University of New Hampshire University of New Hampshire Scholars' Repository

Master's Theses and Capstones Student Scholarship

Spring 2011

Measuring residual strength of liquefied soil with the ring shear device

Jay Hargy University of New Hampshire, Durham

Follow this and additional works at: https://scholars.unh.edu/thesis

Recommended Citation Hargy, Jay, "Measuring residual strength of liquefied soil with the ring shear device" (2011). Master's Theses and Capstones. 828. https://scholars.unh.edu/thesis/828

This Thesis is brought to you for free and open access by the Student Scholarship at University of New Hampshire Scholars' Repository. It has been accepted for inclusion in Master's Theses and Capstones by an authorized administrator of University of New Hampshire Scholars' Repository. For more information, please contact [email protected]. MEASURING RESIDUAL STRENGTH OF LIQUEFIED SOIL WITH THE RING SHEAR DEVICE

BY

JAY HARGY BS Geology (Engineering Emphasis) Northern Arizona University, 1996

THESIS

Submitted to the University of New Hampshire in Partial Fulfillment of the Requirements for the Degree of

Master of Science in Civil Engineering

May, 2011 UMI Number: 1498960

All rights reserved

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. UMT Dissertation Publishing

UMI 1498960 Copyright 2011 by ProQuest LLC. All rights reserved. This edition of the work is protected against unauthorized copying under Title 17, United States Code.

ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 This thesis has been examined and approved.

(deceased)

Thesis Director, Pedro de Alba Professor of Civil Engineering, UNH

Mandar Dewoolkar Associate Professor of Civil Engineering, UVM

Jean Benoit Professor of Civil Engineering, UNH X 'T^U^

JeffiMy S. Melton Research Assistant Professor of Civil Engineering, UNH

s>f/3 / ao// Date DEDICATION

This thesis is dedicated to my advisor, the late Pedro de Alba (1939-2011). It has always been my intention to dedicate my work to him, but with his recent passing this dedication takes on even more meaning.

During my two-plus years of knowing Dr. de Alba, I have gained the utmost respect for him. I knew I was working with a great man after seeing his name referenced in many of the textbooks used during my coursework. It was only recently that I realized that he studied at UC-Berkeley under Dr. H. Bolton Seed, the founder of geotechnical earthquake engineering. Dr. Seed's success is due in part to Pedro's hard work. During his 33 years at UNH, Dr. de Alba continued on to make many more major contributions to geotechnical earthquake engineering and the liquefaction resistance of sands. He was a mentor and a role model to numerous students and colleagues who have become successful engineers in their own right.

Dr. de Alba was also an outstanding gentleman who showed great compassion in his work, and for his students; always asking about my family before we spoke of school work. His hard work and dedication to his profession and students were exemplary, and I am truly honored to have been guided by him. He is missed.

iii ACKNOWLEDGMENTS

This work was funded in part by the U.S. National Science Foundation (NSF)

(award number CMMI-0724080), and was part of a larger project directed by Dr. Mandar

Dewoolkar from the University of Vermont. He and his student Ian Anderson greatly

contributed to the NSF report used as a starting point for this document. Mr. Robb

Wallen and Dr. John McCartney of the University of Colorado were valuable help in

centrifuge testing and must be acknowledged. Ian, Robb, and I spent many long days in

the lab and nowhere near enough time enjoying the many amenities of Boulder.

Besides the thesis work, my success at UNH would not have been possible

without the help from Dr. Jean Benoit. I thank him and Dr. Pedro de Alba for their

direction and guidance, patience with my misuse of words, and general concern for my well-being and success. Dr. Jeffery Melton is thanked for stepping in as advisor on short notice. I thank Drs. Ray Cook, Jo Daniel, and David Gress for providing TA positions

during my stent at UNH. Shawn Wads worth, UNH Civil Engineering Technician, must also be mentioned. He is the go-to guy for help with anything lab related. Also, Steve

Jackson from Axis New England, and Rob Cinq-Mars and Bob Champlin from UNH helped with programming and machine modifications.

Last, but by far not least, I acknowledge my chief editor, wife, and best friend,

Sandy Hargy. I thank her for allowing me to ignore her, and our daughter Ruby, as I spent too much time with my books. We shared the same roof, but I have missed the past two years with them. Their love and support through it all have been invaluable.

iv TABLE OF CONTENTS

DEDICATION iii ACKNOWLEDGEMENTS iv TABLE OF CONTENTS v LIST OF TABLES vii LIST OF FIGURES viii ABSTRACT ix

CHAPTER PAGE

1. INTRODUCTION 1 Project Background 1 Causes and Effects of Liquefaction 2 Earthquakes 3 Collapse Theory 4 Static Liquefaction 6 Structural Failure of Volcanoes 8 Selection of Residual Strength Values 9 Back-Calculation Based on Failure Case Histories 9 Standard Laboratory Testing 11 Project Goal and Components 13 2. THE RING SHEAR DEVICE 16 Testing Machine 16 Sample Chamber 16 Top Ring 18 O-Rings 20 Drive Motor and Controller 23 Data Collection System 27 Rainer 27 Other Key Components 28 Machine Calibration 29 Machine Maintenance and Modifications 30 Sample Chamber, Top Ring, and O-rings 31 Rainer Construction 32 Other Modifications 33

v 3. MATERIALS TESTED 35 Sand 35 Index Properties 35 Shear Strength Properties 40 F-75 Specimen Preparation Methods 44 Pore Fluid 48 4. RING SHEAR TESTING 49 Machine Preparation 49 Constructing a Specimen 52 Sand Deposition 52 Saturation Process 53 Specimen Uniformity 55 Testing Procedure 55 Post Testing Procedures 58 Data Reduction 59 Special Tests 63 Shear Zone Determination 63 Lower Initial Vertical Stress 65 Pore Pressure Dissipation Tests 66 Testing with Holliston 00 Sand 67 5. COMPARISON TESTING 69 Centrifuge Tests 69 Centrifuge Model Setup 69 Centrifuge Specimen Preparation 72 Centrifuge Testing Procedures 73 Centrifuge Test Results 73 Modified Triaxial Tests 83 Modified Triaxial System Description 83 Modified Triaxial Specimen Preparation 84 Modified Triaxial Test Results 85 6. RING SHEAR RESULTS AND COMPARISONS 89 Ring Shear Results 89 Comparison of Test Results 92 Comparison with Back-Calculated Field Values 94 Comparison with Previous Ring Shear Results 100 7. CONCLUSIONS AND RECOMMENDATIONS 103 Conclusions 103 Recommendations 105 8. REFERENCES 108

APPENDIX A-MACHINE CALIBRATION RESULTS 115 APPENDIX B - F-75 SAND MATERIAL PROPERTY TESTING DATA 125 APPENDIX C-RESIDUAL STRENGTH TESTING DATA 130

VI LIST OF TABLES

Table 3.1. Summary of F-75 sand index properties 36 Table 5.1. Centrifuge testing conditions and rough coupon results 74 Table 5.2. Modified triaxial testing conditions and results 88 Table 6.1. Ring shear testing conditions and results 91

vii LIST OF FIGURES

Figure 1.1. Las Colinas Flowslide, El Salvador, (after Evans and Bent, 2004) 3 Figure 1.2. Coal-mine waste flowslide, British Columbia, (after Hungr et al., 2002) 5 Figure 2.1. A schematic of UNH ring shear device 17 Figure 2.2. Sample chamber, torque load cell, and pneumatic bladder 18 Figure 2.3. Top ring 20 Figure 2.4. Motion Planner cyclic motion code PROG23 25 Figure 2.5. Motion Planner monotonic motion code PROG24 26 Figure 2.6. Rainer and leveling gage 28 Figure 3.1. Grain-size distribution curves of F-75 sand 37 Figure 3.2. Summary of F-75 sand friction angle data 41 Figure 4.1. Ring shear testing sheet 50 Figure 4.2. Typical processed data from a ring shear test 57 Figure 4.3. Raw torque load cell data plot 60 Figure 4.4. Typical spread of residual strength values (Test 106) 62 Figure 4.5. Shear band on RS test specimen 64 Figure 4.6. Material strength gain with pore pressure loss, RS specimens 67 Figure 5.1. Typical centrifuge model configuration 70 Figure 5.2. Photograph of a centrifuge model installed on the swing platform 71 Figure 5.3. Typical data plot of rough coupon pulling force 76 Figure 5.4. Centrifuge tests results 78 Figure 5.5. Centrifuge rough coupon data residual strength ratio plot 78 Figure 5.6. Coupon force vs. effective stress (drive motor measurements) 80 Figure 5.7. Coupon force vs. effective stress (slip-ring measurements) 80 Figure 5.8. Coupon shear band 83 Figure 5.9. Modified triaxial system schematic (not to scale) 84 Figure 5.10. Typical modified triaxial test result 86 Figure 5.11. Modified triaxial test results 87 Figure 5.12. Modified triaxial residual strength ratio plot 87 Figure 6.1. Ring shear tests results 90 Figure 6.2. Ring shear data residual strength ratio plot 90 Figure 6.3. Exponential trends in testing results 93 Figure 6.4. Residual strength ratio comparisons 93 Figure 6.5. Comparison of back-calculated Sur with range of F-75 test result trends 95 Figure 6.6. Back-calculated residual strength ratios with F-75 test result trends 95 Figure 6.7. CPT based residual strength ratio comparisons 98 Figure 6.8. Previous ring shear results, Holliston 00 sand 101 Figure 6.9. Current ring shear results, F-75 sand 101

vin ABSTRACT

MEASURING RESIDUAL STRENGTH OF LIQUEFIED SOIL WITH THE RING SHEAR DEVICE by

Jay Hargy

University of New Hampshire, May, 2011

Natural and constructed slopes may contain zones of loose granular soils capable of liquefaction. Liquefied soils behave like heavy fluids and consequent rapid flowslides can produce great damage. The "residual strength" (Sur) of the liquefied soil can be estimated by back-calculation from field case histories; however, very little confirmation laboratory testing has been conducted thus far. A reliable laboratory measurement technique is needed to independently verify Sur values used for mitigation design.

A ring shear device (RSD) designed and built at the University of New

Hampshire (UNH) allows for residual strength testing under controlled strain rates and infinite total strain. The Surof a fine sand, "Ottawa F-75," was analyzed using the RSD.

These results were verified by comparison to residual strength values obtained by geotechnical centrifuge testing. This study indicates that the UNH RSD can be a reliable tool for estimating the residual strength of liquefied soil.

IX 1. INTRODUCTION

Project Background

It has long been observed that saturated loose sand subjected to shock or

earthquake loading liquefy due to elevated pore water pressure during soil consolidation.

The liquefied sand experiences a drastic loss of strength and behaves like a heavy fluid

with little resistance to flow. As long as the liquefied state persists, the soil can flow

down slopes in destructive landslides. The soil gradually regains its strength as excess pore water pressure dissipates.

Modeling liquefied soil behavior for risk studies and engineering design requires

adequate measurements of how shearing strength is lost, the liquefied soil's resistance to

shear deformation, and how soil strength is eventually regained through pore pressure dissipation. Unfortunately, there are no full-scale field measurements of residual strength to guide development of such models, and existing field case histories are limited to observing the final damage due to the liquefaction process.

Controlled laboratory measurements are desirable, but the onset of liquefaction in soil specimens in conventional laboratory tests is usually accompanied by such large strains and deformation that reliable strength measurements cannot be made. Therefore, new test methods need to be developed to obtain reliable measurements. This document presents a detailed description of the University of New Hampshire (UNH) ring shear device (RSD) and ring shear (RS) testing methods used to characterize the residual strength (Sur) of a liquefied sand. RS testing was conducted by the author, concurrent

1 with modified triaxial testing by Dr. Pedro de Alba and civil engineering senior, Kayla

Hampe, from UNH, as well as centrifuge testing at the University of Colorado, Boulder

(CU-Boulder) led by Professor Mandar Dewoolkar from the University of Vermont

(UVM).

Causes and Effects of Liquefaction

A simple definition of a liquefied condition is when a soil's effective stress (a') becomes equal to zero (Konrad and Watts, 1995). The concept of effective stress was

first presented by Karl Terzaghi (1925); the total normal stress (load per area) applied to a soil (atotai) is carried by the inter-granular connections of the soil skeleton and by the porewater within.

<*total = a' + u

Stated another way, the effective stress is equal to the total stress minus the porewater pressure (u).

a' = atotai - u

In other words, fluids in a saturated soil carry a portion of the weight. In theory, when the porewater pressure increases to equal to the total stress (excess pore pressure equals initial effective stress) liquefaction occurs at a loss of all soil strength. However

Castro (1969) showed that, even after liquefaction, sands retain a significant resistance to shear deformation. This strength has been referred to as the undrained steady-state shear strength (Poulos et al., 1985), the undrained residual shear strength (Seed, 1987), the undrained critical shear strength (Stark and Mesri, 1992), the minimum undrained

2 strength (Konrad and Watts, 1995), and the shear strength of liquefied soils (Stark et al.,

1998). The term "residual strength" (Sur) is used herein simply for brevity.

Liquefaction can be triggered by many events and ensuing flow behavior can inflict great economic and emotional damage. Several studies have shown that the residual strength of a liquefied soil is not dependent on triggering conditions (monotonic or cyclic) (Casagrande, 1965; Castro, 1975; and Ishihara, 1993).

Earthquakes

Cyclic ground shaking during earthquakes is the cause most people associate with soil liquefaction. Evidence of liquefaction can be found after almost all major earthquakes.

In El Salvador (Figure 1.1), a steep slope, approximately 450 feet high, of loose sandy soil derived from weathered volcanic rock was stable under static conditions

3 (Evans and Bent, 2004). The steepness of the slope (32°), was indicative of its

considerable shear strength. A 7.6 magnitude earthquake occurred in January, 2001

during the dry season, following a drier than average wet season. The hillside was not

fully saturated; however, the volcanic material was underlain by a paleosol that acted as

an aquitard. The soil directly above the aquitard was saturated and liquefied due to the

earthquake loading and slipped off the hill carrying the soil above with it, leaving an

almost chair-like failure surface.

An estimated 4.6-million cubic feet of soil flowed down slope like a heavy fluid.

Witnesses said that the flowslide occurred suddenly and extremely rapidly. The slide ran

out on a basically flat slope (3°) to just over 2,400 feet in about 45 seconds, indicating an

average velocity of about 35 mph (1,565 cm/sec). It was evident that the low residual

strength of the liquefied soil allowed for both initial slope movement and long runout.

This one flowslide destroyed numerous homes and took approximately 585 lives (Evans and Bent, 2004).

Collapse Theory

Another trigger of soil liquefaction, common in coal-mine waste piles, is collapse of the soil structure. Coal-mine waste is most often end-dumped out of haul trucks over the edge of a pile. As the material cascades down the pile slope, it segregates into steeply dipping layers of rock, sand, and fines. The material is deposited in a loose condition and, as it is covered by more waste, weak zones are created within the interior of the pile.

The angle of repose of a waste pile is usually greater than the effective friction angle of the material (Dawson et al, 1998).

4 If one soil element is deposited such that it is on the verge of collapse, a minute

change in stress conditions, or porewater pressure, or the rumbling of a truck can cause

that soil element to collapse, triggering a failure. When this happens, the load is

transferred to the porewater within the soil skeleton, spontaneously generating excess

pore pressure. Because water cannot resist shear, the load is transferred to adjacent soil,

which also fails, leading to progressive failure of the slope (Hungr et al., 2002). Finite

element analysis suggests that even if the collapse process is confined to a small domain

within the pile, catastrophic displacements and accelerations result (Gu et al., 1993). The

collapse process is rapid enough that excess pore pressure dissipation is retarded, even

with soil permeability in the order of lxlO"3 to lxlO"2 cm/s (Dawson et al., 1998). Figure

1.2 shows the effects of liquefaction due to mine waste collapse.

Figure 1.2. Coal-mine waste flowslide, British Columbia, (after Hungr et al., 2002)

Failures of this type are usually preceded by significant rainfall, but fully saturated conditions are not necessary for collapse to occur. It has been shown that a

5 saturation level of 85% is sufficient to promote excess pore pressure during soil collapse

(Dawson et al, 1998).

Failures of coal mine-waste piles can be slow to start, with crest displacements in

the order of 3 to 16 feet per day. Visible bulging of the pile face at mid-slope is often

noted before "explosive" movement of the toe area and extremely rapid flow-like

movement of the failed material. The flowslide shown in Figure 1.2 is estimated to have reached speeds of 75 to 100 mph (3,350 to 4,470 cm/s) (Hungr et al., 2002).

Once the slide overrides the toe of the pile, the rapid loading generates excess pore pressures in the native soil. If it liquefies, the flow can be carried for great distances

on a layer of liquefied soil, as shown in Figure 1.2.

Given the prevalence of coal mining throughout the world, there is a significant potential for risk. For example, up to 30 percent of active coal-mine waste dumps in

British Columbia have a high potential for runout flowslide hazards. Flowslides initiated by static processes are not well understood and thus are often not adequately considered during design (Dawson, 1998).

Static Liquefaction

Liquefaction of soils and rapid flowslides can occur after a long duration of heavy rainfall. After 16 years of direct observation of numerous flow type landslides in the lower Himalayan region, it was found that not a single landslide was caused by the many, typically severe earthquakes of the region (Bayan, 2008). In fact, prolonged and intense rainfall is the most prominent landslide trigger, a universal landslide survey held in the year 2003 noted that 90% of landslides were activated this way (Orfano, 2009).

6 Static liquefaction can occur after saturation of a soil underlying an impervious

layer, if the soil is in an undrained state. Under such conditions, long periods of heavy

rainfall (i.e. exceeding 2.4 to 3.0 inches per hour and lasting for 2 to 3 days) can elevate

porewater pressure and cause the shear strength of the slope to become lower than the

driving shear stress (Bayan, 2008).

The factors affecting static liquefaction of a liquefiable mass are: the ratio of the undrained steady state strength to the driving shear stress; the strain required to reach the peak undrained soil strength at the in situ void ratio; and the rate at which the peak undrained strength is lost with continued strain. Therefore, the soil type, initial soil

structure, and the driving shear stresses affect the intensity and duration of rainfall necessary to trigger liquefaction (Bayan, 2008). The presence of a static driving shear

stress on a horizontal plan is beneficial for relatively dense soil, where dilative shear behavior is common. However with loose soils (which tend to contract when sheared), the presence of a static driving shear stress can decrease the resistance of the soil to the initiation of liquefaction (Seed and Harding, 1990)

Static liquefaction of soil has also been documented to occur in river beds in some regions of the mid-western United States (Chellis, 1951). When a river bed that may be dry for much of the year fills, it has been noted that the river bottom may liquefy to a distance equal to the depth of water. Even shale bottoms may liquefy. Scour does not occur because no material is removed, but there is loss of load-carrying capacity and lateral support for pile foundations.

7 Structural Failure of Volcanoes

All three of the previously mentioned trigger mechanisms can occur on volcanoes,

and massive slope failures have been recognized on more than one out of six Holocene

volcanoes; although only a limited number of sites have been investigated worldwide. In

the well-studied region of Japan, debris avalanches have been documented at more than

40% of the volcanoes (Siebert, 2002). Liquefaction of soil plays a key role in the speed

and runout distance of devastating flowslides associated with volcanoes.

Volcanoes are inherently unstable due to steep, mechanically unsound slopes,

surface loading by new eruptions, hydrothermal alteration causing reduced permeability, thermal fluid pressurization and slope over-steepening associated with magma intrusion,

and volcanic seismicity. Many volcanic flowslides are deep seated and involve immense volumes of material (from less than 2.4 to over 2,400 cubic miles) that can travel far

distances (less than 6 to over 600 miles) at rapid speeds [estimated by back-calculation to range from 100 to 335 mph (4,470 to 15,000 cm/s)] (Siebert, 2002). High velocities can quickly be attained on the steep volcanic slopes. The pronounced increase in mobility, relative to non-volcanic flowslides, is attributed to the availability of hydrothermal and magmatic fluids and to a greater percentage of fragmented material which facilities interaction between fluids and soil partials (Siebert, 2002).

Flowslides caused by the eruption of Mt. St. Helens in 1980, the most-studied volcano in the United States, had initial velocities in the range of 150 to 175 mph with average velocities approximately 75 mph (3,350 cm/s). The trigger for the eruption was a

5.1 magnitude earthquake that caused the largest debris avalanche in recorded history when the north side of the mountain, 7/10 of a cubic mile of material, slipped away. The

8 release of overburden pressure caused the eruption. The debris flow traveled over 16

miles, covering a 23 square mile area with an average of 148 feet of relatively

homogeneous debris. Dewatering of the slide led, 5 hours later, to mudflows, one of

which traveled all the way to the Columbia River almost 50 miles away, destroying

bridges and highways along the way (Siebert, 2002).

Seattle, Washington is about 58 miles from Mt. Rainier; and Portland, Oregon is

about 47 miles from Mt. Hood. These two volcanoes are possible triggers of rapid

flowslides of liquefied material that have the potential to impact roughly 6.6 million people. There are even greater threats worldwide.

Selection of Residual Strength Values

The selection of residual strength values for analysis and design is a difficult and controversial topic in geotechnical engineering (e.g., Kramer, 1996; Finn, 1998; Stark et al., 1998).

