Real-Time Extraction of Stiffness of Layered Granular Soil Using Machine Learning Considering Dynamic Soil-Roller Interaction for Intelligent Compaction

Xu, Zhengheng

Real-Time Extraction of Stiffness of Layered Granular Soil Using Machine Learning Considering Dynamic Soil-Roller Interaction for Intelligent Compaction

Xu, Zhengheng

Permalink

Publication Type:: Thesis
Issue Date:: 2023

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Adobe PDF

Download thesisAdobe PDF (5.45 MB)

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Xu, Zhengheng
dc.date.accessioned	2023-10-17T21:04:49Z
dc.date.available	2023-10-17T21:04:49Z
dc.date.issued	2023
dc.identifier.uri	http://hdl.handle.net/10453/172743
dc.description	University of Technology Sydney. Faculty of Engineering and Information Technology.	en_US.UTF-8
dc.description.abstract	Intelligent Compaction (IC) has been acquiring a growing interest in real-time quality control of compacted soil layers because of its high efficiency and full-area coverage. However, accurate real-time determination of the soil modulus during compaction based on roller acceleration impacted by soil characteristics has been challenging due to the multi-layered composite nature of the soil and the nonlinearities of the dynamic equation of motion and soil response. The main goal of this study is to implement rigorous numerical modelling to simulate the multi-layered granular soil response subjected to cyclic loading, ranging from small to large strain amplitudes and account for stiffness degradation via adopting the finite element method. The benefits of adopting the finite element method are attributed to its capability to simulate interaction problems by achieving more realistic geometries. True scale three-dimensional models with the isotropic hardening elastoplastic hysteretic model were implemented to carefully simulate the boundary conditions and soil behaviour under cyclic loading. More than 5,000 sets of three-dimensional numerical simulations were run to cover a wide range of frequencies, amplitudes, weights, lift thicknesses and various states of granular soils. The results were adopted in conjunction with the newly proposed extended support vector regression (X-SVR) to establish a robust and accurate method for predicting the real-time unloading-reloading modulus of the compacted soil during compaction. The findings confirm that the adopted numerical models with the HS-Small constitutive soil model were able to evaluate the nonlinear stress-strain response of the soil subjected to cyclic loading, particularly variations of damping and soil stiffness with shear strain. Moreover, the proposed method could predict both the single and double-layered soil stiffness based on the X-SVR algorithm with the Gegenbauer kernel and Gaussian kernel, using the acceleration response of the drum and basic roller properties. Both training and testing unloading/reloading moduli, obtained from the machine learning method, correlated well with the 3D finite element predictions considering the nonlinear elastoplastic soil model and dynamic soil-drum interaction. This research can prove that the inverse solver developed in this study can predict the soil stiffness using the novel kernel-based X-SVR machine learning in a reasonably short time, required for real-time quality control by roller operators on the site. The findings of this study can be employed by practicing engineers to interpret roller drum acceleration data for estimating the level of compaction and ground stiffness during compaction.	en_US.UTF-8
dc.format	Thesis (PhD)
dc.language.iso	en_US	en_US.UTF-8
dc.relation	https://opus.lib.uts.edu.au/bitstream/10453/172743/1/thesis.pdf
dc.rights	info:eu-repo/semantics/openAccess
dc.rights	The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
dc.rights	© 2023 Zhengheng Xu
dc.rights	au.edu.uts.lib/cph
dc.title	Real-Time Extraction of Stiffness of Layered Granular Soil Using Machine Learning Considering Dynamic Soil-Roller Interaction for Intelligent Compaction	en_US.UTF-8
dc.type	Thesis
utslib.copyright.status	open_access	*

Abstract:

Intelligent Compaction (IC) has been acquiring a growing interest in real-time quality control of compacted soil layers because of its high efficiency and full-area coverage. However, accurate real-time determination of the soil modulus during compaction based on roller acceleration impacted by soil characteristics has been challenging due to the multi-layered composite nature of the soil and the nonlinearities of the dynamic equation of motion and soil response. The main goal of this study is to implement rigorous numerical modelling to simulate the multi-layered granular soil response subjected to cyclic loading, ranging from small to large strain amplitudes and account for stiffness degradation via adopting the finite element method. The benefits of adopting the finite element method are attributed to its capability to simulate interaction problems by achieving more realistic geometries. True scale three-dimensional models with the isotropic hardening elastoplastic hysteretic model were implemented to carefully simulate the boundary conditions and soil behaviour under cyclic loading. More than 5,000 sets of three-dimensional numerical simulations were run to cover a wide range of frequencies, amplitudes, weights, lift thicknesses and various states of granular soils. The results were adopted in conjunction with the newly proposed extended support vector regression (X-SVR) to establish a robust and accurate method for predicting the real-time unloading-reloading modulus of the compacted soil during compaction. The findings confirm that the adopted numerical models with the HS-Small constitutive soil model were able to evaluate the nonlinear stress-strain response of the soil subjected to cyclic loading, particularly variations of damping and soil stiffness with shear strain. Moreover, the proposed method could predict both the single and double-layered soil stiffness based on the X-SVR algorithm with the Gegenbauer kernel and Gaussian kernel, using the acceleration response of the drum and basic roller properties. Both training and testing unloading/reloading moduli, obtained from the machine learning method, correlated well with the 3D finite element predictions considering the nonlinear elastoplastic soil model and dynamic soil-drum interaction. This research can prove that the inverse solver developed in this study can predict the soil stiffness using the novel kernel-based X-SVR machine learning in a reasonably short time, required for real-time quality control by roller operators on the site. The findings of this study can be employed by practicing engineers to interpret roller drum acceleration data for estimating the level of compaction and ground stiffness during compaction.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/172743