Image Data-Driven-Deep Learning in Geosystems: Exploratory Investigation into the Stability of Retaining Walls

SPONSOR: NATIONAL SCIENCE FOUNDATION

PI:  Zhen Liu

 

Project Period:  09.01.17 – 08.31.19

The project is proposed to introduce the most recent breakthroughs in computer vision attributed to deep learning to address a rarely discussed yet urgent issue in most engineering disciplines: how to analyze the explosively increasing image data including images and videos, which can hardly be analyzed with traditional methods? Therefore, the proposed work is very time-sensitive. The core concepts enabling the breakthroughs in image recognition and adopted in AlphaGo, i.e., deep convolutional neural nets, will be used to explore the possibility to accurately assessing the safety of retaining walls with image data. The analysis will be validated against traditional methods including limit analysis and numerical simulation. The research is hypothesis-driven and rationally built on our preliminary study, which clearly supports the hypothesis and shows the high potential of the proposed study. In addition to further proving the power of the proposed concept, this collaborative effort between geotechnical engineering and computer science will understand the data and deep learning in geotech analysis, relate image patterns to physical mechanisms, and investigate two key potential issues associated with the methodology.

The proposed work will be organized into three tasks. Task 1 is to understand how the geotechnical data and analysis results can be related to the input and output of deep learning using convolutional nets. This is a key to the successful application of deep learning in geotech. In Task 2, eight factors in three categories, i.e., geology (soil property), topography (geometry) and boundary conditions, will be evaluated regarding their influence on the safety factors. The relationships between these factors, their roles and significance in deep learning, and correlation between the patterns identified by the CNN and the physical mechanisms will be compared and revealed to connect the deep learning to the existing geotech knowledge pool. In Task 3, two additional key aspects representing the unknowns and possible weaknesses of the machine learning method will be analyzed: model robustness and extrapolation. Testing data prepared in ways different from the input data, i.e., appearance and ranges, will be used to assess and improve these two aspects.

Zhen Liu
Zhen Liu
Shiyan Hu
Shiyan Hu

 

Develop and Implement a Freeze Thaw Model Based Seasonal Load Restriction Decision Support Tool

SPONSOR:  MICHIGAN DEPARTMENT OF TRANSPORTATION

PI:  Zhen Liu

Spring (or Seasonal) Load Restriction (SLR) policies that limit the axle loads of trucks have been implemented in many states of the United States and other countries to minimize costly roadway damage that occurs in seasonally frozen areas during the annual spring thaw and strength recovery period (Zarrillo et al., 2012). This is because concrete and asphalt, though look indestructible, can actually be quite fragile in late winter as frost comes out of the ground (County Road Association of Michigan).

The overall objective of the project is to establish a thawing model and a process for setting and removing seasonal load restrictions in a manner that will give industry the most amount of time to prepare for the restrictions and minimize the time to lift the restrictions based on the MDOT Project RC 1619. The overall objective will be accomplished through a series of objectives and tasks leveraging existing research, technology, and resources that MDOT already has in place.

  1. Evaluate existing thaw/freeze depth prediction models, practice for SLR in state DOTs and MDOT’s needs and available resources, and based on that, determine if existing thaw depth models suffice for application as a decision support tool for Michigan or if a refined model would be prudent.
  2. Identify the type, sources, and format of the soil and weather information used for analysis by the decision support tool.
  3. Building on this project and the research of RC 1619, develop a thaw depth model th!it utilizes the existing data sources in Objective 2.
  4. Identify locations for potential virtual Road Weather Information System (RWIS) sites and collect necessary data to implement those locations.
  5. Develop a user friendly decision support tool that could be easily utilized by public and private sector in estimating potential thaw conditions and setting of SLRs for any location on the MDOT road network.
  6. Recommend processes for predicting the time to post and remove SLR signs to protect the pavement structures from excessive damage during the spring thaw season.
  7. Identify opportunities to collect, present, and apply data and develop models to refine pavement designs.
  8. Develop professional training materials and course for training MDOT staff in the use of the decision support tool.
Zhen Liu
Zhen Liu
Stan Vitton
Min Wang
Min Wang