Back-Calculation Based on Failure Case Histories

The current state of practice is to rely on Sur values back-calculated from field case histories of failure that have been related to best estimates of penetration resistance in the liquefied zones. Unfortunately, there is a wide scatter in values due to the uncertainties associated with each case, and with the methods of analysis themselves, which led to significantly different values of residual strength back-calculated from case histories (e.g. Castro, 1995; Stark et al, 1998; Olson and Stark, 2002). Examples of uncertainties include:

9 • assumed limits of the zone of liquefaction, • shear strengths of non-liquefied zones, • location of initial and final sliding surfaces, • location of phreatic surfaces, • potential drainage or pore water redistribution occurring during flow, • hydroplaning, and • kinetics.

Limit-equilibrium or sliding block type back-analysis can be performed on flow

failure or lateral spread case histories. The back-calculated residual strength can then be

related to field test indices such as Standard Penetration Test (SPT) blowcounts (Seed,

1987; Davis et al., 1988; Seed and Harder, 1990; Stark and Mesri, 1992; Ishihara, 1993;

Baziar and Dobry, 1995; Konrad and Watts, 1995; Olson and Stark, 2002), cone penetration resistance (Ishihara, 1993; Jefferies et al., 1990; Robertson, 1990; Olson and

Stark, 2002), shear wave velocity (Fear and Robertson, 1995), or vane shear resistance

(Charlie et al., 1998). Consequently, published case history-based relationships between residual strength and penetration resistance (from SPT or cone penetration testing) are employed in most seismic evaluations.

Depending on the quality and amount of information available from a case history, the following limit equilibrium-based methods have typically been used in back- analysis: (1) pre-failure geometry (e.g., Seed, 1987); (2) post-failure geometry (e.g.,

Seed, 1987; Ishihara et al., 1990; Olson and Stark, 2002); and (3) both pre- and post- failure geometries with the kinetics of failure (momentum) (Davis et al., 1988; Castro,

1995; Olson and Stark, 2002). The pre- and post-failure geometries result in upper and

10 lower bound estimates of residual strength and the kinetics approach generally results in

values that are between the two.

For instance, even for the Lower San Fernando Dam, the most thoroughly

documented case history of a flowslide, the best estimate of Sur back-calculated for the

liquefied layer varies from 5 to 36 kPa (0.75 to 5.25 psi) (Davis et al., 1988, Seed et al.,

1989, Olson and Stark, 2002). Thirty seconds after the 1971 magnitude 6.6 San Fernando

earthquake, large blocks of soil comprising the upstream slope and crest of the dam

essentially floated into the reservoir on a layer of liquefied sand (Seed and Harder, 1990).

Olson and Stark's (2002) best estimate of residual strength for the Lower San Fernando

Dam case history was 18.7 kPa (2.7 psi), considering kinetics.

The impact on design of the chosen value of residual strength is obvious; directly

affecting the outcome of the analysis, and hence the extent of remedial measures and

associated costs.

It should further be noted that, although the post-failure analyses previously

described give a rough overview of the final consequences of failure, they do not provide

insights into the actual behavior of a liquefied material as it deforms and develops large

strains. Understanding this behavior is essential to properly model the flow of liquefied

soil around fixed structures, such as pile supported wharves, or to calculate runout from

flowslides, for example.

Standard Laboratory Testing

The selection of residual strength for design has been be done by using laboratory testing procedures under controlled conditions on various undisturbed and reconstituted

11 specimens (Poulos et al., 1985, Castro et al., 1992, Byrne et al., 1994, Vasquez-Herrera

and Dobry, 1989; Ishihara, 1993; Vaid and Sivathayalan, 1999)

The laboratory testing approach is feasible, but a number of studies suggest that

liquefied soil requires large strains (such as those developed by flowslides) to reach its

minimum residual strength value (Bryant et al., 1983; Eckersley, 1990; de Alba and

Ballestero, 2004, 2005, 2006). Unfortunately, most geotechnical laboratory equipment in

current use is very limited in the amount of strain that can be imposed, and cannot

reproduce the high shear strain rates and large shear strains typical of a sliding mass in

the field. Thus the applicability of most lab results is debatable.

To approach this problem under controlled laboratory conditions, de Alba and

Ballestero (2004, 2005, 2006) studied the behavior of a sphere as it was pulled through

liquefied sand specimens in a modified triaxial chamber, using a deadweight

arrangement. This test series showed that the liquefied sand behaved as a viscous non-

Newtonian fluid, and exhibited a viscosity which decreased with strain rate (de Alba and

Ballestero, 2004, 2005, and 2006). Analysis was complicated however, by the fact that at higher velocities the Reynolds number for the falling sphere indicated a transition to turbulent flow. Further testing using a thin square coupon (thin plate) made the results hydrodynamically easier to analyze because the flow remained laminar over a much larger range of velocities. While minimum residual strength can readily be determined in these types of triaxial experiments, the test is limited in that the strain rate cannot be controlled and total strain is limited.

None of the test methods mentioned above can determine how the liquefying materials loses, and subsequently regains, resistance as pore pressure changes. Such an

12 experiment requires a specimen that does not collapse after the minimum residual

strength is reached so strength recovery can be measured (assuming that realistic drainage

boundary conditions could be simulated).

Project Goal and Components

A reliable measurement technique for independently verifying residual strength is

obviously needed. Centrifuge or shake table tests can be considered as "ideal case

histories" if conducted carefully and are well-documented, but back-calculation of

residual strength from induced liquefaction failures is still not straightforward if typical

analysis techniques are applied.

With an appropriate measurement technique; however, a liquefying centrifuge model would be an ideal "field experiment" for observing the evolution of shearing resistance as earthquake-induced pore pressures change. It was postulated that the shear

strength of liquefying sand can be measured in-flight in a seismic geotechnical centrifuge model using a thin coupon (flat plate) pulled horizontally through the soil model

(Dewoolkar et al., 2010). The large strains and strain rates associated with flow failures would be simulated by moving the coupon relative to the sand before, during, and after earthquake simulation. By measuring the drag force on the coupon, it would be possible to observe the evolving soil shear strength, as it decreases to a minimum and subsequently increases as excess pore pressure dissipates. The centrifuge models could provide realistic field-scale stresses and boundary conditions, and a relatively dense array of instrumentation could facilitate observations of the strength changes in the liquefying sand.

13 The centrifuge test results could then be used to validate companion ring shear and modified triaxial testing. It was envisioned that the combined results of the centrifuge and small-scale laboratory experiments would provide guidance on how to simulate the large-scale tests in smaller laboratory apparatus. The mechanisms and instrumentation used for the centrifuge study could then be adopted by other physical modeling (e.g. centrifuge and shake table) facilities, and the small-scale tests could be used to extend the study of liquefaction behavior to other soil types (silty and clayey sands), and for specific engineering design purposes.

This research project involved three types of laboratory testing: centrifuge, ring shear and modified triaxial, all using the same "soil". The following are brief descriptions of the various components of the overall project:

• Seismic centrifuge tests on models of level ground saturated sand were conducted at the University of Colorado, Boulder (CU Boulder), with thin coupons pulled horizontally through liquefiable sand before, during and after induced shaking. The coupon pulling force was measured along with acceleration and pore pressure in the soil; permitting parametric studies on the effects of coupon speed and sand relative density. The centrifuge tests results were used to compute residual strength values and to analyze recovery of strength as excess pore pressures dissipated.

• Ring shear tests were conducted on saturated sand specimens liquefied under cyclic load at the University of New Hampshire (UNH). Residual strength was measured under strain rate controlled monotonic rotation.

• Modified triaxial tests, similar to the ones described earlier, were conducted at UNH.

• The residual strengths from the ring shear and modified triaxial tests were compared to those determined from the centrifuge tests for validation.

It is expected that the combined results of these experiments will significantly contribute to the modeling of the residual strength of liquefied sand, clarifying its

14 dependency on key factors such as strain rate and partial drainage. This in turn will permit more accurate simulation of problems such as estimating the forces exerted by liquefied soil on obstacles like pile-supported structures, and the prediction of flowslide behavior in general. These results are also expected to give designers an enhanced understanding of how to choose residual strength values.

It should be noted that much of the data collection was conducted using the

United States Customary System of measurement. However, the International System of

Units (SI) was used for presentation of data in a report to the NSF by Dewoolkar et al.

(2010). An attempt has been made to include both English and SI measurement in the following text, but the graphs are usually in SI units only.

15 2. THE RING SHEAR DEVICE

Ring shear tests were conducted using the strain rate-controlled ring shear device

(RSD) designed and built at UNH under the direction of Dr. Pedro de Alba and Dr. Barry

Fussell. The RSD can apply both cyclic and monotonic strain, via a shearing ring, to the top surface of a ring-shaped specimen; and can reproduce the rapid shear strain rates and

large shear strains typical of a sliding mass in the field. The following is a description of the UNH RSD. Additional design and useful information is given by Sandoval (2007),

and Sandoval et al. (2010). Figure 2.1 is a schematic of the testing machine.

Testing Machine

The RSD consists of several main components including: the specimen chamber, the top ring, the o-rings, the drive motor and controller, the rainer, and the data collection

system. Figures 2.2 and 2.3 are photos showing key components of the testing machine.

Sample Chamber

The sample chamber is an annular-shaped, anodized aluminum unit with a weep hole, covered by a porous stone, on the chamber bottom for specimen saturation and backpressure control. The chamber bottom is sloped at a 10-degree angle from the center, with the outer part of the specimen being thicker. In theory, this distributes the vertical shear strain uniformly; however, special testing (discussed in Section 4) indicates that this may not necessarily be the case.

16 Top plate

Mid plate

Main frame

Backpressure chamber

table Pneumatic bladder Figure 2.1. A schematic of UNH ring shear device.

The inner and outer diameters of the sample chamber were measured in three

locations approximately 60° apart with calipers that have a sensitivity of 25.4 |im (0.001 inch). The sample chamber has an average inner diameter of 20.66 cm (8.135 inch), an average outer diameter of 31.03 cm (12.215 inch), and a sample top surface area of 420.7 cm2 (65.21 in.2).

17 The sample chamber is connected to a rigid bottom plate by a torque/thrust load

cell. A pneumatic bladder between the bottom plate and machine table is used to press

the sample chamber against the top ring, creating a vertical force and confining stress on

the specimen. These features are shown in Figure 2.2.

A solid center shaft at the center of the sample chamber is used to align and

stabilize the top ring for testing. The top ring is lowered for testing and securely fastened

at a mid plate by two lateral shafts with in-line load cells that measure vertical loads.

Top Ring

The top ring is also an annular-shaped, anodized aluminum unit. It is lowered into the sample chamber for testing and applies shear to the top of the specimen. A weep hole is present (for specimen saturation) and is covered by a porous stone. Grooves, designed for o-rings that create a seal between the top ring and sample chamber, are

18 located near the base of the ring, near the specimen. The top ring is slightly smaller than

the sample chamber, with inner and outer diameters of 20.85 cm (8.207 inch) and 30.83

cm (12.139 inch), respectively. These dimensions were also determined by use of caliper

and give a total shearing area of 405.6 cm2 (62.83 in.2). Figure 2.3 is a photograph of the

top ring.

Cloth-backed sandpaper is adhered to the top ring to create a rough shearing

surface. A template of the top ring is used to mark the inner diameter and porous stone

location on a standard 12-inch round sandpaper disk. The sandpaper is then cut to shape

using heavy-duty shears and a pen-like utility knife. The hole for the porous stone can be

created using the utility knife, or using the hemispherical head of a ball-peen hammer and

a hole in a metal plate that is the same size as the porous stone. In the second method, the marked location is aligned with the hole and the hammer is used to punch out the porous

stone location. Minor additional trimming of the inner and outer edges of the sandpaper

is usually necessary after it has been adhered to the ring, to avoid rubbing on the sample chamber.

For this study, 40-grit sandpaper was used with an effective particle size of 0.3 mm (Sandoval et al., 2010). Actual 40-grit sand is larger than 0.3 mm, but the particles do not entirely protrude from the sandpaper matrix. A finer grit sandpaper, with an effective particle size around 0.2 mm, may be more appropriate for testing the F-75 sand.

The 40-grit sandpaper was used to be consistent with previous testing done with the machine.

A center shaft housing unit is incorporated with the top ring to align it with the sample chamber for testing. Linear bearings are encased within the housing and create a

19 relatively frictionless connection between the center shaft and housing unit. The center

shaft housing is connected to the drive motor by a coupling designed to transmit only torsional loads.

Figure 2.3. Top ring.

O-Rings

Both inner and outer o-rings are required between the top ring and sample chamber to maintain fluid pressure within the specimen. The friction (torque) caused by the o-rings is measured separately so it can be subtracted from the test result to determine the torque due solely to the shearing reaction of the specimen. O-ring friction is

20 measured with only water in the sample chamber using the same set of o-rings, inner and

outer, used during the test. The same monotonic motion program used for testing is used

for friction measurements.

For this study, torque measurements of the o-rings were generally conducted with

only a small amount of fluid pressure inside the sample chamber. A few measurements

conducted with pore pressures varying up to that typically used during testing showed no

difference due to chamber pressure.

The frictional resistance of the o-rings varies with shearing speed, number of uses of the o-ring, type of lubricant, amount of sand entering the o-ring grooves, location where the o-ring contacts the sample chamber, and minor inherent manufacturing defects of the o-rings. Due in part to these variables, o-ring friction needs to be measured before and after each test.

Shearing Speed. O-ring friction generally increases with speed, with an average torque of 800 inch-pounds at 5 revolutions per minute (rpm), 926 inch-pounds at 10 rpm,

1,009 inch-pounds at 15 rpm, and 1,099 inch-pounds at 20 rpm noted during this study.

Usage. The o-rings would become stretched during repeated use, but no trend in friction could be found related to the amount of use of the o-rings. Very generally, new rings had lower friction readings than older rings, but there was not a constant gradient in between. It was determined that one set of o-rings could be used for two tests before they became too stretched.

21 With a new set of rings, the inner o-ring is smaller than the o-ring groove and inner diameter of the sample chamber. Because of this, the inner o-ring has a tendency to get caught between the inner sample chamber wall and top ring. Care is needed to ensure that this ring is centered before lowering the top ring.

Lubricant. The type of lubricant used on the o-rings has a significant effect on friction values. A lubricating silicone grease was initially used, but due to very high friction values recorded with the grease (generally over 1,500 inch-pounds), several combinations of grease, oil, spray-on lubricant, and graphite powder were tested.

Through trial and error, it was determined that the best lubrication for the o-rings was obtained using a graphite-oil bath followed by spray-on acrylic lubricant. This combination of lubricants achieved the lowest frictional torque during this study with an overall average o-ring torque measurement of 958 inch-pounds. The name brand of graphite-oil and acrylic lubricant used also makes a difference in o-ring friction values.

Results showed that o-ring friction was about 75% of the total measured torque value from a typical test. Previous testing with Holliston 00 sand showed higher test results; consequently, the o-ring torque was reported on the order of 50% of the total measured torque value in those tests (Sandoval et al., 2010).

Sand Abrasion. The o-ring friction may also be elevated by sand entering the o-ring grooves during testing. A grinding noise is evident when this happens; however, the effect on friction measurements was inconsistent. The sample chamber has slightly abraded zones where the o-rings contact the sample chamber walls. Because of these

22 wear zones, o-ring friction likely varies depending on the vertical position of the o-ring

within the chamber. Therefore, o-ring torque measurements should be taken at the same

approximate height as the tested specimen. The chamber should be buffed often with

increasingly finer grades of steel wool to minimize the frictional differences.

Ring Type. Minor inherent manufacturing defects of the o-rings may account for

some of the seemingly erratic behavior of the o-rings. The o-rings used are standard off-

the-shelf supply Vi-inch (8.25-inch inner ring and 11.75-inch outer ring) ethylene

propylene diene monomer rubber (EPDM) o-rings. EPDM o-rings were selected over

nitrile butadiene rubber (buna-N) rings to be consistent with previous testing. Quad-rings

were tested, but they most often became caught by the inner sample chamber wall while

lowering the top ring and did not appear to greatly decrease friction values.

Drive Motor and Controller

A Parker SM Series Brushless Servo Motor is mounted to the mid plate and provides rotary movement to the top ring. It is controlled by a Gemini GV6 Servo Drive/

Controller (M Series, 105) through the Motion Planner™ software. LabVIEW® is used to control the data collection system. Motion Planner is used to program and control the motion profiles used in testing.

Cyclic motion is triggered by the "shake spin" button in Motion Planner

(PROG23). The PROG23 script causes cyclic motion of the top ring until a predefined pore pressure is reached. After which, the "JUMP" command in PROG23 causes a

23 monotonic motion script to run. Both "PROG21" (four shearing speeds) and "PROG24"

(two shearing speeds) were used for monotonic motion.

The PROG23 and PROG24 scripts are included as Figures 2.4 and 2.5 as a

reference to the important coding information that follows. As seen in the scripts, the

amount of shearing is controlled by the rotational velocity and the distance traveled of the

top ring. Rotational velocity is controlled by the "V" command. Velocity is coded in

revolutions per second and the motor has a 40:1 gear head. When V is set as a variable

"VARI", the variable input has a resolution of 0.0001, therefore when VARI6 is set to

33333 in PROG24 (Figure 2.5), the substitution of the variable yields a velocity of

3.3333. This velocity equates to 5 revolutions per minute (rpm). Since the average

shearing radius is 12.92 cm, a V = 3.3333 makes a shear velocity of 6.8 cm/sec (3.3333

rev per sec / 40:1 gear head x 2TI x 12.92 cm = 6.76 cm/sec).

The amount the rotation is dictated by the number of motor counts set by the

distance "D" command. One rotation, 360 degrees, is equal to 32 motor counts.

Therefore, when D is set as a variable equal to 1600000 (VARI5 in Figure 2.5), the number of motor counts is equal to 160, creating 5 revolutions.

The Motion Planner command "TTRQA" (Transfer Actual Torque/Force) in

PROG24 is used to record the torque applied by the drive motor. Further discussion of this command is provided in the following section. A review of drive system manual

(Parker Automation, 2001) should help clarify how to adjust the motion profiles for particular testing needs.

24 ;PROGRAM 23 OSCILLATES THE TOP RING A SET NUMBER OF TIMES AND FINISHES UNLESS INPUT #1 IS ACTIVATED. IF INPUT #1 IS ACTIVATED, IT JUMPS TO MONOTONIC ROTATION PROGRAM PROG24 **********************************************

DEL PROG23 ;Delete and redefine PROG23 DEF PROG23

WAIT(IN.2=B1) ;Pore pressure cutoff value set in LabVIEW VARI1=9000 ;Set angle of oscillation - "0.6 degrees

LHO ;Set motor controls MCO MAO DRIVE1 PSETO COMEXCO

DMONAV24 ;Set monitors DMONBV5

A100 ;Set acceleration and velocity settings AA50 AD100 ADA50 V25

L50 ;Loop 50 times VARI2=VARI1*-1 ;Make negative oscillation variable D(VARI2) ;Move in negative direction first GO D(VARIl) ;Move in positive direction second GO IF(IN.1=B1) ;lf pore pressure is above set input value JUMP PROG24;Jump to monotonic monition PROG24 NIF LN

END

.********************************************************************** / Figure 2.4. Motion Planner cyclic motion code PROG23.

25 1 ;PROGRAM 24 ROTATES THE TOP RING FOR 5 REVOLUTIONS AT 5 RPM, HAS A 3 SECOND PAUSE , THEN ROTATES AT 20 RPM FOR 15 REVOLUTIONS. ************************

DEL PROG24 ;Delete and redefine PROG24 DEF PROG24 DRIVE1 PSETO VARI5= =1600000 ;Set initial distance to 5 revolutions MCO MAI VARI6==3333 3 ;Set initial to velocity 5 rpm VARI7:= 0 ;Move counter VARI8== 0 ;Ready to increment LO IF(VARI8=VARI7) ;lf start of test V(VARI6) ;Set velocity D(VARI5) ;Set distance VARI8=1 COMEXC1 GOl ;Start motion NIF TTRQA ;Transfer actual torque TO. 5 ;Wait a half second IF(PE>= VARI5);lf total distance traveled = 5 revolutions TTRQA ;Transfer actual torque T3 ;Wait 3 seconds VARI6= VARI6+99999 ;lncrease velocity by 15 rpm VARI5= VARI5+4800000 ;lncrease distance by 15 revolutions VARI 8=0 NIF IF(PE>=6400000) ;lf total distance traveled >= 20 revolutions JUMP PROG10;Jump to END script NIF ;should change PE to >=800000 so total distance is 25 revolutions LN END

DEL PROG10 DEF PROG10 COMEXCO END / J Figure 2.5. Motion Planner monotonic motion code PROG24.

26 Data Collection System

As mentioned earlier, the load cells in the lateral shafts measure the normal load

on the specimen, and a load cell at the base of the sample chamber measures the torque.

The load cell at the base of the sample chamber also measures thrust, however, the

measured thrust values are erratic at times. On occasion, the load cell would record

torque as evaluated thrust. Recorded thrust values were usually not equal to the loads

measured by the lateral shafts load cells; therefore, thrust measurements were felt to be

unreliable and were not analyzed during this program.

A removable linear variable differential transformer (LVDT) is installed prior to testing to monitor vertical displacement of the top ring during testing. A differential pressure transducer is located in line with the backpressure and sample chamber to monitor specimen pore pressure. Differential pressure transducers are also used to monitor the bag pressure of the pneumatic bladder between the bottom plate and machine table. All of the above-mentioned systems are monitored continuously during the test using a LabVIEW software application designed by Rob Cinq-Mars, analytical instrumentation engineer from UNH's University Instrumentation Center. LabVIEW produces a data file that includes: a time stamp, torque load cell readings, thrust measurements, data from both lateral shaft load cells, pneumatic bladder bag pressure, specimen pore pressure, LVDT displacement, and ring rotational position.

