificial neural network model for settlement prediction is evaluated and improved using measurement records from a test embankment in The Netherlands. Based on a learning pattern that focuses on convergence of the settlement rate, the basic model predicted settlements which were in good agreement with the measurements, when the amount of measured data used as teach data for the model exceeded a degree of consolidation of 69 %. For lower amounts of teach data the accuracy of settlement prediction was limited. To improve the accuracy of settlement prediction, it is proposed to add short-term predicted values that satisfy predefined statistical criteria of low coefficient of variance or low standard deviation to the teach data, after which the model is allowed to relearn and repredict the settlement. This procedure is repeated until all predicted values satisfy the criterion. Using the improved network model resulted in significantly better predictions. Predicted settlements were in good agreement with the measurements, even when only the measurements up to a consolidation stage of 35 % were used as initial teach data.

AB - Earth-fill structures such as embankments, which are constructed for the preservation of land and infrastructure, show significant amount of settlement during and after construction in lowland areas with soft grounds. Settlements are often still predicted with large uncertainty and frequently observational methods are applied using settlement monitoring results in the early stage after construction to predict the long term settlement. Most of these methods require a significant amount of measurements to enable accurate predictions. In this paper, an artificial neural network model for settlement prediction is evaluated and improved using measurement records from a test embankment in The Netherlands. Based on a learning pattern that focuses on convergence of the settlement rate, the basic model predicted settlements which were in good agreement with the measurements, when the amount of measured data used as teach data for the model exceeded a degree of consolidation of 69 %. For lower amounts of teach data the accuracy of settlement prediction was limited. To improve the accuracy of settlement prediction, it is proposed to add short-term predicted values that satisfy predefined statistical criteria of low coefficient of variance or low standard deviation to the teach data, after which the model is allowed to relearn and repredict the settlement. This procedure is repeated until all predicted values satisfy the criterion. Using the improved network model resulted in significantly better predictions. Predicted settlements were in good agreement with the measurements, even when only the measurements up to a consolidation stage of 35 % were used as initial teach data.

KW - Measurement record

KW - Neural network

KW - Observational method

KW - Settlement prediction

KW - Soft ground

UR - http://www.scopus.com/inward/record.url?scp=84899953246&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84899953246&partnerID=8YFLogxK

U2 - 10.1007/s10706-014-9745-8

DO - 10.1007/s10706-014-9745-8

M3 - Article

VL - 32

SP - 687

EP - 697

JO - International Journal of Mining Engineering

JF - International Journal of Mining Engineering

SN - 0263-4546

IS - 3

ER -