Phase Composition Curves in Frozen Porous Materials

SPONSOR:  MICHIGAN SPACE GRANT CONSORTIUM

PI:  Zhen Liu

The relationship between unfrozen water content and temperature, which is referred to as the Phase Composition Curve (PCC) in frozen soils, has long been observed. However, this relationship has not been extensively studied and widely used, possibly due to the lack of a physical understanding. Recent studies of the Pl succeeded in obtaining a physical description, a physically-based equation, and a physic-empirical prediction method for this relationship. Based on the common nature of porous materials, it is hypothesized that there is a relationship between unfrozen water content and temperature in all frozen porous materials. This study will experimentally investigate the existence of the PCC in typical porous materials. Optimization analyses will be conducted for the design of a Time Domain Reflectometry sensor. The TOR sensor together with thermal couples, which is suitable for the measurement of the PCC in the selected porous materials, will be fabricated and calibrated. The sensor will be utilized to measure the PCC by strictly following a specially designed procedure. The measured PCCs will be analyzed using the physically-based equation proposed by the PL The parameters in the equation will be obtained by means of curve fitting to the measured results. The values of the parameters for different porous materials will be categorized and compared.

The research will not only offer a definite answer to the wide existence of the PCC, but also obtain the characteristics of that of different porous materials. The research will provide a clear understanding of phase transition of water in porous materials which is currently absent. The resultant conclusions may advance many engineering applications involving the freezing process of porous materials. The research thus will lay down a necessary basis for the exploration in extraterrestrial environments, where both porous materials and the phase change of water or other liquid are very likely to exist. Also, this study will open a new research area for the PI and will answer a key question for preparing a solid proposal which will be submitted to the NSF.

Zhen Liu
Zhen Liu

Exploratory Investigation of Thermally-Induced Water Flow in Soils

SPONSOR:  NATIONAL SCIENCE FOUNDATION (NSF)

PI:  Zhen Liu

This project aims to answer a very fundamental yet very old scientific question: “Why and how does water move due to temperature gradients in porous materials?” This thermally induced water flux ubiquitously exists in porous materials, whenever both heat transfer and water movement are present. A scientific understanding of this phenomenon is an essential base for many important scientific and social challenges: climate effects on geomaterials, geothermal energy applications, behavior of porous materials under extreme conditions, and recovery of non-conventional fossil fuels such as gas hydrates and shale gas. However, despite the significance, this phenomenon has been an historically unsolved and perplexing issue affecting many science and engineering areas involving porous materials from traditional applications in civil engineering, soil science and petroleum engineering to emerging needs in microfluidics, material processing and biomechanics.

This award supports the exploration of a new research concept/methodology and its application to reveal the physical mechanisms underlying thermally induced water flux for a complete scientific description and analysis framework for this phenomenon. As an exploratory study, which pioneers a very high-risk but possibly high-return concept, the success of the study may provide the geotechnical community a new understanding to tackle many issues which are hard to solve in the existing frameworks, and also provide a way to integrate porous material research which is currently distributed in various disciplines. In addition to supporting a doctoral student, the project will support outreach activities for rural, low-socioeconomic students and native tribal communities in the Upper Peninsula of Michigan. An annual summer program will be established to engage K-12 students in hands-on-learning for understanding of porous materials.

Zhen Liu
Zhen Liu

 

Development of Advanced Ultrasonic Techniques for Air Void Size Distribution in Early-Age and Hardened Concrete

Sponsor:  Michigan Tech Transportation Institute (MTTI)

PI:  Zhen Liu, Qingli Dai

The air void size distribution has significant impacts on mechanical, thermal and transport properties of concrete and long-term durability such as freeze-thaw resistance. Measuring the characteristics of air voids in concrete (especially at early-ages) is thus very important in assessing its long-term durability. Nondestructive ultrasonic technique will be developed for potential concrete mixture quality control both in lab and field applications.

Research Objective:

1. Develop ultrasonic air void size distribution measurement techniques for hardened air-entrained concrete and evaluate the accuracy with ASTM C 457 measurements

2. Develop ultrasonic air void size distribution measurement techniques for early-age air-entrained concrete and evaluate the accuracy with the ASTM C 457 measurements

3. Develop the testing procedures and data processing tools for potential field applications

Implementation Plan:

1. Prepare air-entrained concrete samples with/without internal curing reservoirs (using light-weight fine aggregates) for air void distribution measurements

2. Design the ultrasonic measurement system for both hardened and early-age concrete sample measurements and develop the signal processing programs for the air void distribution evaluation of both types of samples

3. Evaluate the measurement accuracy by comparing with RapidAir ASTM C 457 and propose the testing procedures for potential field mixture quality control

Zhen Liu
Zhen Liu
Qingli Dai
Qingli Dai