Rainer

A new, form-fitted rainer, constructed out of 1/32-inch PVC sheet and window screen (equivalent to a #18 mesh) was made to pluviate sand into the sample chamber. A

27 leveling gage is used to smooth the surface of the sand in the rainer before creating the specimen. Figure 2.6 is a picture of the rainer and leveling gage. The blade on the leveling gage has its corners trimmed at 45 degree angles to make a windrow of sand along the inner and outer rainer walls. The sand in the windrows compensates for the presence of the rainer in the sample chamber, once it is removed, and create a more level specimen surface. Markings on the gage are used to incrementally adjust the blade height, ensuring a level surface.

Figure 2.6. Rainer and leveling gage.

Other Key Components

The main frame of the RSD has a winch attached to the top plate that is used to lower and raise the mid plate (and top ring and drive motor). Four main shafts align the

28 top, mid, and bottom plates and guide the mid plate while it is raised and lowered. Two

lateral shafts with load cells firmly bolt the mid plate to the main frame during testing to

avoid unacceptable vertical deformations of the testing machine.

Other key components include:

• an external backpressure fluid chamber to control specimen pore pressure, • a compressed air system for backpressure and bag pressure, • a vacuum source to purge air from the specimen prior to saturation, • a CO2 tank that aids in the purging process, • de-aired water supplied by a water tank designed to de-air tap water, • heavy-duty jacks used to tilt the table during the saturation process, and • small green jacks used to support the bottom plate during testing.

Machine Calibration

All transducers were checked using an air pressure gauge and a digital multimeter before the testing program began. For calibration, the bag pressure sensor was adjusted slightly using the resistance screw (10 psi = 1 volt). The backpressure sensor was adjusted with the gain dial to be in the range of typical specimen initial pore pressure (15 to 20 psi). The functionally of the load cells was checked by securing the top plate and applying bag pressure, knowing that 1000 kg = 1 volt. Initial readings and calibration results are included in Appendix A.

The torque load cell between the sample chamber and bottom plate was initially checked using a custom-made wood lever that fits on the sample chamber, and a spring scale. This load cell was calibrated on three occasions using the wood lever and a steel cable connected to a stack of hanging weights. A single pulley mounted to the RSD table

29 was used to create a 90-degree directional change in the steel cable. A photo of the

calibration setup is included in Appendix A. Torque load cell calibration was done at the

beginning and end of the initial testing phase and prior to follow-up ring shear testing

after the centrifuge testing. Calibration prior to follow-up testing was after a RS testing

program conducted by UNH civil engineering senior, Nic Emerson, on an embedded load

cell in the top ring for potential direct force measurement of shear forces. His results are

included on the accompanying compact disk.

Machine Maintenance and Modifications

General maintenance of the RSD includes occasional buffing of the slightly

abraded zones on the sample chamber walls using increasingly finer grades of steel wool.

Also, the wire connections should be cleaned periodically with compressed air electronic

cleaner to avoid stray load cell measurements.

During this study, a new spider bushing was installed in the coupling between the

drive motor and center shaft housing. The connection of the coupling to the motor shaft began to slip after conducting special testing (discussed in Section 4). The coupling was removed, cleaned, and replaced with loctite® applied to the screw threads. Slippage was apparently observed during previous testing programs as well (personal communication with Pedro de Alba). The connection should be retooled to include a keyed joint that prevents slippage.

Two new linear rotary bearings were installed in the center shaft housing after ball bearings were noted falling out. During replacement, it was observed that the top bearing had migrated up to the top of the housing (away from the sample chamber) and

30 became jammed in place. Since the center shaft does not completely penetrate the housing, the movement of the upper bearing reduced the length of contact between the two. To prevent excess vertical migration of the bearings, the less worn of the two old bearings was reinstalled in the center shaft housing above the new ones as the top bearing. This stops the bearings from migrating excessively and allows the majority of the shaft/housing connection to be supported, maximizing stability of the upper ring.

As a preventive measure, the end of the center shaft was smoothed and slightly rounded with a file to lessen the initial contact between the shaft and lowest set of bearings when lowering the top ring. No loose bearings were observed after replacement and modifications.

Several other modification made to the testing machine during this testing program are discussed below.

Sample Chamber, Top Ring, and O-rings

Originally, the bottom of the sample chamber had Holliston 00 sand adhered to it by a two-part epoxy, but this coating slowly flaked off with use. The intention of the adhered sand was to prevent the specimen from slipping in the sample chamber. An attempt was made to epoxy F-75 sand to the chamber bottom; however, the process was aborted before the epoxy had cured because of the difficulty in obtaining an even, level surface. However, it has been shown that shearing is limited to a narrow band where the specimen contacts the top ring (Sandoval et al., 2010). It appears that the specimen does not slide at the base and excess roughness is not required.

31 On the top ring, self-adhesive sandpaper was initially used, but the glue failed due to the wet conditions encountered during testing. The sandpaper often peeled around the porous stone after one use. It was found through trial and error that water-resistant contact cement provided adequate adhesion, giving the sandpaper a much longer life. It is very important to have proper alignment when fixing the sandpaper, because it is difficult to scrape the cement off the smooth top ring.

The o-rings used to maintain pressure with the sample chamber are standard size rings (VWnch, 8.25-inch inner ring and 11.75-inch outer ring). They are slightly too small for the device when new and become stretched with usage. In an attempt to solve this problem, custom o-rings were made by cutting EPDM cord to length with a razor blade and gluing the ends together with superglue. The superglue bond could hold about

1 kilogram (Kg) in tension before breaking. Roughening the cord ends slightly with sandpaper increased the bond strength to approximately 4 Kg.

The homemade o-rings became stretched out too quickly with testing, although pore pressure in the sample chamber was maintained on one test even though the bond had all but failed, with the ends attached by only the very smallest amount.

Rainer Construction

The original rainer used for specimen preparation, constructed of cardboard and hexagonal screen, was is a state of disrepair and needed to be replaced. To make the new rainer (shown in Figure 2.6), two strips of PVC sheeting were cut to length. Each strip was formed into a ring by gluing the ends together with all-purpose PVC cement and a 1- inch overlap. These rings became the inner and outer walls of the rainer. Because of the

32 flexing of the PVC sheet, the bonded area needed to be supported until fully cured. PVC

welding to bond the strip ends proved to be a failure due to too much heat induced

deformation of the thin material, and the weld did not stick.

A roughly 13-inch diameter section of screen was glued with "E6000 glue" (with

a !/2-inch overlap) to the outer wall as the ring was held in shape by a 5-gallon bucket.

Six large paperclips were bent to match the angles of the sample chamber bottom and

connect the inner and outer rainer walls. This way, the rainer slopes with the chamber bottom. The inner portion of the screen was then cut and glued to the inner wall.

This rainer could be improved. The inner wall is snug with the sample chamber and would be improved by a slightly larger radius. New PVC strips have been cut for inner and outer walls but they have not been formed into rings. Also, an appropriate gauged wire could be used to connect the inner and outer walls, instead of paper clips. A coarser rainer screen may be needed depending on the material being tested.

Other Modifications

It was noted early on during this testing program that the reported time on the data from LabVIEW was in error. The time stamp was looping back and repeating numbers.

This was eliminated by lowering the data acquisition rate in the LabVIEW data acquisition (DAQ) device to a read of 400 at a rate of 2,000 hertz; however, the time stamp is still not correct. Based on real time measurements with a stop watch, and the number of Motion Planner TTRQA data points (collected every lA second), the Lab View time stamp is consistently twice as fast as reality (i.e., a test run of one minute has a time

33 stamp of two minutes on the last data point). This author was not familiar enough with

LabVIEW to find and correct the problem.

It was noted at one point during the testing program that metal shavings were

collecting on the top and mid plates. The connection of the winch to the top plate had

become misaligned, allowing the cable to rub against the top plate. A spring, located

where the winch is connected to the mounting plate, was too short and did not keep the

winch aligned. Additional washers were added to the spring to better set the alignment.

An automatic shut-off switch on the winch is present to stop the mid plate from being

lifted too high, but this author is unsure if the shutoff switch was ever set properly.

Caution is needed when lifting the mid plate to avoid pinching the drive motor power

cables against the top plate.

Finally, the compressibility of the dummy blocks used to determine the specimen height was investigated. On two occasions, the dummy blocks were placed in the machine and incrementally loaded and unloaded. Dial gauge readings were taken at bag pressures ranging from 0.75 psi to 23.7 psi. The best linear relation between bag pressure and dummy height was found to be the un-deformed average dummy height (0.599 inches) minus 0.001 times the bag pressure. This relation was used to determine the dummy height at the pressure the dummy readings were taken (for specimen height, volume, and relative density). Graphs of dummy height readings and associated bag pressures is included in Appendix A.

34 3. MATERIALS TESTED

Sand

F-75 foundry sand from the U.S. Silica Company's Ottawa, Illinois plant was used in all testing (ring shear, centrifuge, and modified triaxial). The F-75 is a natural, white, quartz silica sand that is fine to very fine-grained, poorly graded, subrounded to rounded. This sand is commonly used in geotechnical laboratory testing for research purposes. Based on visual inspection, the sand is over 95% quartz, with less than 5% unidentified black flakes. The U.S. Silica product specification indicated the sand is

99.8% silicon dioxide.

Index Properties

Key index properties of F-75 sand were determined by performing in-house laboratory testing and from literature review. Table 3.1 shows the results by source.

Discussions on test methods and sample preparation techniques used by the referenced authors are presented at the end of Section 3.

Specific Gravity. Following the procedures described in ASTM D 854 "Standard

Test Methods for Specific Gravity of Soil Solids by Water Pycnometer," Goulding

(2006) determined that the specific gravity of F-75 sand was equal to 2.65. This agrees with the values stated by the manufacturer.

35 Table 3.1. Summary of F-75 sand index properties.

CO

o o o 13

Max void ratio - 0.805 0.77 0.856 0.840** - 0.78 -- Min void ratio - 0.486 0.46 0.681 0.536** ~ 0.52 -

Uniformity coefficient - - 1.7 1.73 2.56 -- ~ ~

Coefficient of gradation - ~ -- 0.83 0.69 ~ -- -- Notes: * estimated from grain size distribution curve, ** average value.

Grain Size Distribution. Figure 3.1 presents a grain size distribution curve of the sand obtained by mechanical sieve analysis at UNH; which is in good agreement with the manufacturer-specified grain size distribution and published data. UNH sieve analysis results are presented in Appendix B.

The apparent deviation of the UNH grain size distribution from the other gradation curves in the 0 to 30 percent finer range is because the very fine #140 sieve

(0.106 mm) was mistakenly not used in the analysis. The uniformity coefficient and coefficient of gradation calculated therefore moderately agree with the values found in the literature.

According to the Unified Soil Classification System, F-75 sand is a poorly graded fine sand (SP). The mean particle size (D5o) was found to be 0.2 mm. U.S. Silica Co. specifications state a D5o of 0.18 mm for the F-75, but the value has been reported as low

36 as 0.18 mm (Chen, 2006), as 0.22 mm (Alshibli et al, 2003, William, 2004, and Zornberg et al. 2005), and as high as 0.23 mm (Goulding, 2006).

100 r-4 >— \ ^1 90 w \s • UXH Sieve Amh

20

10 * % \ .> •.. 0 * 1.00 0.10 0.01 Sieve Size (mm)

Figure 3.1. Grain-size distribution curves of F-75 sand.

The literature review reveals that the Ottawa F-75 sand is an ideal sand for geotechnical testing because of its uniform mean particle size and gradation. Because of the regularity of this sand, uniform specimens can be constructed and consistently replicated with strictly controlled specimen preparation conditions. F-75 sand was used for highly successful, very low-confining pressure triaxial tests conducted on board the

NASA space shuttle orbiters Atlantis and Endeavour (Alshibli, 2002, Alshibli et al.,

2003, and Sture et al., 1999).

37 As a comparison to the Holliston 00 sand, another commonly-used sand in laboratory testing, test results from F-75 sand are more likely to be repeatable and comparable across the industry. The Holliston 00 sand is very non-uniform and results from similarly-designed test can vary widely (R. Collins, pers. comm.).

Void Ratio. Goulding (2006) showed that the maximum void ratio (emax) for uniform spheres comes from a simple cubic packing of grains and result in an emax of

0.91. A minimum void ratio (emin) of 0.34 is achieved by close tetrahedral packing.

Although the sand grains of F-75 sand are subrounded to rounded, they are not uniform spheres.

The minimum and maximum void ratios for F-75 sand reported in the literature vary greatly (see Table 3.1). For instance, the reported emax varies from 0.77 to 0.856, and emin varies from 0.46 to 0.681. Only one author indicated the method used to determine these key parameters.

The max and min void ratios used for this study (emax = 0.805 and emin = 0.486) are from Dr. Khalid Alshibli at Louisiana State University. Alshibli and Hasan (2008) stated these values were determined using ASTM D 4253 "Standard Test Methods for

Maximum Index Density and Unit Weight of Soils Using a Vibratory Table" and D 4254

"Standard Test Methods for Minimum Index Density and Unit Weight of Soils and

Calculation of Relative Density." These values were also used for interpreting centrifuge testing at CU-Boulder.

The emax and emin of F-75 sand was ascertained at UNH using the Test Method for

Maximum and Minimum Densities of Sands (JSF T 161-1990), also known as the

38 Japanese Standard Method. This method is valid for sand sizes between 2 and 0.075 mm

in diameter (medium to fine sand). The maximum void ratio (loosest condition) is

determined by slowly raising a specifically-sized paper cone filled with the sand,

allowing the sand to fall with minimum drop height into a calibrated cylindrical mold within 20 to 30 seconds. Minimum void ratio (densest condition) is determined by filling the mold in ten lifts, lightly tapping each lift 100 times with a rubber mallet while turning the mold approximately lA turn every 5 taps. The tests were conducted twice and indicated an emax between 0.826 and 0.854 and an emin between 0.530 and 0.533

(laboratory testing data is found in Appendix B). These values are comparable to those given in the literature; however, the ASTM test methods used by Dr. Alshibli are more precise than the methods used by this author, therefore his values were used for the testing program.

Relative Density. Relative density (Dr) is used to indicate the denseness or looseness of granular soil and is a function of the in situ void ratio (e) of the soil and maximum and minimum void ratios for the soil.

^max "" ^min

Qualitative descriptions of the Dr of F-75 specimens in the literature indicate that the sand is "loose" when Dr is less than 45%) and "dense" when Dr is over 75%) (Zornberg et al. 2005, Dijkstra, 2004, and Chen, 2006). This is in agreement with descriptions given by Das (2008) where a "medium" soil has a Dr between 50%o and 70%o. The specimens tested during this program ranged from a loose 26.0%o Dr to a medium dense

39 64.7% Dr. Major factors that typically affect laboratory specimen density are funnel

aperture size and the drop height used during specimen preparation. Discussion on

specimen preparation techniques described in the literature is provided later in Section 3.

Permeability. In order to assess the effects of strain rate application during the

centrifuge testing (discussed in Section 5) the permeability of F-75 was determined by

constant and falling head tests. One loose specimen (30.5%o Dr) was prepared in a

standard proctor-sized apparatus using a rainer system similar to that used for the ring

shear and centrifuge specimens. A dense specimen (84.4% Dr) was created by tapping

the mold on the table while raining the sand. Test results show a permeability of F-75

sand in the order of 0.94 to 1.4 xlO"2 cm/sec, depending on the density of the specimen.

These values are within the range of soil permeability given by Dawson et al. (1998) that

are capable of retarding pore pressure during the collapse process (see Section 1). Test

data are provided in Appendix B.

Shear Strength Properties

Peak Friction Angle. The angle of internal friction () of granular quartz sand is

typically 33°, but <(> is a function of the state of packing of the sand and can vary greatly

with relative density. Review of the literature indicates that the c|> of F-75 sand increases

with Dr, as would be expected. In general, the published results were determined by triaxial tests (Alshibli et.al, 2003; Chen, 2006; and William, 2004), direct shear tests

(Goulding, 2006), or plane strain tests (Alshibli et.al, 2003 and Zornberg et. al., 2005) on

40 specimens prepared by air-pluviation methods under strictly-controlled conditions, with pre-defined discharge rates, aperture size, and drop height.

Published values of peak friction angle for F-75 sand range from 29° for loose specimens (20% Dr) (Chen, 2006) to near 50° for very dense specimens (97%o Dr)

(Alshibli et al., 2003). Figure 3.2 shows the published peak cj) values. Generally, peak friction angles between 40 and 50 degrees were reported for relative densities between 40 and 90 percent. With the exception of Zornberg et al. (2005), void ratios of the tested specimens were provided by the authors. In cases where the max and min void ratios used by others differed from those used in this study, the reported relative densities were adjusted using the specimen void ratio and max and min void ratios used for this study.

5£ ~ BAhsmbi et al (2003) A William (2004) % == 0 2273x + 29 403/ 50 - R2 = 0 7571^"^ - •Zornberg et al (2005) • UDcwoolkar ct al (2U10) (F-75 Sand) 48 - ODewoolkar ct al (2010) (Rough Aluminum /r-75) ^•" ^^ AWiIham (2004) Critical State *A^ & 46 ' A c 44 - < c - o A '& 42 - "w

ra 40 - (Critical state \ allies eu • a. ^ not included in A A trend line ) 38 - A - 36 - 8 A

...... , ...... L , f f ! 34 - , - - -t- ' ' ' ' ' ' 1 ' ' J—J—f i I L_J—| 1 I 1 t _i_J 1—1 L_J 40 50 60 70 80 90 100 Relative Density, Dr {%)

Figure 3.2. Summary of F-75 sand friction angle data.

41 Direct shear tests conducted at the University of Vermont (UVM) determined a peak friction angle of 36.1° for a relative density of 45% (Dewoolkar et.al 2010). This value seems low when compared to the general trend of the published results (see Figure

3.2), but it is within the range of scatter noted by William (2004).

Microgravity tests described in detail by Sture et al. (1999) showed that low

confining pressures result in high peak friction and dilation angles in cohesionless soil, and that both angles decrease as confining pressure increases. The experiments on the

space shuttle showed strength properties two to three times greater and stiffness properties ten times greater than expected, based on conventional thought (Alshibli,

2002). The space specimens bulged uniformly in the region of peak stress and deformation was accommodated by complex multiple symmetrical radial shear bands. In contrast, failure of typical plane strain specimens is characterized by distinct shear bands accompanied by stress response softening that depends on the specimen density and confining pressure (Alshibli et al., 2003).

Critical State Friction Angle. A limited amount of information about the critical state friction angle of F-75 sand was found in the literature. Grains of a dense soil interlock when sheared and continued shearing causes dilation of the soil. Peak soil strength (discussed in regards to peak friction angle in the previous section) occurs when the volume of the soil has expanded to a point where the interlocked grains can slip past each other. A lower shearing force (and therefore lower friction angle) is required for shearing to continue at the new soil volume, which may stay constant with continued shearing. Soil strength at constant volume shearing is called the critical state of shearing

42 resistance and the associated friction angle is the critical state friction angle (Wood,

2004). Loose soils tend to contract in volume until a constant volume is reached and may not develop a peak strength above the critical state. Residual strength is likely related to the critical state friction angle.

Critical state values in the density range between 72% and 75%) are given by

William (2004) and are plotted on Figure 3.2. The values range widely, but William's results suggest the critical state friction angle is on average 9° lower than the peak friction angle.

Alshibli et al. (2003) reported the same residual stress, regardless of confining pressure. The specimens continued to dilate and never reached the critical state condition. More research and testing is required to fully understand the critical state friction angle of F-75 sand.

Apparent Cohesion. Although F-75 sand is a non-cohesive granular material, apparent cohesion of the sand has been studied by some when testing wet, unsaturated sand. The ring shear specimens were tested in saturated conditions. Chen (2006) indicated that the strength of saturated unreinforced F-75 specimens under undrained loading can be represented by no cohesion intercept. However, Goulding (2006) presents information that indicates F-75 has an apparent cohesion at relatively high percent saturation.

Goulding did not test 100% saturated specimens, but did show that friction angle is relatively constant as a function of water content. An increase in effective stress due to capillary inter-particle forces increases the frictional resistance of the soil. The effect of

43 adding even a small amount of water had a very small change in saturation but a significant change in terms of shear strength (Goulding, 2006).

In a study of loading rate on pile capacity, Dijkstra (2004) noted that all of her tests on wet, unsaturated F-75 specimens were executed in soil with apparent cohesion conditions. Again, further research is needed to understand this behavior.

While determining the best method for ring shear specimen preparation (discussed in Section 4), it was found that adding just a small amount of water had a significant effect on apparent cohesion. The moisture content of the dry sand in storage (laboratory air dry) was found to be 0.03%. Small amounts of water were added to see if lower placement densities could be achieved with a more moist sand. When the moisture content was 0.5% or above, the sand grains clumped together and would not pass through the rainer screen. A moisture content of 0.25% (1 ml water to 400 g sand) was acceptable and appeared to slightly lower Dr, but sand preparation to this specification

(just a few drops of water per specimen) was deemed to not be worth the effort.

Therefore, dry sand was used for specimen preparation.

F-75 Specimen Preparation Methods

Since F-75 sand had not been tested on the UNH RSD before, a literature search was conducted to determine the best method for specimen preparation. In general, many of the published data are from specimens prepared using air pluviation methods; with pre-defined discharge rates, aperture size, and drop height. Repeatable densities can easily be achieved with F-75 sand using strictly-controlled conditions. The following

44 summarizes various methods of specimen preparation and testing methods found in the

literature review.

