Files in this item

FilesDescriptionFormat

application/pdf

application/pdfBALTAJI-DISSERTATION-2021.pdf (20MB)Restricted to U of Illinois
(no description provided)PDF

Description

Title:Enabling technologies for pervasive sensing and deep learning in geotechnical engineering
Author(s):Baltaji, Omar
Director of Research:Hashash, Youssef
Doctoral Committee Chair(s):Hashash, Youssef
Doctoral Committee Member(s):Singer, Andrew; Olson, Scott; Ghaboussi, Jamshid
Department / Program:Civil & Environmental Eng
Discipline:Civil Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):Wireless communication
inverse analysis
constitutive modelling
Self
Abstract:The observational method has been historically employed to adapt to the numerous uncertainties encountered during geotechnical construction. It is based on measuring geotechnical systems performance, updating estimates of their future performance, and implementing changes to construction activities and design if needed. Underground sensing systems embedded in soil and rock formations and inverse analysis frameworks are powerful tools that allow engineers to better realize the potential of the observational method in the age of pervasive sensing and big data. Nevertheless, these tools suffer from important limitations: below ground sensing systems often rely on wired communication that is costly, vulnerable to damage, and of limited versatility; and inverse analyses are typically performed manually by ad hoc methods or computationally by sophisticated and time-consuming methods. In this study, advances are made in wireless through-soil communication and algorithmic development of the deep-learning-based inverse analysis simulation framework, SelfSim. A communication system, termed SoilComm, was developed, enabling communicating sensor data wirelessly through soil, and hence the deployment of underground sensors that are more economical, versatile, and robust than wired sensors. This system was tested in laboratory and field environments, and its performance was demonstrated by transmission of piezometer measurements and digital images. The latest prototype achieved a 10-m communication range with a power efficiency estimated to enable its batteries to last for around 3 years of operation. The deep-learning-based inverse analysis simulation framework, SelfSim (Self Learning Simulation), is re-implemented in a modern computational framework and is highly optimized for speed and computational efficiency, enabling much faster learning of geotechnical systems behavior. The performance of the new SelfSim framework was demonstrated through inverse analyses of simulated and real triaxial laboratory tests. SelfSim run times dropped significantly, completing some analyses in less than 5 minutes, compared to hours or days using prior implementations. In addition to their employment in geotechnical applications, both tools can potentially be employed in many other fields including agricultural, mining, petroleum, material, and bio engineering.
Issue Date:2021-12-01
Type:Thesis
URI:http://hdl.handle.net/2142/113975
Rights Information:Copyright 2021 Omar Baltaji
Date Available in IDEALS:2022-04-29
Date Deposited:2021-12


This item appears in the following Collection(s)

Item Statistics