For comparison of strain localization in sand due to plane strain and triaxial

compression, Alshibli et al. (2003) ensured uniform specimen density by air pluviation of the sand, or 'raining' from a funnel into a mold. The specimens were prepared in a terrestrial laboratory for standard laboratory testing and for testing onboard the space

shuttle at microgravity. The raining intensity and velocity were controlled by funnel

opening size and distance between the funnel and the mold.

Chen (2006) investigated the drained and undrained behavior of fiber-reinforced

sand using consolidated triaxial compression tests. An under-compaction process was used to produce homogeneous reinforced and unreinforced specimens. The F-75 sand was mixed to a nominal 10% moisture content for loose specimens and 3% moisture content for medium-dense specimens and allowed to hydrate overnight prior to compaction. The specimens were backpressure-saturated at an effective consolidation

stress of 2.5 psi using the ASTM D 4767 dry mounting method. It took Chen approximately 5 days to bring the Skempton's pore pressure coefficient "B-value" of the specimens to 0.96.

Dijkstra (2004) was investigating the influence of loading rate on pile capacity in unsaturated sand. Specimens were prepared by flowing water from the bottom of a sand- filled tank to redistribute the grains into a very loose undisturbed soil structure. The wall of the tank was then immediately vibrated with water still in the tank, to obtain greater compaction of the sand. Dijkstra reported that it was difficult to obtain repeatable

45 densities using this method even when vibration time, soil type, and fluidization parameters were held constant.

During the study of tensile strength, shear strength, and effective stress in unsaturated sand, Goulding (2006) used a Wykeham-Farrance direct shear machine to measure vertical and horizontal deformations, shearing force, and rate of deformation.

Both dry and partially-saturated specimens were tested. The dry sand specimens were pluviated into the shearing device using a small funnel to obtain relatively loose conditions. A belt sander was used to vibrate the sides of the device to obtain relatively dense conditions. Partially-saturated specimens were created using pre-moistened sand

(16-hour cure time) and an under-compaction method. A sliding hammer that provided energy equivalent to standard Proctor compaction achieved homogeneous specimens and reproducible densities.

Sadrekarmim and Olson (2010) used both air pluviation and moist tamping methods to prepare specimens during their study of particle damage in drained and constant volume ring shear tests. They indicated that air pluviation achieved by gently raising a funnel to deposit the particles with a nearly zero drop height reduced segregation of the fine and coarse particles. This method produced the loosest uniform structure possible by air pluviation. However, moist tamping using an under-compaction method produced looser specimens. Ottawa 20/40 sand was used in their study, along with Illinois River and Mississippi River sands.

An air pluviation method was used by William (2004) in the development of a true triaxial apparatus. Dry sand was passed through a funnel into a stack of sieves which

46 rained the sand into the mold. This method yielded specimens with consistent void ratios within ±0.006.

Zornberg et al. (2005) used an air pluviation method for a series of centrifuge tests examining the failure mechanisms in pipelines bridging a void in dry sand backfill.

Sand was pluviated from specific elevations using pre-defined discharge rates into the testing apparatus under controlled conditions. Friction angles reported by Zornberg et al.

(2005) were determined by plane strain tests.

An investigation of specimen preparation method for the RSD conducted by

Sandoval (2007) with Holliston 00 sand included both dry and water pluviation methods.

It was found that the sand would segregate when using the water pluviation method, producing non-uniform specimens. Also, the hopper and screen tool used to rain the sand was difficult to attach to the RSD. Therefore a dry pluviation method was devised where a circular rainer was placed in the sample chamber, sand was scooped in, and the rainer was pulled slowly to keep the drop height as small as possible (Sandoval, 2007). A vacuum cleaner with an adjustable blade was supported on the lip of the sample chamber and moved around to level the surface of the specimen.

The method used by Sandoval (2007) proved difficult to work with the F-75 sand.

When the vacuum nozzle was rested on the sample chamber and used to level the sand surface, it would often dip into the specimen and create divots in the surface when the vacuum hose got entangled in the testing machine. The method also required 100%) recovery of the specimen to determine Dr. It was found that a 99.9%> F-75 specimen recovery is possible when care is taken. With practice, the Sandoval method may prove

47 reliable; but a new specimen construction method (detailed in Section 4) was devised for the F-75 sand.

Pore Fluid

Two types of pore fluids were used during this study. De-aired water was used for all of the ring shear tests and modified triaxial tests and most of the centrifuge test.

Methylcellulose was used as a substitute pore fluid in two centrifuge models to prolong the excess pore pressure dissipation phase. Two methylcellulose solutions were used;

0.7% by mass and 1.4% by mass. The viscosities of the solutions were 7 and 22 centistokes (cSt), respectively (Dewoolkar et al., 2010). Dewoolkar et al. (1999) and

Stewart et al. (1998) showed that methylcellulose is an acceptable substitute pore fluid for seismic centrifuge testing involving saturated sand models.

48 4. RING SHEAR TESTING

Machine Preparation

All equipment, including the drive motor, should be plugged in and turned on a minimum of 30 minutes before testing or collecting data. The computer used to control the drive motor and data acquisition system should be turned on before anything else.

The LabVIEW interface is accessed through the "2009 RingShear" shortcut on the

computer desktop. The Motion Planner interface is accessed through the "MP 2010"

shortcut.

The o-rings need to be cleaned by hand using an orange pumice solution and air- dried. They are then drawn through a small bowl of graphite oil. The oil can be applied after depositing sand in the sample chamber (discussed later in Section 4). Spray-on acrylic lubricant is applied immediately prior to o-ring installation, just before lowering the top ring for specimen saturation and testing. The o-rings are installed after specimen deposition to minimize evaporation of the lubricant.

Before placing sand into the sample chamber, the height of dummy blocks are measured so the height of the specimen can be determined. Three dummy blocks are placed in the empty chamber at designated locations approximately 120 degrees apart.

The top ring is lowered and bolted in place, and bag pressure is increased to the same bag pressure used during testing. The top of the shearing ring is measured to the nearest

0.001 inch using a dial gage at the three dummy block locations. Dial gage readings and bag pressure are recorded on a testing sheet (shown as Figure 4.1) and the relative density

49 UNH Ring Shear Device Testing Sheet

Test # Date

1 Measure friction of O rings at same height as test

2 Measure dummy heights

location 1 LC1 2 LC2 3 Bag

Time 3 Race sample in chamber a Insert rainer in chamber b Place sample of known weight sn rainer using a funnel

sample weight recovered weight (if necessary)

moisture content tare (if necessary) w+t d+t

c Level sample with level gauge d Lift rainer at a smooth and constant rate e Does it look right? if not, redo f Install O-nngs g Lower top ring and bolt in place h Apply bag pressure so load cells combined reads about 855 kg (about 17 19 psi) allow to stabilize

LC1 LC2 Bag

i Measure top of ring to 0 001' with dial gauge at the 3 locations

location 1 final at after 2 load test test _pst 3 time

Dr

Figure 4.1. Ring shear testing sheet.

50 Test 8

4 Prepare sample for test

Tfme a M=^=HM Apply 5 mm Hg vacuum to top drain for 10 minuies b Jack table end so top drain is the highest point c Attach flask to monitor bubbles install LVDT d Circulate CC^ at slow rate for 20 minutes e Stop C02 for 5 minuies f Recirculate CO, for an addsttonal 5 minutes ^^^[ g Seep in de aired water at a very slow rate for about 20 to 30 minutes until water appears h Stop water for 2 minutes, the circulate until no bubbles are observed i Circulate de aired water for 20 minutes j Stop water for 5 m mutes, open valve and check for bubbles k Recirculate water for an additional five minutes if bubbles are present I Shut top valve turn watenoackpressure switch m Apply 15 psi backpressure Open top valve to let out bubbles, then shut valve n Let sit for 15 minutes with back pressure on check for leaks o Open top valve to let bubbles out circulate water for 5 minutes if it bubbles

p Shut backpressure valve by tank and check if pore pressure is constant

Race of loss psi seconds

q Cneck B value apply an additional 9 5 psi on bag record new pore pressure

Bag initial Bag final Difference Pore initial Pore final Difference B Value

r Remove additional bag pressure s Open backpressure valve, lower table slowly disassemble flask bubbler t Put bottom plaie suppori jacks in place and hand tighten

5 Run test LC1_ Time a Start new LabVIEW file (new file = VIliWill (date)) LG2 _ b Close backpressure valvt Bag~ c Run Motion Planner program Pore initial" Pore final

7 Clean up

a Measure sample height b Remove green jacks c Turn water/backpressure switch to water side start to dram sample d Remove backpressure, open bacl^ressure valve e Remove bag pressure and open top valve f Note sample condition and remove for drying overnight and weighing if necessary g Recover any remaining sand from chamber the next morning dry in oven for at least 1 hour and add to recovered sample for weighing

Figure 4.1. Ring shear testing sheet (page 2).

51 calculation spreadsheet (included on the accompanying computer CD). Both sides of the double-sided testing sheet are shown on Figure 4.1, and give the key steps in running a test.

Constructing a Specimen

Sand Deposition

A new specimen deposition method was devised for this testing program, using an air-pluviation method similar to that previously used with the ring shear device. The rainer is placed into the sample chamber and filled with a known amount of sand

(typically 1.5 kg). The sand is pluviated into the rainer at a minimal drop height through a 10-mm aperture funnel, taking care to fill the rainer evenly. The leveling gage is used to push the high spots into nearby low areas, then level the surface by moving the gage around the sample chamber, adjusting the blade up or down until the surface is level, with very little excess sand. Excess sand is feathered out on the surface. The rainer is then pulled up steadily, maintaining a minimal drop height, allowing the grains to redistribute as they passed through the screen.

This method produced F-75 specimens with initial placement Dr between 29% and 44%. Sandoval et al. (2010) reported specimen relative densities between 19% and

36%o and stated that different densities were achieved by slightly varying the speed at which the rainer was lifted. This author found little, if any, relation between lifting speed and density, with the Dr at test time being more affected by the saturation process.

It should be noted that the Dr of the F-75 specimens at test time are 25 to 30%) higher than the range of Dr reported for Holliston 00 sand specimens previously tested on

52 the RSD (Sandoval et al., 2010, and Sandoval, 2007). This is more than what would be expected due to material differences. This noted difference prompted studies into the

sample chamber and top ring dimensions, dummy block compressibility, specimen recovery percentage, and change of Dr during specimen preparation. This author believes the major difference in reported Dr results from when density was determined during the

specimen preparation method. It appears that the Dr of the specimens tested by Sandoval was determined before sample saturation and consolidation and not immediately before testing as was the specimens tested during this study. Further discussion on testing with

Holliston 00 sand can be found in the "Special Tests" subsection at the end of Section 4.

Saturation Process

To saturate a specimen, lubricated inner and outer o-rings are installed on the top ring before it is lowered and secured. A vertical confining stress is induced via the pneumatic bladder to replicate stress conditions at a depth of 7.6 meters (25 feet), a typical depth observed to be prone to liquefaction. The vertical load is measured by the load cells in the lateral shafts. When the combined readings of the load cells is 855 kg

(1885 lb), the vertical confining stress on the specimen is 206.7 kPa (30 psi). The control of pneumatic bladder inflating is not very precise which makes it easy to over-shoot the desired vertical confining stress. Because of this, the average vertical confining stress used for F-75 testing was slightly higher, 227.4 kPa (33.0 psi).

A 5-inch Hg vacuum is applied to the top of the specimen for 10 minutes to purge the specimen of air. To assist in the saturation process, the RSD table is lifted on one side (tilted) using jacks to make the bottom valve on the sample chamber the low point

53 and the valve on the top ring the high point. After the vacuum, carbon dioxide (CO2) is

flushed through the specimen from the bottom for 20 minutes, followed by a 5-minute

rest period and another 5 minute CO2 purge. De-aired water is then slowly seeped in

from the bottom of the chamber to saturate the sand. Flow rates for both the CO2 and de-

aired water are visually monitored with the aid of a bubbler system.

Once water is observed exiting the valve on the top ring (after 20 to 40 minutes),

de-aired water is circulated through the specimen for a minimum of 20 minutes and until

no bubbles are observed exiting the top valve. The saturated specimen is allowed to rest

for 5 minutes, after which de-aired water is again circulated to check for entrapped air

exiting the top valve. Back pressure is then applied and the specimen is allowed to

consolidate for 15 minutes at 125 kPa (18 psi) effective stress (on average). The top valve is opened once again to check for entrapped air. Opening of the top valve should be minimized because the specimen consolidates a little more every time it is opened.

Skempton's pore pressure coefficient "B-value" is checked after specimen

consolidation by increasing the bag pressure 10 psi and monitoring the increase in

specimen pore pressure. The B-value check did not appear to work well with the F-75

sand. A reported average B-value of 0.92 was obtained during previous Holliston 00 testing (Sandoval et al., 2010). B-values measured while testing F-75 sand had an average of 1.00 and ranged from 0.77 to 1.30. Most of the B-value measurements taken are potentially erroneous, with pore pressure often increasing more than the vertical stress applied. All transducers were checked and in good working order. It should be noted that it took Chen (2006) approximately 5 days to bring the B-values of his backpressure- saturated F-75 specimens to 0.96. Nonetheless, this author believes the B-value of the

54 specimen is not necessary an important variable, since the object of the shear testing was

to determine post-liquefaction behavior, rather than the precise number of cycles needed

to initiate liquefaction.

The specimen volume, used to determine Dr, is measured after the specimen is

saturated, consolidated, and B-value assessed; just prior to cyclic motion.

Specimen Uniformity

The specimen uniformity was not investigated as part of this study. However, the

uniformity of ring shear specimens was investigated by Sandoval (2007). He conducted a

gelatin impregnation test on one Holliston 00 specimen in a dummy sample chamber.

Sandoval et al. (2010) reported a maximum difference of about 2.4%) in relative density,

with a standard deviation of 1.16%. Considering that the dummy sample chamber used

for the impregnation test was slightly larger than the true sample chamber, and the previous rainer was not fitted to the sample chamber slope, it is believed that uniformity

of the F-75 specimens would be equal to or better than that of the Holliston 00

impregnation test specimen.

Testing Procedure

Just prior to testing, green jacks are installed under the base plate to prevent the sample chamber from moving down. Specimen height measurements are then collected and recorded. The average specimen height of F-75 specimens was 2.2 cm (0.87 inch).

To run a test, the LabVIEW data collection system is started first, prompting a file name. File names used for this project included the date of file preceded by T# (Test and

55 sequential number) or OR (o-ring measurement). O-ring files also included an a, b, or c

if more than one friction measurement was taken on a given day. For example, file

ORb 021411 would be the second o-ring friction measurement collected on February

14,2011.

To start motion of the top ring, ensure that the backpressure valve is shut, then

start cyclic motion by hitting the "shake spin" button in Motion Planner (to trigger

PROG23). A cyclic load with ±5 degrees of rotation (equal to approximately 1.35%

strain at the average diameter of the specimen) is applied by the top ring and liquefies the

specimen. Once the specimen pore pressure increases under the cyclic load to above a set

level (default of 30 psi), a program of progressively increasing monotonic rotation is

automatically triggered. Liquefaction usually occurred in no more than two complete

cycles.

For this study, post-liquefaction monotonic testing was initially conducted at four

increasing top ring rotation velocities (referred to as shearing speeds in subsequent

discussions) of 6.8, 13.5, 20.3, and 27.0 cm/s (5, 10, 15, and 20 revolutions per minute)

for 10 revolutions at each speed. Follow-up ring shear tests after centrifuge testing were

conducted at only 6.8 and 27 cm/s. Sadrekarimi and Olson (2010) conducted ring shear tests at a rate of 18.6 cm/s so large shear displacements (more that 10 meters) could be completed in one day, and stated that numerous investigators have shown the shearing rate has negligible effect on shearing behavior.

The progressively increasing monotonic rotation program was usually repeated on the liquefied specimen after the first run had stopped, so that each specimen had two runs

56 of shearing. Once the second session of shearing is complete, LabVIEW data collection is stopped. Typical processed torque data from a test are shown on Figure 4.2.

20 rpm 130 Torque vs. Time {Test T12) 120

110 20 rpm

*g 100 t 15 rpm r 10 rpm %/VVr

5 rpm -vn KA^v 70

60 Run 1 «x~. Torque Load Cell Drive Motor

50 _j -j . -_ i --•-- p—j -,--1 | | —j— r r~~i i i r r~"~t r* "i i 60 120 180 240 300 360 420 480 Time (seconds)

Figure 4.2. Typical processed data from a ring shear test.

As seen in the figure, torque measurements collected by the drive motor (Motion

Planner) were, on average 22%, lower than those collected by the torque load cell.

However, torque due to the specimen shearing reaction is the difference between the test result and the o-ring friction; both the torque load cell and the drive motor returned similar values in this respect. The torque load cell monitored by LabVIEW was calibrated on three occasions during the testing program. The drive motor was not calibrated. Torque data plots for each successful test are included in Appendix C.

Electronic files of all tests are included on the accompanying CD.

57 Post Testing Procedures

Specimen height measurements should be taken after testing for comparison to pre-testing values. The green jacks are then removed from under the base plate and the

sample chamber is drained through the bottom valve. Bag pressure is released to

depressurize the pneumatic bladder. Pressure in the backpressure chamber is decreased by shutting the compressed air source valve. All pressures should be released slowly to avoid stressing the transducers.

Once all pressures have been released, the valve on the top ring is opened and the lateral shafts are unbolted from the mid plate. When raising the mid plate, it is important to do so slowly, because the top ring can be stuck in the sample chamber (due to suction when lifting the top ring). When this happens, the entire bottom plate is lifted, stretching the pneumatic bladder. If the top plate is lifted too high or fast, the bottom plate may slam back down once the top ring is out of the sample chamber, potentially damaging the torque load cell. Markings have been drawn on the main frame to indicate the maximum height the mid plate should be raised before the top ring is clear of the sample chamber.

Moist sand is then recovered from the sample chamber. Note that the surface of the sand should be smooth. If "waves" are present, the top ring may have lost contact with the specimen during testing, and the validity of the results may be in question.

Measure the thickness of the shear band if it is obvious during sand recovery. Remove and wash the o-rings. Ensure that all sand is removed from the o-ring grooves. Since the weight of the sand is determined prior to testing, it is not necessary to completely recover it. Any sand remaining in the sample chamber will dry overnight and can be vacuumed out.

58 The friction of the o-rings should be re-measured soon after testing. A typical testing sequence was developed: the friction of a new set of o-rings is measured, the first test is conducted, o-ring friction is re-measured, the second test is conducted, and then a final o-ring friction measurement is made. The second o-ring friction measurement, between the two tests, served as the post-friction value for the first test and the pre- friction value for the second test. This sequence can be conducted within an 8-hour period, once the operator becomes familiar with the testing machine and method.

Data Reduction

LabVIEW produces an Excel data worksheet that contains all measurements collected by the load cells and transducers. The worksheet includes measurements of: time, torque load cell response, thrust, lateral shaft loads, pneumatic bladder bag pressure, specimen pore pressure, LVDT displacements, and ring rotational position. The majority of data reduction is associated with the torque measurements. In the worksheet data file, the torque data recorded by LabView can be plotted versus time to easily distinguish the reading from the different shearing speeds. The motion profile includes a three-second pause between the testing speeds that causes a significant drop in torque values, which serves as a visual break between the shearing speeds. Figure 4.3 is a plot of typical raw torque load cell data recorded by LabVIEW.

The torque values from the drive motor are only recorded on the Motion Planner interface. They need to be selected and copied from the window display and pasted into the LabVIEW data file. After the drive motor torque values have been copied to the

Excel worksheet, they need to be converted to meaningful numbers. The Motion Planner

59 TTRQA command returns the percent of full-scale output of the drive motor and is reported as "*TTRQA12.43," for instance. The percent is extracted from each individual

data entry (using the MID function in Excel). The drive motor has a 24 Newton-meters

(N-m) full scale output automatically set according to the configuration utility. A 40:1 gearhead makes the true full-scale output equal to 960 N-m. Therefore, the drive motor torque values reported are a percentage of 960 N-m. The torque readings can be converted to equivalent in.-lb torque readings by multiplying the extracted values by

84.9672 (960 N-m x 8.85075 in.-lb per N-m /100 = 84.9672 in.-lb).

Torque vs. Time (Test T12) 20 rpm - Raw torque load edl data 1150

Run 1 Run 2 15 rpm Excess o-ring friction during monotonic motion 20 rpm Breaks between fit4 10 rpm shearing speeds A V :MV*' 15 rpm \/

5 rpm r*T

i 1 r r" 1— 1 1 1 i 1 1 r 0 30 60 90 120 150 180 210 240 270 300 330 360 390 420 450 480 Time (seconds) Figure 4.3. Raw torque load cell data plot.

60 Once the drive motor data have been converted, the breaks in the data between the different testing shearing speeds need to be found. Usually there are only two obvious data points per break in motion. The drive motor torques values do not always show a

significant change during the break between shearing speeds, because the torque is held by the stopped motor. For graphic display purposes, rows should be inserted between the drive motor shearing speeds data sets to align them better with the LabVIEW data. A fully reduced data set is shown graphically in Figure 4.2. Fully reduced data sets for successful tests are included in Appendix C.

Not all torque values recorded during monotonic motion are representative of the residual strength of the specimen. For instance, the o-ring friction may be excessively low at the beginning of the shearing or may be high near the end, or vice versa.

Therefore, good judgment is required to ensure proper data analysis.

The beginning or end of the data set from each shearing speed is usually trimmed where it deviates from the apparent overall trend in the data set. The overall trend in torque values generally increases over the time of shearing; or less often, it decreases with time, but it is rarely constant over the shearing session. On few occasions, erratic torque values within the middle of the dataset are removed if obvious excess o-ring friction is noted. The average value of the remaining data points is considered the measured value for that shearing speed data set.

The torque attributed to the shearing of the liquefied sand is the difference between the torque measured during the test and the o-ring friction. The difference in torque is divided by the average ring radius and the shear area to obtain the residual strength of the liquefied sand.

61 Based on the above-described testing scheme, there are two values of test torque

for each shearing speed (from Run 1 and Run 2) and four values of o-ring friction (two

runs before, and two runs after). Each test run, first or second, was compared only to the

associated run of the o-ring measurements. This gives four potential values of residual

strength for each shearing speed of each test. However, since torque measurements were

collected with both the torque load cell on the sample chamber (LabVIEW) and by the

drive motor (Motion Planner), there are potentially eight values of residual strength for

each shearing speed. Figure 4.4 shows a typical spread of the eight results (only two

shearing speeds were tested for this specimen). Similar plots for all the successful tests

are included in Appendix C.

1.70 A

t^* 1.50 - y%

*•a. •* 1.30 - "A" LabVIEW, o-ring before, Run 1 CuO • LabVIEW, o-ring before, Run 2 c CJ —•—La&VIEW, o-ring after, Run i 4-» 1.10 - u* =#—LabVIEW, QM-ing after, Run 2 CO —A— Motion Planner, o-ring before, Run 1 3 0.90 J t3 O Motion Planner, o-rfng before, Run 2

Some, none, or all of the torque measurements recorded could be in error due to loss of o-ring lubricant, sand grains migrating into the o-ring channel, or other reasons.

Of the eight potential values, unusually high or low (erratic) results are disregarded and the average of the remaining values is reported as the residual strength of the specimen and shearing speed.

62 Special Tests

Shear Zone Determination

One test incorporated colored, fine-grained craft sand to determine the shear band thickness of the specimen. Sandoval (2007) and Sandoval et al. (2010) had previously shown that shear failure in the liquefied sand did not extend through the full thickness of the specimen. For this test, a column of colored sand that extended from the bottom of the sample chamber to the top of the specimen surface was made in the rainer while the with the F-75 sand was being placed. Care was taken not to disturb the boundary between the sands when leveling the surface before pulling the rainer.

Figure 4.5 is a photograph of the shear band on top of the colored sand column as it was excavated after testing. During this test the shear band varied between 5 and 6 mm in thickness, which is comparable to the thickness reported by Sandoval for the Holliston

00 sand (6 mm).

With the F-75 and colored sand specimen, the shear band was 25 to 30 times the

D50 of the sand. On occasion, with other F-75 specimens, the shear band appeared obvious due to a change in void ratio and ranged from 2 to 5 mm, 10 to 25 times D50.

Alshibi and Hasan (2008) indicated an average increase in void ratio as high as 24.7% within the shear band. The shear zone noted by Alshibi and Hasan were on average only

11 to 14 times D50. This range was noted on one F-75 specimen. It should be noted that the shear band thicknesses for F-75 reported in the literature are from specimens tested by direct shear or triaxial test, where the failure occurred within the specimen. With the

UNH RSD, failure is concentrated at the top of the specimen, where it is in contact with

63 the top ring. The shear bands noted during this testing program may be thicker than what they should be due to the coarse-grit sandpaper used for roughness on the top ring.

Shear band thickness is used to compute shear strains and shear strain rates.

However, this author believes that due to the observed differences in shear band thickness, it would be premature to compute shear strains and shear strain rates based on only one special test. More special testing with colored sand should be conducted to determine an average thickness of the sheared zone. This way, a good estimate of the shear strain rate can be determined.

Finally, it should be noted that no obvious sand particle damage was observed within the shear band. Sadrekarimi and Olson (2010) reported significant particle damage in the shear band occurs in ring shear test, but stated that damage typically occurs when the normal stress is above about 28 kPa (4 psi) in constant volume tests. They also

64 concluded that particle damage in specimens with large initial void ratios was not

significant because the response was entirely contractive during shear, resulting in a

substantial decrease in effective stress, a typical response of very loose sands that experience flow liquefaction. This author would not expect any particle damage within the shear band because of the zero effective stress conditions of testing.

Sadrekarimi and Olson (2010) noted that considerable particle damage in flowslides can likely contribute to long runout because an increase in fines content decreases hydraulic conductivity. This may slow pore pressure dissipation and may account for significant decrease in apparent frictional resistance and high mobility of many long runout slides.

Lower Initial Vertical Stress

One test (T12) was conducted at a lower total vertical stress (129 kPa, 18.6 psi) so a less dense specimen could be obtained. For this test, the total stress applied by the pneumatic bladder was roughly one-half of what was normally applied, but backpressure was not lowered accordingly. This condition gave a Dr of 45.8% at test time; however, the effective stress was very low (25 kPa, 3.6 psi) in comparison to the normal specimens

(124 kPa, 18 psi on average).

The T12 test results are in line with results from specimens tested at higher vertical total and effective stresses (see Figure 6.1 in Section 6), suggesting that initial vertical effective stress has little impact on Sur values obtained by RS testing. Alshibli et al. (2003) noted that all of their specimens showed nearly the same residual stress regardless of the confining pressure value. The very low initial effective stress of T12 is

65 not a real-world condition normally found with flowslides caused by earthquakes, but

low initial effective stress conditions may be present in flowslides initiated by static

liquefaction. However, the low initial effective stress does affect the results from this test

greatly when Sur is normalized by initial effective stress, as some authors have done (see

Section 6 for discussion on Sur normalization).

Pore Pressure Dissipation Tests

A few attempts were made to monitor post-liquefaction soil strength recovery as excess pore pressure dissipates. This was accomplished by adjusting the pressure in the backpressure chamber to equal the pore pressure in the liquefied specimen. The valve between the sample and backpressure chambers was opened and the pore pressure lowered very slowly by turning the backpressure air supply valve. Figure 4.6 shows that as effective stress increased, a delay in soil strength recovery was noted. Then, a very rapid increase of soil strength would occur and seize the machine. This style of strength gain is very different from what was observed during centrifuge testing and is likely an aberration of the testing machine (see Figures 5.6 and 5.7 for comparison).

Dissipation could not be controlled finely enough with the RSD to see a complete strength gain curve. A better control of backpressure release (perhaps with a servo valve) may allow for more precise strength measurements as effective stress increases, although it must be remembered that the testing machine can seize when soil strength is over roughly 50 kPa because the machine is not designed to shear non-liquefied soil. The

66 evolution of shearing strength with pore pressure dissipation could not be thoroughly assessed by the RSD.

Effective Stress (kPa)

Figure 4.6. Material strength gain with pore pressure loss, RS specimens.

Testing with Holliston 00 Sand

Due to the great difference in relative specimen densities obtained with the F-75 sand as compared to those reported for Holliston 00 specimens (Sandoval, 2007 and

Sandoval et al., 2010), two tests were conducted using Holliston 00 sand (D5o = 0.3 mm, approximately 85 to 90%) quartz, 5%o feldspar, and 5 to 10%o mica). Trace amounts of coarse sand was retained in the rainer during specimen deposition.

The main purpose of these tests was to observe the change in Dr during the saturation process. Relative densities reported by Sandoval for Holliston 00 sand ranged

67 from 19%o to 36%o. A range of Dr for the F-75 sand was usually between 46%o and 63%o.

A step-by-step testing procedure given by Sandoval (2007) shows that the specimen

height readings were taken before the saturation process only. The Dr of the two special- test Holliston 00 specimens (see Section 2) were 46%o and 50%o after total stress load had been applied, but before the saturation process. These densities are higher than the range reported by Sandoval (19%o to 36%); but more importantly, the Dr of these specimens

increased an average of 7.8%o during the saturation process.

Furthermore, it should be noted that the average relative density for the Holliston

00 impregnation test specimen discussed earlier in Section 4 (Sandoval, 2007 and

Sandoval et al., 2010) was reported as 27.1%. This Dr is a typical value reported for the

Sandoval Holliston 00 test specimens. The impregnation test specimen was not subjected to confining pressure or consolidated as the test specimens were, and therefore the reported Dr for that specimen should have been lower than those reported for the test

specimens.

As mentioned earlier, this author believes the major difference in reported relative densities of the F-75 and Holliston 00 specimens is due to differences in the point at which the specimen height was measured during the specimen preparation process.

Relative density of the F-75 specimens increased on average 7.7%o during saturation, but the range was from 2%> to 18.2%o. The amount of change appeared to be influenced by the number of times the top valve was opened during saturation. Based on these findings, it is safe to say that Dr values reported by Sandoval are low, possibly by roughly 7.5%o on average; however, the actual Dr of his specimens at the test time are unknown.

68 5. COMPARISON TESTING

As mentioned before, ring shear testing was conducted concurrently with

centrifuge and modified triaxial testing. The centrifuge results are considered the "field" values, to which the ring shear and modified triaxial results are compared.

Centrifuge Tests

The centrifuge testing was conducted at CU-Boulder as part of a NSF-funded

Network for Earthquake Engineering Simulation Research (NEESR)-Payload project.

The following is a summary of detailed descriptions of centrifuge testing given in

Dewoolkar etal. (2010).

Centrifuge Model Setup

The centrifuge models were constructed in a rigid aluminum container 4 feet long,

1 foot wide, and 17 inches tall (121.9 cm by 30.5 cm by 43.5 cm). The models were brought to either 25 or 50 times the earth gravity and subjected to a shaking motion by a shake table mounted on the centrifuge swing platform. Pore pressure transducers (PPT) were installed to monitor excess pore pressure development and dissipation.

Accelerometers (AC) embedded in the specimens monitored soil response to the induced shaking. Linear variable differential transformers (LVDT) measured specimen consolidation during spin-up and dynamic testing. A typical model configuration is shown in Figure 5.1.

69

30 5 cm

7 62 cm

43 5 cm

1 9 cm (after Dewoolkar et al, 2010) AC12 Sectional View Figure 5.1. Typical centrifuge model configuration.

Two aluminum coupons (1 inch by 3 inches by 0.1 inch, 76.2 mm by 25.4 mm by

2.25 mm) were pulled through the specimens at various speeds before, during, and after

the simulated earthquake. One coupon was made "rough" by gluing on a thin layer of

F-75 sand. The peak contact friction angle between the sand and "rough" aluminum was

determined at UVM using a direct shear machine by gluing F-75 sand onto a smooth

aluminum insert to make it "rough." The test results show the contact friction angle

between F-75 sand and rough coupon (36.6°) is very similar to the internal friction angle

determined for the sand (36.1°). Therefore, the rough coupon data could be analyzed using the internal friction angle and critical state angle of the sand.

The coupons were pulled horizontally through the sand by a flexible, nylon- coated, 7x7 strand stainless steel wire rope with a coated diameter of 1.1684 mm. Two

70 identical motor assembly and pulley systems were mounted on lightweight aluminum billet machined frames on opposite ends of the container to offset each other's weight.

The systems were offset horizontally so as to allow unobstructed coupon travel through the sand. A specially-designed pulley box was used to create a 90-degree directional change for the pulling wire at the coupon level. Figure 5.2 is a photograph of a centrifuge model on the testing machine with key components identified.

Figure 5.2. Photograph of a centrifuge model installed on the swing platform.

71 Centrifuge Specimen Preparation

The preparation method used for the centrifuge models was very similar to that

used for the ring shear specimens and is described in detail in Dewoolkar et al. (2010). In

summary, specimen preparation began with a clean and dry model container. Each

model was constructed on a large floor scale so relative density could be determined

based on the amount of sand used.

A thin layer of clean fine gravel was placed in the bottom of the container and

covered with filter fabric to assist with specimen saturation. The specimen was built on

top of the fabric in approximately 2.5-inch (6.5 cm) thick lifts using a rainer constructed

of window screen (equivalent to a #18 mesh) and a thin plywood frame slightly smaller

than the model chamber. After leveling each lift, the rainer was lifted slowly to keep the

drop height as small as possible as the sand passed through the screen. Coupons and

accelerometers were installed on the top of each lift where necessary. (The pore pressure

transducers were installed later, during the saturation process.)

The sand-filled model container was installed on the centrifuge, covered with an

air-tight lid, and a 5-inch Hg vacuum was applied to the top of the model for about 20

minutes. A 40-minute C02 flush followed the vacuum, and de-aired water was then

seeped in from the bottom of the specimen to saturate the sand. The flow rate of CO2 and

de-aired water was visually monitored with the aid of a bubbler. De-aired water was

allowed to seep through the sand until it was approximately 1 cm above the specimen.

Near the end of the saturation process, the lid was removed so the motor

assemblies, cables, and LVDTs could be installed (as seen in Figure 5.2). Pore pressure transducers were put in place once the sand at a certain level was wetted. The height of

72 sand and water in the model chamber was measured when the LVDTs were installed and once testing was complete. As mentioned earlier, either water or methylcellulose was used as the pore fluid in the centrifuge specimens.

Centrifuge Testing Procedures

Testing procedures used in the centrifuge testing program are described in detail in Dewoolkar et al. (2010). In summary, after the specimen was brought up to the desired g-level (25g or 50g at the coupon level) and had consolidated (as noted by stable

LVDT readings), the dynamic testing program was initiated. The coupons were pulled at a desired speed for a given distance. Cyclic motion was applied via the shake table to liquefy the specimen once coupon movement had begun. Coupon pulling continued after the shaking ceased to monitor soil strength recovery as the excess pore pressures dissipated. A consistent testing procedure was followed for all tests with only coupon speed, distance travel, g-level, or specimen Dr as variables.

Centrifuge Test Results

Results from the centrifuge testing program included residual strength values and soil strength recovery behavior. Both smooth and rough coupons were used in the testing program, and both the drive motors and custom-built slip-ring transducers were used to collect data. The drive motor data were derived from the motor winding currents

(directly proportional to motor torque), whereas the slip-ring transducers monitored torsional strain at the coupling between the motor-gearbox output shaft and winch spool.

The different measuring techniques gave different pulling force results, with the slip-ring

73 transducers usually producing higher values and greater scatter overall. Testing conditions and Sur results of successful rough coupon centrifuge tests are included in

Table 5.1.

Table 5.1. Centrifuge testing conditions and rough coupon results.

C/5 Q S 3 I o > > E

CO -a CD B Initia l Vertica Initia l Por e Pressur Initia l Por e Pressur fro m Roug h Coupo n (kPa ) Roug h Coupo n Forc e (N ) Exces s Por e Pressur H Effectiv e Stres s (kPa ) Estimate d (kPa ) Measure d (kPa ) a, o Residua l Strengt h (kPa ) 8-18-1 44.2 116 124 113.6 30 50 WateOH r 14.0 3.6 100 (12.0) (3.1) 8-19-1 38.4 58 62 59.6 5 25 Water 15.0 3.9 55 (36.0) (9.3) 8-19-2 56.9 57 62 59.7 5 25 Water 29.0 7.5 53 (48.0) (12.4) 8-21-1 45.6 118 126 115.3 60 50 Water 15.0 3.9 108 (32.0) (8.3) 8-25-1 51.0 116 117 103.2 5 50 Methylcellulose 47.4 12.2 114.5 1.4% (83.0) (21.4) 8-25-2 64.7 115 117 112.8 2 50 Methylcellulose 104 1.4% (~) 8-26-1 48.3 119 126 117.6 5 50 Methylcellulose 33.5 8.7 107 0.7% (51.0) (13.2) Note: ~ no values, * wire broke. Drive motor results are in bold, slip-ring results are in parentheses ().

Liquefied Soil Behavior. Two values of initial pore pressure are given in Table

5.1; one estimated from the height of water above the sample and testing g-level, and one measured by the pore pressure transducers (PPTs). Pore pressures measured by the PPTs were lower, generally within 10% of values expected from model construction. This difference could have been due in part to possible evaporation of water during centrifuge spin-up; however, the post-testing water level above the specimen lowered at most by

74 1/4-inch (6.3 mm) from the initial measurement, which would account for only 2.5% difference at 50-g.

The estimated initial effective vertical stresses at the coupon level in Table 5.1 is based on the saturated unit weight of the sample (monitored by LVDT) and the initial height of water above the sand (estimated pore pressure). Initial measured values from the PPTs were used to monitor excess pore pressure.

Methylcellulose was used in tests 8-25-1, 8-25-2, and 8-26-1, to extend pore pressure dissipation. It appears that excess pore pressure may not have fully dissipated between test 8-25-1 and test 8-25-2 (tested immediately after on the same model). The observed specimen consolidation during the first test only accounts for 1.2 kPa (0.2 psi) of the 9.6 kPa (1.4 psi) difference measured by the PPTs before shaking began.

Figure 5.3 is typical of the centrifuge test data and shows filtered measurements of the pulling force on the rough coupon during Test 8-18-1, as measured by both the drive motors and slip-ring transducer. Data in excess of 20 Hz were digitally filtered out.

The change in pore pressure (excess pore pressure) measured during the test is shown on the figure to assist with data interpretation.

As seen in the figure, the wire broke during pulling at about 1.3 seconds. Once the wire broke, the motor assembly and the slip-ring transducer registered the resistance of just the wire against the sand. This resistance was considered the "zero" measurement and used to calibrate the forces after the wire broke to an approximately zero load.

The residual pulling force (12 to 14 N) is marked as dashed lines on Figure 5.3, and was typically the flat portion of the pulling force at the end of shaking, just before excess pore pressure began to dissipate. The lower pulling force at the beginning of the

75 Force Slip-ring 1 (Rcugh) (Tes 8/16-1) 300- Test 6/18-1 roo^ Rough coupon Slip-ring measuremef tt 300 §-* 50cs i 30 cm/s speed Wat-r m JZD = 44% m a w ®

o ri

-100- 05 1 5 25 ime (sec)

Fore© Drive Motor 1 (Rougi) (Test 8/18-1) 300^

l Test 8/18-1 700 Fough coupon Onva mo or msasuremeit A^ Wire broke m 500- 50 cs 30 cm/s speed (W § LWater 3 5001- D = 44% 0> Li ! Pulling force 0> 400 h LoL i C/J

Figure 5.3. Typical data plot of rough coupon pulling force.

76 shaking period, and immediately after the wire broke, are assumed to be a spring like reaction within the wire due to the quick release of tension. In water-saturated tests, the excess pore pressure began to dissipate as soon as the shaking stopped. In the test with methylcellulose, dissipation was delayed.

To determine residual strength of the specimen, the coupon drag force was divided by the total area of the two flat coupon surfaces (6 in.2, 3.871 xlO'3 m2). Forces on the leading edge of the thin plate coupon and the two side edges were considered insignificant (estimated to be less than 10 percent of the total force) and hence not considered when determining Sur.

Figure 5.4 shows the Sur results and trend lines plotted versus specimen relative density of the different data sets (smooth and rough coupons, drive motors and slip-ring transducers values). A slow sensors response time may have limited the sensitivity to low pulling force values, and likely accounts for much of the perceived scatter in the data. However, the coupon speed of the plotted results ranged from 2 to 60 cm/s and the g-level was either 25g or 50g (hence different initial effective stresses), both of which probably contributed to the perceived scatter in data.

A comparison of trend lines on Figure 5.4 show that the drive motor results from the rough coupon tests had the lowest proportion of variability within the data set as indicated by the coefficients of determination (R2) values. Therefore, the rough coupon drive motor results were considered representative of the residual strength of the material.

The slip-ring measurements were generally higher than the drive motor measurements for both rough and smooth coupon results (in the order of 45% higher for the rough coupon data). The smooth coupon results were very scattered and are not discussed further.

77 25 -r • Rough Coupon R2 = 0.2282 Dn vc Motor / O Rough Coupon — 20 Slip-Rmg • Smooth Coupon Dnvc Motor R2 - 0.4879 OSmooth Coupon tuD 15 4 Slip-Ring

15 10

'35 R2 = 0.2829

5 4 Smooth Coupon (slip-ring) R2 = 0.0131 0 .j i l I 1 | i i I L_ 30 40 50 60 70 Relative Density (%) Figure 5.4. Centrifuge tests results.

• Rough Coupon 0.25 T Dnvc Motor •Rough Coupon Slip-Ring 0.20 4 re ec f, 0.15 c 41 &- +-»

"5 0.10

0.05 Test 8-19-1 a: 38.4% Dr

0.00 -h _i i i_ 30 40 50 60 70 Relative Density (%) Figure 5.5. Centrifuge rough coupon data residual strength ratio plot.

78 Normalization of residual strength by pre-shaking vertical effective stress (a'v) has been used by some researchers when analyzing residual strength values of field case histories (Olson and Stark, 2002; Idriss and Boulanger, 2008; and Robertson, 2010).

Figure 5.5 shows the residual strength ratios (Sur/cj'v) of the rough coupon results plotted versus relative density. With the exception of one data point (38.4% Dr) and allowance made for experimental scatter, the data appears to be very congruent (even with the varying coupon speed and g-levels) and appears to benefit from SUr/G'v normalization.

The trend lines and R values in Figure 5.5 do not include the 38.4% Dr test results (Test 8-19-1). This test is somewhat unique: it has the lowest Dr, a lower pulling

(shearing) speed, and lower initial effective stress than most of the other tests. One would expect that the combination of the loose soil condition and slow shearing speed during Test 8-19-1 would result in a very low residual strength, which is difficult to measure accurately. Test 8-19-2 had the same shearing speed and initial effective stress as Test 8-19-1, but the Dr was 18.5% higher. Results from that test appear to "fit" better with the other normalized test data. Further in-depth analysis of the centrifuge data may be helpful in understanding the behavior of loose sand at low shearing speeds and low effective stress.

Post-Liquefaction Strength Recovery. The centrifuge testing program examined the evolution of soil shear strength as it decreased to a minimum and subsequently increased upon excess pore pressure dissipation. It was found that the recovery of soil strength increases more or less linearly as effective stress increases.

79 Figures 5.6 and 5.7 show the rough coupon force measurements, post shaking, as measured by the drive motor and the slip-ring transducer plotted against increasing vertical effective stress. Although the relative densities and coupon speeds were different for the tests, the figures show that the results generally fall within a narrow band that can be represented by a single line. The lines in both the drive motor and slip-ring plots have the same slope (approximately 7.14 N/kPa).

When analyzing the strength recovery data, Dewoolkar et al. (2010) found that the forces estimated by a simple force equilibrium model were significantly smaller than the measured forces. Therefore, the resistance offered by the leading edge of the coupon could be more significant during the return of soil strength than during liquefaction. To analyze the results, three solutions for "deep" anchors in dry sand were used: Ovesen

(1964), Biarez et al. (1965), and Merifield and Sloan (2006). The vertical stress from dry soil in the original equations was replaced by vertical effective stress.

O 20 40 60 80 ° 20 40 60 80 Vertical Effective Stress (kPa) Vertical Effective Stress (kPa)

Figure 5.6. Coupon force vs. effective Figure 5.7. Coupon force vs. effective stress (drive motor measurements). stress (slip-ring measurements).

80 Dewoolkar et al. (2010) noted that the solutions used were only very crude

models of the behavior of the coupon and may not be applicable because (1) the solutions

were developed for vertical anchors where the anchor reaction to pulling is due to the

area perpendicular to the pulling direction, which is generally greater than the anchor

thickness. The coupon on the other hand is thin, with the leading edge area (57 mm )

orders of magnitude smaller that the total flat surface area (7,742 mm2) parallel to the pulling direction. (2) The solutions predict zero coupon force at zero effective vertical

stress, whereas a small coupon force is present due to the residual soil strength. (3) The

friction angle assumed for the solutions significantly affects the result. Figures 5.6 and

5.7 suggests that the rate of regain of coupon force is not strongly dependent on the

friction angle (as related to Dr).

Dewoolkar et al. (2010) also noted that the vertical effective stresses assumed for

analysis were based on pore pressure measurements made at a significant distance from the coupon (up to 9 inches, 22.8 cm), whereas the actual effective stresses in the vicinity

of the coupon could be different. In a study of ploughs cutting saturated soil, Palmer

(1999) showed a dramatic increase in cutting force with speed. Because most sands dilate when they are sheared, increasing the void space between the particles, additional pore fluid has to flow in from the surrounding soil. When the deformation speed is rapid, a large pressure gradient in the pore fluid develops to drive the rapid flow into the void space. Because of the lower fluid pressure in the shear area, the effective stress between the soil particles increases, which increases the drag forces required for soil deformation.

Palmer (1999) indicated that during rapid deformation, the inward flow rate of the pore fluid is determined by the permeability and compressibility of the soil. Plow size and

81 cutting depth also have an influence, but the speed effect appears at lower shear rates if the sand is fine and comparatively impermeable. F-75 sand is fine and testing conducted

at UNH showed the permeability of F-75 sand in the order of 1.5 to 2.2 xlO"2 cm/sec, which is relatively low for clean sand.

The analysis of soil strength evolution during pore pressure dissipation based on pulled coupon force measurements is a complicated issue. Since the coupon is much less bulky than a pulled plow or anchor plate, the volume of soil directly affected by the motion of the coupon is likely quite small in comparison. The associated pore pressure

equilibrium may not be significantly affected by coupon speed. However, the

experimental and theoretical force measurements show that the coupon force increases with the vertical effective stress in a linear fashion, indicating that the recovery of a soil's

shear strength is very likely to have a linear relationship with excess pore pressure dissipation.

Shear Band Thickness. On one centrifuge test (8-17-1), colored sand was incorporated in the model at the coupon level to assess the thickness of the shear band.

While excavating the coupon after the test, a shear band, approximately 3-4 mm thick was noted (see Figure 5.8).

The colored shear band was about 1 to 2 mm thicker than the coupon (2.25 mm) and 15 to 20 times the D50 of the sand. The thickness of the shear band represents two shearing surfaces, considering that shearing occurred on both the top and bottom of the coupon. The shear band thickness for shear strains and shear strain rate computations is therefore 7.5 to 10 times D5Q. This is thinner than the shear zone noted by Alshibi and

82 Hasan (2008) (11 to 14 times D50). This may be due to the centrifuge specimen being

liquefied when sheared, where as the Alshibi and Hasan specimens were sheared until

liquefied. More colored-sand testing is needed in the centrifuge specimens to obtain a

range of shear band thickness. The centrifuge specimen shear band thickness testing and

findings discussed herein were not reported in Dewoolkar et al. (2010).

Figure 5.8. Coupon shear band.

Modified Triaxial Tests

Modified Triaxial System Description

Figure 5.9 is a schematic of the modified triaxial device. Tests were carried out by forming triaxial specimens of F-75 sand around a smooth 1 inch (2.54 cm) square titanium coupon, 0.06 inch (0.16 cm) thick, attached to a fine, 0.03 inch (0.08 cm) diameter) wire which could slide through an o-ring seal in the chamber base.

83 The specimens were approximately 9.5 inches (24 cm) tall and 2.8 inches (7.1 cm) in diameter. The coupon was placed in the specimen so that it could travel approximately 7 inches (17.8 cm) before the weight hanger hit the support. This would

stop the coupon about 1 inch (2.5 cm) above the base of the specimen. The test setup required that the coupon be located initially at ~ lA inch (1.25 cm) below the top cap, and then the specimen formed around it. The coupon's displacement and resistance to motion were measured by an LVDT and load cell arrangement as shown in the figure.

Cyclic Loader

(after de Alba and Ballestero) 1 \ ^*- Hanger Figure 5.9. Modified triaxial system schematic (not to scale).

Modified Triaxial Specimen Preparation

Specimens of F-75 sand were produced by two pluviation techniques. Loose specimens were formed as follows: a mold/membrane stretcher was assembled around the specimen base, and the coupon on its wire was positioned at 8 inches (20.3 cm) above

84 the base. The mold was filled with dry sand, and a 7.9 inch (20 cm) long sleeve fitted

with a cap at the free end was attached to the top of the mold. The mold and sleeve

assembly was then slowly rotated several times to re-deposit the sand. During rotation,

the coupon was held in position by a fine thread that extended through the cap. This

technique was found to create uniform, reproducible specimens in the 25% to 35%

relative density range, but it could not attain denser values.

For dense specimens (40% to 55% Dr), a conventional pluviation technique was

employed. A flask was filled with dry sand and sealed with a stopper having a 0.25 inch

(6.4 mm) diameter opening. The sample mold was then filled while maintaining the

inverted flask approximately 5.5 inches (14 cm) above the deposited sand. As before, the

coupon was held in place during the formation process using a fine thread. The effects of minor differences in sand grain structure between the two pluviation techniques were considered negligible once liquefaction occurred.

Specimens were saturated and tested using the conventional cyclic triaxial test procedure. Tests were carried out at an average initial effective stress of 118.4 kPa (17.2 psi). The modified triaxial system was set up to stop cyclic loading after initial liquefaction (pore pressure ratio, ru = 100%) when axial strains exceeded ± 2%.

Modified Triaxial Test Results

Typical results are shown in Figure 5.10, an example from a 47.8%) Dr specimen.

The figure indicates that after initial liquefaction, effective stress decreases to zero and the coupon starts to move (increasing velocity). The shear forces on the coupon then decrease to a minimum value before increasing with increasing velocity (shearing speed).

85 All tests at relative densities over 40%) showed a similar tendency, with coupon resistance

reaching a minimum at coupon velocities on the order of 55 cm/sec. The decrease in

shear resistance after initial increase with velocity is assumed to be a boundary effect as

the coupon approached the specimen base. In looser specimens, the coupon resistance

was also seen to increase, but at a much slower rate that decreased with increasing

velocity. The effective stress shown before liquefaction in Figure 5.10 is not the initial

effective stress on the specimen because only the last cycle of loading is plotted.

Coupon stops when 120 4 weight hanger system hits support I 100 Initial liquefaction •O 80 1 Effective stress (kPa) ° P ** «• ii.ti/*ll'r|M|f|l|,|,%'\.A £ * *i uv1;-1' *' 60

: Minimum j 2 40 4 Coupon shear (kPa) coupon ! resistance I * 20 + Coupon velocity (cm/sec) i -

38.5 38.7 38,9 39.1 Time [sec] Figure 5.10. Typical modified triaxial test result.

Figure 5.11 shows the results plotted verses relative density. Figure 5.12 shows the residual strength ratios plotted verses relative density. Residual strength values reported for the modified triaxial tests are the minimum observed values. Table 5.2 shows the testing conditions and results.

86 20 T

R2 = 0 8135

20.0 30.0 40.0 50.0 60.0 Relative Density (%) Figure 5.11. Modified triaxial test results.

u.zu - 0.18 - .2 0.16 - n * 0.14 - JZ ; f> 0.12 - R2 = 0 8065 & 0.10 - • 1 0.08 - "O / • S 0.06 - cc 0.04 - • 0.02 - 0.00 --__, , , , 1 , , , , 1 , , , , L— J—'—'—'—1 20.0 30.0 40.0 50.0 60.0 Relative Density (%) Figure 5.12. Modified triaxial residual strength ratio plot.

87 Table 5.2. Modified triaxial testing conditions and results

Relative Effective Velocity at Residual Density Stress Minimum Strength Test (%) (kPa) Sur (cm/sec) (kPa) A 26.0 120 52 1.23 B 27.5 120 90 1.34 C 44.1 115 58 11.4 D 53.0 117 55 8.7 1 E 47.8 120 48 4.7

88 6. RING SHEAR RESULTS AND COMPARISONS

Ring Shear Results

Figure 6.1 presents the residual strength values obtained from successful ring

shear tests on F-75 sand. The data are tabulated in Table 6.1 and show that, as expected, residual strength increases with both relative density and shearing speed. Exponential trend lines of the different shearing speed results generally have high coefficients of determination (R2) values, indicating a low proportion of variability within the data set.

Test results from the 6.8 cm/s shearing speed have a relatively low R2 value and

show a decrease in Sur with relative density. This behavior is considered an artifact of the testing system. The average residual strength measured for all speeds and relative densities is about 1 psi (7 kPa). This is small in comparison to the average friction caused by just the o-rings, about 3 psi (20 kPa) on average. The 13.5 cm/s shearing speed results have a much higher R2 value than the 6.8 cm/s speed, but they appear relatively flat, with the exception of the result from Test 12 (45.8% Dr). Residual strength at higher shearing speeds was easier to detect and separate from the o-ring friction due to greater liquefied soil resistance, and less variability in results were obtained.

Figure 6.2 shows the RS results plotted versus the residual strength ratio (Sur/a'v).

The data have more scatter when plotted this way, with generally lower R values.

Results from both the 13.5 cm/s and 6.8 cm/s shearing speeds decrease with Dr in this plot.

89 R2 = 0.9349

R2 = 0.9928

R2 = 0,7043

R2 = 0.3831 • 2.0 + Test 12 i 45.8% Dr 0.0 —I • « « • 1 • « « « 1— 45.0 50.0 55,0 60.0 65.0 Relative Density [%) Figure 6.1. Ring shear tests results.

0.35 -] Shearing • speed (em/sec) 0.30 ; # • 27.0 D 20.3 - A 13.5 o O 6.8 *rf 0.25 - ftJ - : a to 0.20 : - OaJ s X- 4-* CO 0.15 CD . 3 : A to 2 0) 0.10 • ; Test 12 27.0cm/s_ R - 0.4320 al 45.8% Dr 2 R = 0.8943 R2-0.0182 0.05 4 R2 = 0.5019

0.00 45.0 50.0 55.0 60.0 65.0 Relative Density [%) Figure 6.2. Ring shear data residual strength ratio plot.

90 1. Ring shear testing conditions and results. Relative Effective Initial Pore Shearing Residual Test Density Stress Pressure Speed Strength Number (%) (kPa) (kPa) (cm/s) (kPa) 4 56.1 111.6 102.7 6.8 3.6 13.5 8.0 20.3 27.0 11.5 5 50.0 113.6 103.4 6.8 13.5 20.3 6.8 27.0 9.0 11 58.0 125.0 103.4 6.8 3.6 13.5 7.9 20.3 9.5 27.0 11.4 12 45.8 25.0 103.4 6.8 13.5 3.1 20.3 5.6 27.0 7.5 18 53.3 138.2 102.7 6.8 4.2 13.5 7.5 20.3 8.2 27.0 9.7 19 62.8 138.7 103.4 6.8 3.9 13.5 7.7 20.3 11.7 27.0 12.8 103 52.5 116.8 105.5 6.8 4.9 27.0 9.1 106 53.1 120.5 105.5 6.8 5.7 27.0 9.1 | Note: —, no valid result.

Test 12 is an outlier on Figure 6.2 because the initial effective stress was very low

(explained in Section 4), making the Sur/a'v ratio much higher. The exponential trend lines and R2 values shown in Figure 6.2 do not include the Test 12 results.

91 Comparison of Test Results

Results from the different tests (centrifuge, ring shear, and modified triaxial) have similar trends of increasing residual strength with relative density. Figure 6.3 is a comparison of the trend lines of results from the various test methods.

It should be noted that Sur minima observed after liquefaction were not at the same velocity (shear strain rate) between the different tests; however, the values determined by the different tests are comparable. Sandoval et al. (2010) concluded from ring shear tests on Holliston 00 sand that the liquefied sand can be modeled as a shear- thinning fluid, in which resistance (Sur) increases with increasing shear strain rate, but at a decreasing rate. The Holliston 00 data suggests that at relative densities in excess of

40%, Sur might be modeled as a constant, and not dependent on shear strain rate (velocity of shear application). At higher densities, the effect of shear velocity seems to have much less influence. Further discussion of this topic can be found later in Section 6.

As seen in Figure 6.3, the trends in the RS results (separated by shearing speed) fall between and slightly below the trends of the rough coupon centrifuge results measured by the drive motor (lower) and slip-ring (higher). The trend of the modified triaxial test results seems to fit well with the centrifuge and RS testing results. (However, the modified triaxial results are from smooth aluminum coupons and may not be directly comparable.) It is important to note that all tests agree on the order of magnitude of residual strength values.

If the centrifuge results are considered to be "true" field values, both of the small- scale laboratory experiments are seen to produce similar results. Based on the R2 values of the trend lines, the ring shear device produced the greatest consistency and least scatter

92 20,0 -r

18.0 4

16.0 4

14.0 4

12.0 4 00 c 0) 10.0 4 4-* */> 8.0 4 3 no 6.0 4 CC 4.0 4

2.0 4 Xv&Xt o.o 4 u-t-J H- H-^ 25 30 35 40 45 50 55 60 65 Relative Density (%) Figure 6.3. Exponential trends in testing results.

0.25 T

0.20 4

a: to 0.15 4 c 1/5 § o.io 4 tf-39»

a: 0.05 4 rJSSSs tfl^' jl* 0.00 1 1 I I i 1 I I I I i 1 -+" liillillllltltll 25 30 35 40 45 50 55 60 65 Relative Density (%) Figure 6.4. Residual strength ratio comparisons.

93 in data overall. Since the RS results have the same magnitude and trend of the centrifuge results, they can be considered a valid representation of "true" residual strength values.

Figure 6.4 is a comparison of residual strength ratio trend lines. Again, the trend

of the modified triaxial test data seems to fit well with the centrifuge and RS results and

gives the overall trend to lower relative densities. However in contrast to the centrifuge trends, the slope of the RS trends do not increase significantly with Dr. As mentioned earlier, the Sur/o'v has been used when analyzing field case histories. The centrifuge

"field" results appear to benefit from normalization, having less scatter in the data.

However, the laboratory-based RS data have more scatter when plotted this way. Further discussion about residual strength normalization is provided in the following section.

Comparison with Back-Calculated Field Values

The post-liquefaction residual strengths obtained from the different tests are compared to residual strengths obtained by back-calculation of liquefaction flow failure case histories in Figures 6.5 and 6.6. In the field case histories, back-calculated Sur values are correlated with equivalent clean-sand normalized Standard Penetration Test

(SPT) blowcounts [(Ni)6o-cs] derived from field SPT values, as originally proposed by

Seed (1987); although more recent analysis of the field case histories suggest that the clean-sand normalization is not necessary (Gutierrez et al., 2006). In order to compare with the laboratory data, it was necessary to convert the relative density of the test specimens to equivalent normalized clean-sand blow-counts. An approximate correlation proposed by Mayne et al. (2001) for clean sands was used:

2 (Ni)6o-cs = 60 (Dr/100)

94 50 19 Group 1 — Case histories with an adequate amount i e 3 ofin-siiu measurements fe g, SPT, CPT) and reasonably complete geometric details CO 40

w 30

S 2o

^ " i r2 ~"y J%*XL**** ' ' .lid- «+• t i 0 4 § 12 16 20 Equivalent clean sand SPT corrected blowcount (NJq^^ (after Idriss and Boulanger, 2008)

Figure 6.5. Comparison of back-calculated Sur with range of F-75 test result trends.

T—pi—r I i / Recommended Curve / for conditions where void redistribution effects A> / are expected to be negligible t£fi / -# •f

/ Recommended Curve * for conditions where -*- void redistribution effects -* cpuld be significant

shear OS >* ••;:;. .

i >AoC J L-J !_i ! !_L 10 15 20 25 30

Equivalent clean sand SPT corrected blowcount, (W^60cs_Sr (after Idriss andBoulanger, 2008) Figure 6.6. Back-calculated residual strength ratios with F-75 test result trends.

95 Idriss and Boulanger (2007) critically reviewed analyses of the field case histories by Seed (1987), Seed and Harder (1990) and Olson and Stark (2002) and listed Sur values they felt were best-documented. The larger symbols in the figures represent the values

Idriss and Boulanger considered to be best-documented. It should be noted that some field Sur values represent the same case history as interpreted by different researchers.

Figure 6.5 shows the range of residual strength trends from the different test methods investigated with the field case studies results and a design curve proposed by

Idriss and Boulanger (2008) based on the back-calculated values. From a practical point of view, the Sur trends from the fine F-75 sand plot below the design curve. Malvick et al. (2006) suggested that the design curve constitutes a lower bound for failure conditions in which drainage is impeded, for example by an overlying low-permeability layer.

Impeded pore pressure dissipation may produce a void redistribution and consequent loosening of the upper part of the confined sliding mass. Because the laboratory-obtained

Sur trends shown on Figure 6.5 were obtained under essentially zero effective stress; the design curve proposed by Idriss and Boulanger (2008) may represent situations where some level of drainage was possible. Indeed, it is difficult to visualize a sliding mass subjected to large deformations that will not break up in such a way as to allow some dissipation of pore pressure during sliding.

Figure 6.6 is a comparison of the residual strength ratio trends to the case histories. Idriss and Boulanger (2008) state that Sur/o'v is more effective for describing stress-strain behavior in undrained monotonic laboratory tests, up to moderate strain levels. It is believed that Sur/o'v better reflects the potential effects of strength loss induced by void redistribution. Void redistribution could cause shear resistance to locally

96 diminish to zero if a film of water forms, but the average shear resistance over a large area is unlikely to be zero (Idriss and Boulanger, 2008). The potential for void redistribution to cause significant slope deformation decreases quickly with increasing Dr due to less water being expelled by contracting zones and more that can be absorbed by dilating zones.

Two design curves are give in Figure 6.6 (the Sur/o'v plot) above an (Ni)6o-cs of about 9 blowcounts. The upper curve is based on conditions where dissipation of excess pore pressure would not be impeded by stratigraphy, and dissipation would be accompanied by consolidation of soil at all depths (Idriss and Boulanger, 2008). The lower curve corresponds to conditions where void redistribution effects are significant and dissipation of excess pore pressure is impeded. When compared to the field case histories, the Sur/o'v trends from this study also tend to plot below the design curves. It is interesting to note, however, that the centrifuge results follow the upper curve. These specimens consolidated during the simulated earthquake and excess pore pressure dissipated without hindrance. The RS results, on the other hand, are from constant volume tests with dissipation impeded, and tend to follow the lower curve.

Figure 6.7 shows the laboratory-obtained post-liquefaction residual strength trend lines compared to the case-history-based back-calculated values which have been correlated with normalized (CPT) clean sand equivalent cone resistance [Qtn,cs]

(Robertson, 2010). An approximate correlation proposed by Mayne et al. (2001) for clean sands was used to estimated cone resistance:

2 Qtn,cs = 300-(Dr/100)

97 Only one design curve is proposed by Robertson (2010). The F-75 results tend to

plot below the design curve, with the trend of centrifuge results following the design

curve best.

030 — n —Prcoosed

• ClassA -0 25 • Class B

So 20

it .So 15

-010 • «r s*gi 3 3* 10D5 • _^^i>s— .-?f .-«•• .«** */lQ3*' * 030 20 40 60 80 100

Normalized CPT clean sand eqnhalent cone resistance, CLlt (after Robertson, 2010) ,BA' Figure 6.7. CPT based residual strength ratio comparisons.

Gutierrez et al. (2006) argue against Sur normalization because of the relatively

small influence of overburden stress on the void ratio of sand and gravels at the shallow depths typical of liquefaction. Gutierrez et al. note that the critical state concept suggests that residual strength is only a function of grain shape, grain-size distribution, and void ratio. They also note that the critical state for a given void ratio should therefore be the

98 same, regardless of initial effective stress. However, Ishihara (1993) suggests that residual strength should be analyzed using the quasi-steady-state strength, when the soil changes from a contractive to dilative response, instead of the critical state strength

(constant volume). The quasi-steady-state is related to consolidation stress and can be normalized by initial effective stress (Gutierrez et al., 2006).

Idriss and Boulanger (2008) note that the void redistribution process is not fully clear at this time and that neither the direct correlation with Sur nor the Sur/o'v method fully account for the numerous factors that influence it. With regards to the Sur/a'v plot,

Physical and analytical models indicate that void redistribution is potentially most severe for loose sand and is likely to have played a role in many of the currently available case histories. This would suggest that the two design relations should be somewhat different at the lower penetration resistances, but the current state of knowledge does not provide a basis for incorporating any difference at this time (Idriss and Boulanger, 2008).

The centrifuge and ring shear Sur/o'v trends differ at higher densities and appear to differ at lower densities, but additional testing would be necessary to confirm this. It should be noted that the design curves presented by Idriss and Boulanger (2008) and

Robertson (2010) are not exponential fits of the data. Thus, any comparisons of the results of this research with those design curves are approximate at best, and should not be used outside the scope of this study.

Wijewickreme et al. (2010) suggest that back-analysis for the estimation of Sur is considered more suitable than laboratory testing, because the latter is not able to simulate the void redistributions, or water film effects, that take place after liquefaction, particularly in layered deposits with contrasting permeability. Field conditions are more complex than idealized laboratory experiments and the knowledge gained from four

99 decades of research on clean sands does not directly translate to natural and man-made fills (Thevanayagam et al., 2002).

Nonetheless, the values obtained with F-75 sand are much lower than what conventional practice would suggest, and caution should be used when choosing residual strength values for design. Wijewickreme et al. (2010) reconciled that the evaluation of laboratory data provides important information towards understanding soil response in a fundamental manner, as well as to support and confirm field-based approaches; however, differences between field and laboratory conditions with respect to initial stress states, stress-strain history, loading mode, and drainage conditions are important considerations.

Comparison with Previous Ring Shear Results

Although previous testing with the UNH ring shear device was conducted on a different sand, a comparison of results is warranted. Figure 6.8 shows the Holliston 00 sand results reported by Sandoval (2007). In this plot, residual strength is plotted verse shearing speed, not Dr. Figure 6.9 shows the F-75 results plotted the same manner for comparison. In both plots, increasing Sur is observed with higher shearing speed, and the rate of increases decreases with speed, like a shear-thinning fluid. The Holliston 00 data was related to shear strain rate (Yuan, 2009) using the Herschel-Bulkley model

(Hemphill et al, 1993):

m

where y is the shear strain rate (in the order of 50 to 100 sec*1). In general, shear stress

(x), yield shear stress (x0), and empirical parameters (K) and (m) are expected to depend

100 30.0 +

CC 25.0 Q- -* "•—^ s: to 20.0

5.0 +

0.0 6.6 13.3 19.9 26.6 Strain rate (cm/s) Figure 6.8. Previous ring shear results, Holliston 00 sand.

30.0

as 25.0 ja£. _c to 20.0 c L

5.0 +

0.0 6.8 13.5 20.3 27.0 Strain rate (cm/s) Figure 6.9. Current ring shear results, F-75 sand.

101 on factors such as grain-size distribution, fines content and plasticity, and relative density of the particular sand. It was shown that increasing relative density decreases the m parameter, and the contribution of shear strain rate to Sur becomes progressively smaller

(Sandoval et al., 2010). Also, the relative importance of yield shear stress increases with relative density, and likewise decreases the contribution of shear strain rate.

Sandoval et al. (2010) noted that Sur tended towards a constant value with increasing strain rate, as conventionally assumed. The Sandoval data also clearly show that for relative densities less than about 40%, a range common in liquefaction flowslides, residual strength would be significantly influenced by the shear strain rate.

Only three of the twenty specimen results used as part of the current study had a Dr under

40%. Therefore, the residual strength values obtained from different test shearing speeds and methods are considered comparable.

Figures 6.7 and 6.8 also show that Sur results from F-75 sand were much lower than the Holliston 00 sand, which made determining Sur at lower shearing speeds difficult. The lower shearing speed limit of the testing machine, when testing F-75 sand, appears to be between 13.5 and 20.3 cm/s. Due to the limit in progressive shearing velocity results per Dr, the F-75 results could not be readily be modeled as a shear- thinning fluid.

As mentioned in the Special Testing section of Section 4, the range of relatively densities reported for the different sets of RS data (F-75 and Holliston 00) varied by 25 to

30%. This author believes the major factor in the differences between reported Dr is from when the specimen density was determined during the specimen preparation process.

Material properties differences also likely contribute some to the observed variation.

102 7. CONCLUSIONS AND RECOMMENDATIONS

Conclusions

Liquefied material within a rapid flowslide retains a relatively small internal

strength, a residual strength (Sur). Most geotechnical lab equipment in current use cannot reproduce the high shear strain rates and large shear strains which occur in a sliding mass

in the field, and thus cannot measure Sur. Although centrifugal testing is the ideal "field

experiment" to obtain "field" Sur values, centrifuge tests of this nature can be expensive to conduct. Therefore, a relatively simple and reliable measurement technique, like the ring shear device, is needed. The ring shear device at the University of New Hampshire

is designed to measure the low Sur values of a liquefied soil.

A study involving centrifuge, modified triaxial, and ring shear testing was undertaken to better understand residual strength, and the change in soil strength from the onset of liquefaction through dissipation of excess pore pressure. The centrifuge results are considered to represent true field values. A fine Ottawa F-75 sand was chosen for testing.

The ring shear device has its limitations due to relatively high and irregular friction caused by the o-rings, which are necessary to maintain pore pressure. Although the overall success rate of the testing program was about 30 percent, the testing procedure is relatively simple and testing can be conducted by one person in roughly one-half of a day. Many tests can be conducted efficiently in a relatively short time, allowing for

103 quick identification and analysis of trends in the data. Also, testing at different shearing

speeds can be conducted on a single specimen (at a constant relative density).

Centrifuge testing, on the other hand, can be time consuming, taking a full day per

specimen, and the testing expense is great. With the method used, only two shearing

speeds could be conducted per specimen. The Drof the specimen increases during a test,

precluding additional tests on the same specimen (the consolidated relative density of the

specimen after testing is usually well above that of natural sand deposits susceptible to

liquefaction and flowslides).

A third possibility, modified triaxial testing, is limited in the total amount of possible shearing, and shearing speed cannot be controlled, making data analysis

difficult. However, loose specimens for testing are easily obtainable, unlike the

centrifuge and rings shear test methods.

Comparison of Sur values from the three experiments previously described show that the small-scale laboratory tests produce similar values and variation with relative density as the centrifuge tests, considered to be the true "field" values for a liquefied deposit of the very fine uniform sand. The UNH ring shear device produced values and trends in data very close to the centrifuge results; therefore, it appears to have been proven reliable.

From a practical point of view, the Sur values from this study tend to plot below proposed design curves that may represent situations where some level of drainage was possible (see Figures 6.4 and 6.5). With regard to the actual values, it should be noted that the measured residual strengths were obtained under conditions of zero effective stress and thus may approximate true minimum values. Consequently, Sur values

104 obtained from laboratory tests, such as those described in this research, may represent a conservative lower bound to those that might occur in the field.

The combined results of these experiments have significantly contributed to modeling of large-strain shear strength of liquefied sand, clarifying its dependency on key factors such as relative density, strain rate, and increase in post-liquefaction strength with dissipation of excess pore pressure. Previous testing with the RSD on the coarser

Holliston 00 sand (Sandoval et al. 2010) resulted in higher Sur values when compared to the F-75 sand results, which strongly suggests that residual strength is also sensitive to the grain size and gradation of the granular soil involved.

Recommendations

As far as the testing machine, it is recommended that the thrust/torque load cell be replaced by a torque-only load cell to avoid the occasional loss of data when torque is recorded as thrust. The coupling of the connection of the center rod housing and drive motor should be keyed to prevent slippage. The time stamp on the LabVIEW data should be fixed to make data reduction somewhat easier.

New dummy blocks should be made of a relatively non-compressible material.

The ones currently used to determine sample volume, a very important variable, compress about 0.02 inches when being loaded for measurement. Due to the small specimen size, this has the effect of changing the specimen Dr by over 11%, if compression is not accounted for.

The sample chamber should be modified to reduce the friction of the o-rings. If a new chamber is created, a flat sample bottom (not sloped) should be chosen because, (1)

105 it appears that shearing is limited to the top of the sample and vertical shear strain is not uniform even though the specimen is thicker on the outside. (2) Uniform height dummy blocks could be used.

Improvement to the back pressure chamber to allow for precise control of pore pressure would allow for better characterization of soil strength loss and subsequent regain at low effective stresses. Effective stresses are greater than zero at the onset of strength loss in loose soils and can remain above zero even at large displacements

(Konrad and Watts, 1995).

The specimen preparation method was revised for this study from what was previously used to better suit the material being tested. Updates to the testing program were also made. Because of this, ring shear testing of the Holliston 00 sand should be conducted to identify valid trends in the data for comparison to the previous results.

Because the RSD appears to have been proven reliable, it can be extended to specific materials involved in seismic hazards studies. Higher shearing speeds should be investigated and may improve the noise to signal ratio of the testing machine; however, caution should be used when increasing shearing speeds until the machine reaction is understood.

Sands with different gradations and fines contents should be tested. An associated variable, the content of fines, must therefore be considered as well. De Alba and Ballestero (2008) concluded, from modified triaxial tests on sands with fines, that fines content over 10% (and the plasticity of the fines) made a significant difference in residual strength. Wijewickreme et al. (2010) concluded, from normal triaxial tests on

106 tailing specimens, that when the fines content just fill the void spaces of the coarser material (a homogeneous mixture referred to as "paste rock"), strain-softening accompanied by loss of shear strength did not occur and the material was unlikely to experience flow failure. Thevanayagam et al. (2002) concluded, from triaxial tests of collapse potential on a range of soils from pure sand to pure silt, that silty sand is

somewhat stronger that clean sand, and sandy silt is typically stronger than pure silt.

However, depending on if the fines just fill the intergranular void spaces or if they begin to separate the coarse grains, the stability of the soil can vary greatly. All of the aforementioned references should prove valuable to a study of this nature.

Articles of interest not used in this text can be found at the end of reference section that follows.

107 8. REFERENCES

Alshibli, K.A. (2002). Educational Brief Using Space for a Better Foundation on Earth Mechanics of Granular Materials, National Aeronautics and Space Administration, Educational Product EB-2002-03-53-MSFC. Alshibli, K.A., Batiste, S.N., and Sture, S. (2003). "Strain localization in sand: plane strain versus triaxial compression," Journal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 129, No. 6, 483-494. Alshibli, K.A. and Hasan, A. (2008). "Spatial variation of void ratio and shear band thickness in sand using X-ray computed tomography," Geotechnique, Vol. 58. No. 4, 249-257. Bayan, G.K. (2008). "Static Liquefaction for Flow Type Landslide at Karshingsa - A Case Study," The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, Goa, India. Baziar, M.H. and Dobry, R. (1995). "Residual strength and large deformation potential of loose silty sands," Journal of Geotechnical Engineering, Vol. 121, No. 12, 896- 906. Biarez, I., Boucraut, L.M., and Negre, R. (1965). "Limiting equilibrium of vertical barriers subjected to translation and rotation forces," Proc. Of the 6th Int. Conf. on Soil Mechanics and Foundation Engineering, Vol. II, Montreal, Canada, 368-372. Bryant S.M., Duncan J.M., and Seed H.B. (1983). "Application of tailings flow analyses to field conditions," Report UCB/GT/83-03 to U.S. Bureau of Mines, Department of Civil Engineering, University of California, Berkeley, 312 p. Byrne, P.M., Imrie, A.S., and Morgenstern, N.R. (1994). "Results and implications of seismic performance for Duncan Dam," Canadian Geotechnical Journal, Vol. 31, No. 6, 979-988. Casagrande, A. (1965). "Second Terzaghi Lecture: The role of "calculated risk" in earthwork and foundation engineering," Journal of Soil Mechanics and Foundations Division, ASCE Vol. 91, SM4, 1-40. Castro, G. (1995). "Empirical methods in liquefaction evaluation," Primer Ciclo d Conferencias Internationales, Leonardo Zeevaert, Universidad National Autonoma de Mexico, Mexico City.

108 Castro, G. (1975). "Liquefaction and cyclic mobility of saturated sands," Journal of Geotechnical Engineering, ASCE Vol. 116, No. 5, 805-821. Castro, G. (1969). Liquefaction of Sands, Ph.D. Dissertation, Harvard University. Castro, G., Seed, R.B., Keller, T.O., and Seed, H.B. (1992). "Steady-state strength analysis of Lower San Fernando Dam slide," Journal of Geotechnical Engineering, ASCE, Vol. 118, No. 3, 406-427. Charlie, W.A., Siller, T.J., and Doehring, D.O. (1998). "Measurement of post- liquefaction shear strength from piezovane field tests," Proceedings of the National Science Foundation Workshop on Post-Liquefaction Shear Strength of Granular Soils, University of Illinois at Urbana-Champaign, April 1997, Stark, Olson, Kramer, and Youd, (eds), 107-112.

Chellis, R.D. (1951). Pile Foundations, Theory-Design-Practice, McGraw-Hill Book Company, Inc., New York, New York. Chen, C-W. (2006). "Drained and Undrained Behavior of Fiber-reinforced Sand," Midwest Transportation Consortium of Student Papers, Transportation Scholars Conference, Iowa State University, Ames, Iowa. Collins, R. (2010). Personal communication with UNH Civil Engineering department head following second seminar presentation of this research, September, 10. Das, B.M. (2008). Fundamentals of Geotechnical Engineering, Third edition, Cengage Learning. Davis, A.P., Castro, G., and Poulos, S.J. (1988). "Strengths back figured from liquefaction case histories," 2nd Int. Conf. On Case Histories in Geotechnical Engineering, Rolla, Missouri. de Alba, P. A. and Ballestero T.P. (2008). "Effect of Fines on Sand Residual Strength after Liquefaction," Geotechnical Earthquake Engineering and Soil Dynamics IV Conference, Sacramento, CA, May. de Alba, P.A., and Ballestero, T.P. (2006). "Residual strength after liquefaction: a rheological approach," Soil Dynamics and Earthquake Engineering, Vol. 26, No. 2-4, 143-151. de Alba, P.A. and Ballestero T.P. (2005). "Liquefied granular materials as non- Newtonian fluids: a laboratory study," Proceedings, Geo-Frontiers 2005, Geotechnical Special Publication 133, ASCE. de Alba, P.A. and Ballestero T.P. (2004). "Residual Strength after Liquefaction: a Rheological Approach," Proceedings, 11th Intl. Conf. on Soil Dynamics and Earthquake Engineering and 3rd Intl. Conf. on Earthquake Geotechnical Engineering, Vol. 2, 513-520.

109 Dewoolkar, M.M., de Alba, P.A., Hargy, J.A, and Anderson, I (2010). Measurement of the strength of liquefied soil in physical models, Report to the National Science Foundation, Washington, D.C., NSF Grant No. 0724080. Dewoolkar, M.M., Ko, H.Y., Stadler, A.T., and Astaneh, S.M.F. (1999). "A substitute pore fluid for seismic centrifuge modeling," Geotechnical Testing Journal, Vol. 22, No. 3, 196-210. Dijkstra, J. (2004). Influence of loading rate on pile capacity in unsaturated sand, M.S. Thesis, Delft University of Technology, Delft, Netherlands. Eckersley, J.D. (1990). "Instrumented Laboratory Flowslides," Geotechnique, Vol. 40 No. 3, 489-502. Evans, S.G. and Bent, A.L. (2004). "The Las Colinas landslide, Santa Tecla: A highly destructive flowslide triggered by the January 13, 2001, El Salvador earthquake," Natural hazards in El Salvador, The Geological Society of America, Inc. Boulder, Colorado, 25-37. Fear, C.E. and Robertson, P.K. (1995). "Estimating the undrained strength of sand: a theoretical framework," Canadian Geotechnical Journal, Vol. 32, No. 5, 859-870. Finn, W.D.L. (1998). "Seismic safety of embankment dams: development in research and practice 1988-1998," Proc. Geotechnical Earthquake Engineering in Soil Dynamics III, Dakoulas, Yegian and Holtz (eds), Geotechnical Special Publication No. 75, Vol. 2, 812-853. Goulding, R.B. (2006). Tensile strength, shear strength, and effective stress for unsaturated sand, Ph.D. Dissertation, University of Missouri - Columbia. Gu, W.H., Morgenstern, N.R., and Robertson, P.K. (1993). "Progressive failure of the St. Fernando Dam," Journal of Geotechnical Engineering, Vol. 119, No. 2, 333-348, in Hungr, et al., 2002. Gutierrez, M. (2006). Probabilistic residual shear strength criteria for post-liquefaction evaluation of cohesionless soil deposits, Final Technical Report to the U.S. Geological Survey, Department of the Interior, USGS Award No. 04HQGR0076. Hemphill T., Campos W., and Pilehvari A. (1993). "Yield-Power Law Model More Accurately Predicts Mud Rheology," Oil & Gas Journal, Vol. 91, No. 34, 45-50. Hungr, O., Dawson, R.F., Kent, A., Campbell, D., and Morgenstern, N.R., (2002). "Rapid flow slides of coal-mine waste in British Columbia, Canada," Catastrophic Landslides: Effects, Occurrence, and Mechanisms, Review in Engineering Geology Volume XV, The Geological Society of America, Inc. Boulder, Colorado, 191-208.

110 Idriss, I.M and Boulanger, R. (2008). Soil liquefaction during earthquakes, Earthquake Engineering Research Institute (EERI), Oakland, California. Idriss, I.M and Boulanger, R. (2007). "SPT and CPT-based relationships for the residual shear strength of liquefied soils," Proc. 4th Int. Conf. on Geotechnical Earthquake Engineering-Invited lectures, 1-21, Springer. Ishihara, K. (1993). "Liquefaction and flow failures during earthquakes," Geotechnique, Vol. 43, No. 3,351-415. Ishihara, K., Yasuda, S., and Yoshida, Y. (1990). "Liquefaction-induced flow failure of embankments and residual strength of silty sands," Soils and Foundations, Vol. 30, No. 3,69-80. Jefferies, M.G., Been, K., and Hachey, J.E. (1990). "Influence of scale on the constitutive behavior of sand," Proc. Canadian Geotechnical Engineering Conference, Laval University, Quebec, Canada, Vol. 1, 263-273. JSF T 161-1990 (1990). "Japanese Industrial Standard Test Method for Minimum and Maximum Densities of Sands," Japanese Industrial Standards Association, 8 p. Konrad, J.M. and Watts, B.D. (1995). "Undrained shear strength for liquefaction flow failure analysis," Canadian Geotechnical Journal, Vol. 32, No. 5, 783-794. Kramer, S.L. (1996). Geotechnical Earthquake Engineering, Prentice Hall. Malvick, E.J., Kutter, B. L., Boulanger, R.W. and Kulasingam, R. (2006). "Shear Localization due to liquefaction-induced void redistribution in a layered infinite slope," J-Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 132, No. 10, 1293-1303. Mayne, P.W., Christopher, B.R. and DeJong, J. (2001). "Manual on Subsurface Investigations," National Highway Institute, Pub. FHWA NHI-01-031. Merifield, R.S. and Sloan, S.W. (2006). "The ultimate pullout capacity of anchors in frictional soils," Canadian Geotechnical Journal, Vol. 43, No. 8, 852-868. Olson S.M. and Stark T.D. (2002). "Liquefied strength ratio from liquefaction flow failure case histories," Canadian Geotechnical Journal, Vol. 39, No. 3, 629-647. Orfano, F. (2009). Landslides, edited and published by Stonecypher, L., available at http://www.brighthub.com/engineering/civil/articles/60082.aspx. Ovesen, N.K. (1964). "Anchor slabs, calculation methods and model tests," Bulletin No. 16, Danish Geotechnical Institute, Copenhagen, Denmark. Palmer, A.C. (1999). "Speed effects in cutting and ploughing," Geotechnique, Vol. 49, No. 3, 285-294.

Ill Parker Automation (2001). Gemini Series Programmer's Reference, Revision E, Parker Hannifin Corporation. Poulos, S.J., Robinsky, E.I., and Keller, T.O. (1985). "Liquefaction resistance of thickened tailings," Journal of Geotechnical Engineering Division, 111 (GT12), 1380-1394. Robertson, P.K. (2010). "Evaluation of flow liquefaction and liquefied strength using the cone penetration test," Journal of Geotechnical and Geoenvironmental Engineering, Vol. 136, No. 6, p. 842-853. Robertson, P.K. (1990). "Evaluation of residual shear strength of sands during liquefaction from penetration tests," Proc. Canadian Geotechnical Engineering Conference, Laval University, Quebec, Canada, Vol. 1, 257-262. Sadrekarimi, A. and Olson, S.M. (2010). "Particle damage observed in ring shear test on sands," Canadian Geotechnical Journal, Vol. 47, No. 5, 497-515. Sandoval, J. (2007). Liquefaction and ring shear device. DRAFT Ph.D. Dissertation, University of New Hampshire. Sandoval, J., de Alba, P.A., and Fussell, B. (2010). "Residual Strength of Liquefied Sand Measured in a Ring Shear Device," Geotechnical Testing Journal, Vol. 33, No. 1, 1-7. Seed, H.B. (1987). "Design problems in soil liquefaction," Journal of Geotechnical Engineering, ASCE, Vol. 113, No. 8, 827-845. Seed, H.B., Seed, R.B., Harder, L.F., and Jong, H.L. (1989). Reevaluation of the Lower San Fernando Dam: Report 2, Examination of the post-earthquake slide of February 9, 1971, U.S. Army Corps of Engineers Contract Report GL-89-2. Seed, R.B. and Harder L.F. (1990). "SPT-Based Analysis of cyclic pore pressure generation and undrained residual strength," Proceedings of the H. Bolton Seed Memorial Symposium, Vol. 2, BiTech Publishers Ltd. 351-376. Siebert, L. (2002). "Landslides resulting from structural failure of volcanoes," Catastrophic Landslides: Effects, Occurrence, and Mechanisms, Review in Engineering Geology Volume XV, The Geological Society of America, Inc. Boulder, Colorado, 209-235. Stark, T.D. and Mesri, G. (1992). "Undrained strength of liquefied sand for stability analysis," Journal of Geotechnical Engineering, ASCE, Vol. 118, No. 11, 1727- 1747.

112 Stark, T.D., Olson, S.M., Kramer, S.L., and Youd, T.L. (1998). "Shear Strength of Liquefied Soils," Proceedings of the National Science Foundation Workshop on Post-Liquefaction Shear Strength of Granular Soils, University of Illinois at Urbana-Champaign, Urbana, Illinois, April 1997, p. 288. Stewart, D.P., Chen, Y.R. and Kutter, B.L. (1998). "Experience with the use of methylcellulose as a viscous pore fluid in centrifuge models," Geotechnical Testing Journal, Vol. 21, No. 4, 365-369. Sture, S., Batiste, S.N. Lankton, M„Costes, N.C., Alshibli, K.A., Swanson, R.A., and Frank, M., (1999). "Mechanics of Granular Materials: Final Report, STS-79 and STS-89 Experiments," NASA Contract NAS8-38779, Report to NASA, Johnson Space Center. Terzaghi, K.A (1925). Erbdaumechanik aufbodenphysikalischer Grundlage, Deuticke, Vienna, Austria. Thevanayagam, S., Shenthan, T., Mohan, S., and Liang, J. (2002). "Undrained fragility of clean sands, silty sands, and sandy silts," Journal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 128, No. 10, 849-859. Vaid, Y.P. and Sivathayalan, S. (1999). "Fundamental factors affecting liquefaction susceptibility of sands," Proc. Physics and Mechanics of Soil Liquefaction, Balkema, Rotterdam, 105-120. Vasquez-Herrera, A. and Dobry, R. (1989). "Reevaluation of the Lower San Fernando Dam, Report 3, The behavior of undrained contractive sand and its effects on seismic liquefaction flow failures of earth structures," Contract Report Gl-89-2, U.S. Army Corps of Engineers, Waterways Experiment Station, Vicksburg, MS. Wijewickreme, D. Khalili, A., and Wilson, G.W. (2010). "Mechanical response of highly gap-graded mixtures of waste rock and tailings. Part II: Undrained cyclic and post-cyclic shear response," Canadian Geotechnical Journal, Vol. 47, No. 5, 566-582. William, H.S. (2004). Development of a true triaxial apparatus for soil testing, M.S. Thesis, Louisiana State University. Wood, D. (2004). Geotechnical Modelling, Applied Geotechnics Volume 1, Spon Press Yuan, C. (2009). Nonlinear analysis on shear strain data, Graduate short term paper, Department of Mathematics & Statistics, University of New Hampshire, 4 pages. Contact [email protected]. Zornberg, J.G. Costa, Y.D. and Bueno B.S. (2005). "Failure Mechanisms in Pipelines Bridging a Void," Geosynthetics Research and Development in Progress (GRI- 18), Geo-Frontiers 2005, Proc. of the Sessions of the Geo-Frontiers 2005 Congress.

113 Other References of Interest

Daouadji, A., Al Gali, H., and Darve, F. (2006). "Triggering mechanisms of soil instability," Monitoring, Simulation, Prevention and Remediation of Dense and Debris Flows, Wit Press, Lorenzini, G., Brebbia, C.A., and Emmanouloudis (eds), 273-282. Fourie, A.B., Rowe, D., and Blight, G.E. (1999). "The effects of infiltration on the stability of the slopes of a dry ash dump," Geotechnique, Vol. 49, 1-13. Handy, R.L. (2008). " "Liquefaction": Hydraulic compaction of soil near rammed piers," From Research to Practice in Geotechnical Engineering, ASCE Geotechnical Special Publication No. 180, Laier, J.E., Crapps, D.K., and Hussein, M.H. (eds), 251-258. Hungr, O. (1995). "A model for the runout analysis of rapid flow slides, debris flows, and avalanches," Canadian Geotechnical Journal, Vol. 32, 610-623. Mayoral, J.M. and Romo, M.P. (2008). "Geo-seismic environmental aspects affecting tailings dams failures," American Journal of Environmental Sciences, Vol. 4, No. 3,212-222. Mitchell, J.K. and Hon. M. (2008). "Mitigation of liquefaction potential of silty sands," From Research to Practice in Geotechnical Engineering, ASCE Geotechnical Special Publication No. 180, Laier, J.E., Crapps, D.K., and Hussein, M.H. (eds), 433-451- Rahhal, M.E. and Lefebvre, G. (2000). "Understanding the effects of a static driving shear stress on the liquefaction resistance of medium dense granular soils," Soil Dynamics and Earthquake Engineering, Vol. 20, 397-404. Sasitharan, S., Robertson, P.K., Sego, D.C., and Morgenstern, N.R. (1994). "State- boundary surface for very loose sand and its practical implications," Canadian Geotechnical Journal, Vol. 31, No. 3, 321-334. Sladen, J.A., D'Hollander, R.D., and Krahn, J. (1985). "The liquefaction of sands, a collapse surface approach," Canadian Geotechnical Journal, Vol. 22, No. 4, 564- 588. Thevanayagam, S. and Mohan, S. (2000). "Intergranular state variables and stress-strain behavior of silty sands," Geotechnique, Vol. 50, 1-23. VanDine, D.F. (1985). "Debris flows and debris torrents in the Southern Canadian Cordillera," Canadian Geotechnical Journal, Vol. 22, No. 1, 44-68. Wijewickreme, D., Khalili, A., and Wilson, G.W. (2010). "Mechanical response of highly gap-graded mixtures of waste rock and tailings. Part I: Monotonic shear response," Canadian Geotechnical Journal, Vol. 47, 552-565.

114 APPENDIX A

MACHINE CALIBRATION RESULTS UNH Ring Shear Device Initial Calibration Check - 6/16/2009

Bag pressure (transducer by keyboard) (10 psi = 1 volt) psi volts initial 10.0 0.993 (adjusted with resistance screw) final 10.0 1.000 20.0 2.000

Back pressure (monintor at top of chamber) 1 (10 psi = 1 volt) psi volts initial 10.0 1.000 20.0 2.048 30.0 3.085 40.0 4.119 (adjusted with gain dial) final 10.0 0.986 15.0 1.496 20.0 2.014 30.0 3.036 40.0 4.049

Torque (16" torque arm crank and spring scale) torque (10 psi = 1 volt) load (lb) (inxlb) volts initial 0 0 0.131 30 480 0.592 45 720 0.803 (check correlation / calibration in LabVIEW)

Load cells (applied bag pressure) (psi) LCI (volts) LC2 (volts) 0.0 0.157 0.128 10.0 0.273 0.308 20.0 0.425 0.469

All transducers appear to be in working order

116 UNH Ring Shear Device Calibration Check - 7/13/2009

Bag and Pore pressure (psi) Bag Pore | Applied pressure pressure 0.0 0.07 0.13 5.0 4.95 4.85 10.0 10.1 9.95 15.0 14.9 14.9 20.0 20.0 20.1 25.0 24.9 25.1 1 30.0 29.9 30.3

^Pressure transducers appear to be in working order:

30.0

to 20.0 CO

Q_ <=> TJ Q_ to CO 10.0 CD

0.0 Applied Q Q 10.0 20.0 30.0 •>*•• Bag pressure *H«Pore pressure Applied Pressure (psi) Load Cell Check |Zero load Cell #1 Cell #2 Initial -52.5 -38 | Final -1.6 -1.5

117 UNH Ring Shear Device Torque Cell Initial Calibration -06/26/09

Torque LabVIEW LabVIEW Weight Applied Reading Reading (kg) (inxlb) (pre) (post) 0 -290 65 211 223 352 360 492 360 1000 875 982

(Initial adjustment, should calibrate for the 1,000 in-lb range.) (Maximum load cell capacity is 10,000 inch-pounds.)

2000

£ 1500 c c S> 1000

CD 13 P"

500

0 Actual o ro O CJ1 o — & Pre-adjustment o o o o o •'•if Post-adjustment o o o Applied Torque (in*lb)

118 UNH Ring Shear Device Torque Cell Calibration Check - 06/29/09

Torque LabVIEW LabVIEW Weight Applied Reading Reading (kg) (inxlb) (load) (unload) 0 0 -1.8 -1.2 14 494 480 520 28.5 1005 904 955 45 1587 1550 1640 1 57 2011 1980 2070 2000

1500

1000

500

0 Actual O O o o Load o o o •• Unload o o o o Applied Torque (in*lb)

119 UNH Ring Shear Device Torque Cell Calibration Check - 07/08/09

Torque LabVIEW LabVIEW Weight Applied Reading Reading (kg) (inxlb) (load) (unload) 0 0 -1.2 -3.9 14 494 505 611 28.5 1005 1000 1160 45 1587 1550 1840 1 57 2011 2040 1990

2000

£ 1500 c

C

CD CD 1000 or

500

0 Actual o Ol _». _*. ro o o en o — B- Load o o o o ...4,.. Unload o o o Applied Torque (in*lb)

120 UNH Ring Shear Device Torque Cell Calibration Check- 07/20/10

lorque LabVltW LabVltW Weight Applied Reading Reading (kg) (inxlb) (load) (unload) 0 0 11 -1.9 176 164 195 317 275 314 13 459 408 500 17 600 529 645 21 741 670 801 25 882 789 945 29 1023 914 1125 33 1164 1035 1251 37 1305 1153 1412 41 1446 1279 1542 45 1587 1393 1612 49 1728 1538 1701 53 1870 1647 1774 57 2011 1797 1823 2000

1500

D) C CO CD or 1000 CD

ro o cn o — &• Load o o o o o o ...&.. Unload Applied Torque (in*lb)

121 UNH Ring Shear Device Torque Cell Calibration Setup

Looking down at the machine. Note: Lateral shaft must be removed for calibration.

122 UNH Ring Shear Dummy Block Compressibility

7/24/09 0.610 \ 0.605 \ .1 0.600 \ X eh 0.595 \ CD

^ 0.590 ^ y = -0 00"x +05996 £ ^ISSL* R2 = 0 9925 £ 0.585 - Q 1 0.580 0.575 1 1 1 1 1 1 1 1 • i i i 0.0 5.0 10.0 15.0 20.0 25.0 Bag Pressure (psi) 7/28/09 0.610 0.605 "S o c 0.600 \ .g> 0.595 CD F A\ 0.590 V, >> y = -0 001x +0 5987 £ F ^"s< R2 = 0 991 £ 0.585 Nk F *^ Q ^>^ 0.580 ^^2*^ 111 1 0.575 1 1 1 1 i i i i i i i i ir*^BK J 0.0 5.0 10.0 15.0 20.0 25.0 Bag Pressure (psi)

123 UNH Ring Shear Device Top Ring and Sample Chamber Dimensions

Top ring Sample chamber inner outer inner outer | location diameter diameter | location diameter diameter 1 8.207 12.145 1 8.138 12.219 2 8.212 12.136 2 8.133 12.211 3 8.202 12.135 3 8.135 12.215 1 average (in.) 8.207 12.139 average (in ) 8.135 12.215 2 area (in. ) 62.83 area (in. ) 65.21

<^+Wty>mm&yq^^'tyt

124 APPENDIX B

F-75 SAND MATERIAL PROPERTY TESTING DATA

125 UNH Sieve Analysis F-75 Sand

Date: 6/15/09 Material: F-75 sand Tested by: J. Hargy Total weight tested (grams): 141.8

Sieve Opening Weight Retained Percent Percent | Sieve Size (mm) (grams) Retained (%) Passing (%) #30 0.600 0 0.0 100.0 #40 0.425 0.2 0.1 99.9 #50 0.300 4.8 3.5 96.5 #80 0.180 80.5 60.3 39.7 #100 0.150 17.9 72.9 27.1 #200 0.075 36.6 98.7 1.3 1 Pan 1.7 99.9 0.1 |

1 D60 0.2075 1 1 Cu 2.56 1 D50 0.200 | | Cc 0.69 | D30 0.1075 | D10 0.081 |

1.00 0.10 0.01 Sieve Size (mm)

126 UNH Maximum and Minimum Void Ratio Tests F-75 Sand

Date: 6/15/09 Test Methods: Japanese Standard Methods Tested by: J. Hargy Material: F-75 sand Mold I ID|60mm I depth 40 mm I weight 827.2 g volume 113.1 cm

Minimum void ratio Maximum void ratio total weight of total weight of trial weight sand (g) trial weight sand (g) 1 989.8 162.6 1 1023.2 196.0 2 989.9 162.7 2 1022.7 195.5 3 988.4 161.2 1 3 1022.4 195.2 4 988.5 161.3 1 average 195.57 1 5 987.9 160.7 1 unit weight (g/cm ) 1.729 average 161.70 | Void ratio 0.5325 unit weight (g/cm ] 1.430 1 Void ratio 0.8535

Date: 7/7/09 Test Methods: Japanese Standard Methods Tested by: J. Hargy Material: oven dried F-75 sand Minimum void ratio Maximum void ratio total weight of total weight of trial weight sand (g) trial weight sand (g) 1 991.8 164.6 1 1023.0 195.8 2 991.6 164.4 2 1023.5 196.3 3 990.8 163.6 1 3 1022.9 195.7 4 991.5 164.3 average 195.93 5 991.2 164.0 unit weight (g/cm3) 1.732 6 989.9 162.7 | Void ratio 0.5297 1 7 992.9 165.7 average 164.19 unit weight (g/cm ] 1.452 | Void ratio 0.8255

127 UNH Permeability Tests F-75 Sand - Loose

Date: 11/2/10 Material: F-75 sand Tested by: J. Hargy Specimen volume (cm ] 944 Weight of mold (g] 3742 Mold and specimen (g] 5207 Dry density (g/cm3] 1.55 Specific gravity 2.65 Void ratio 0.7076 Maxe 0.805 Mine 0.486

%Dr 30.5

Constant head test | Test number 1 2 3 4 5 6 1 Average flow, Q (cm3] 250 250 250 250 250 250 Time, t (sec] 42.84 43.04 43.56 43.22 43.72 44.10 Head difference, h (cm] 57.5 57.5 57.5 57.5 57.5 57.5 Specimen diameter, D (cm] 10.16 10.16 10.16 10.16 10.16 10.16 Specimen length, L(cm] 11.643 11.643 11.643 11.643 11.643 11.643 | Specimen area, A (cm ] 81.07 81.07 81.07 81.07 81.07 81.07 | | Permeability, k (cm/s) 0.0146 0.0145 0.0143 0.0144 0.0143 0.0142 | Average k (cm/s] 0.0144 k=QL/A/7t

Falling head test | Test number 1 2 3 4 5 6 1 Specimen diameter, D (cm 10.16 10.16 10.16 10.16 10.16 10.16 Specimen length, L(cm 11.643 11.643 11.643 11.643 11.643 11.643 Specimen area, A (cm ] 81.07 81.07 81.07 81.07 81.07 81.07 Standpipe area, a (cm ] 0.4964 0.4964 0.4964 0.4964 0.4964 0.4964 Starting head, hi (cm] 106.5 106.5 106.5 106.5 106.5 106.5 Ending head, h2 (cm] 56.5 56.5 56.5 56.5 56.5 56.5 1 Time, t (sec] 3.47 3.44 3.50 3.34 3.34 3.34 1 Permeability, k (cm/s) 0.0130 0.0131 0.0129 0.0135 0.0135 0.0135 | Average k (cm/s] 0.0133 k=L/t*a// \*\n{hl/h 2)

128 UNH Permeability Tests F-75 Sand - Dense

Date: 11/2/10 Material: F-75 sand Tested by: J. Hargy Specimen volume (cm ) 944 Weight of mold (g) 3737 Mold and specimen (g) 5366 Dry density (g/cm3) 1.73 Specific gravity 2.65 Void ratio 0.5357 Maxe 0.805 Min e 0.486

%Dr 84.4

Constant head test | Test number 1 2 3 4 5 6 7 8 9 1 Average flow, Q (cm3) 250 250 250 250 250 250 250 250 250 Time, t (sec] 60.56 61.15 62.63 63.66 64.22 66.09 66.12 66.84 67.97 Head difference, h (cm) 57.5 57.5 57.5 57.5 57.5 57.5 57.5 57.5 57.5 Specimen diameter, D (cm) 10.16 10.16 10.16 10.16 10.16 10.16 10.16 10.16 10.16 Specimen length, L (cm) 11.64 11.64 11.64 11.64 11.64 11.64 11.64 11.64 11.64 | Specimen area, A (cm ) 81.07 81.07 81.07 81.07 81.07 81.07 81.07 81.07 81.07 | Permeability, k (cm/s) 0.0103 0.0102 0.0100 0.0098 0.0097 0.0094 0.0094 0.0093 0.0092 | Average k (cm/s) 0.0097 k=QL/Aht

Falling head test | Test number 1 2 3 4 5 6 7 Specimen diameter, D (cm] 10.16 10.16 10.16 10.16 10.16 10.16 10.16 Specimen length, L (cm] 11.64 11.64 11.64 11.64 11.64 11.64 11.64 1 T Specimen area, A (cm ) 81.07 81.07 81.07 81.07 81.07 81.07 81.07 Standpipe area, a (cm2) 0.4964 0.4964 0.4964 0.4964 0.4964 0.4964 0.4964 Starting head, hi (cm) 106.5 106.5 106.5 106.5 106.5 106.5 106.5 Ending head, h2 (cm) 56.5 56.5 56.5 56.5 56.5 56.5 56.5 1 Time, t (sec] 5.06 5.03 4.87 5.00 4.93 4.93 4.97 1 Permeability, k (cm/s) 0.0089 0.0090 0.0093 0.0090 0.0092 0.0092 0.0091 | Average k (cm/s] 0.0091 k=L/t*a/A*\n(hl/h2)

129 APPENDIX C

RESIDUAL STRENGTH TESTING DATA

130 UNH Ring Shear Test Summary Sheet

Date: 07/10/09 Test1 T04 Material: F-75 sand e max1 0.805 Wdry (gr): 1494.7 Total vol' 0.0324 ft3 e min: 0.486 Wdry(lb)' 3.30 Dry UW- 101.69 pcf Dr (%) = 56.1 Void ratio1 0.626 Sample Height Sample Sample Dummy # Dial reading reading Difference height 1 0.300 0.574 0.274 0.854 2 0.300 0.605 0.305 0.885 3 0.300 0.557 0.257 0.837 Bag pressure (psi): 19.0 Average H (in) = 0.859 H dummy (in): 0.580

Total load kg lb B Value Test LCI 437 964 O lange in pore presure (psi)' 12.0 LC2 449 990 ChangiB in external \Dressur e (psi)' 10.2 886 1954 Angel of Friction 42.2 Coeffent of external pressure 0.329 Total stress* 31.1 psi B Value 1.09 Initial pore U: 14.9 33.9 : Final U Initial effective: 16.2 -2.8 • Final effective stress

Residual Strength Lab View Ruii l Test (a) O-ring (b) (a-b)/5.08'762.83 " Speed Speed Torque Before After Before After (rpm) (in/sec) (in-#) (ln-#) (ln-#) (psi) (psi) 5 2.6 1320 1147 0.54 ~ 10 5.2 1838 1486 Not 1.10 ~ 15 7.9 2015 1860 Measured 0.49 ~ 20 10.5 2149 2053 0.30 ~ Run 2 5 2.6 1563 1406 0.49 ~ 10 5.2 2018 1631 Not 1.21 ~ 15 7.9 2310 1781 Measured 1.66 - 20 10.5 2648 2118 1.66 -

Motion Planner Runl 5 2.6 - 10 5.2 iMot ^~ - ~ 15 7.9 Available ~ ~ 20 10.5 ~ - Run 2 5 2.6 ^^ = — - — 10 5.2 Not -^_ - ~ 15 7.9 _ Available ~ — 20 10.5 = -^ Jr - ~

131 UNH Ring Shear Test Summary Sheet

3UUU i Torque vs. Time (Test 4)

2750 20 rpm

2500 15 rpm /

2250 20 rpm 15 rpm *lfl 10 rprn/**^ ^ 2000 •o 10 rpm c 3 i 1750 - LabView

5 rpm

E" 1500

5 TP™^^j*J

1250 Run 2 ; ^ Run 1

1000 • i i i I I I - 'I "T • I— a ooooooooooooooooooooo ^^^-CNCMCMeOCOCOO^^t'tWWlOCDCO Time (seconds)

2.00 n

Potential Sur values (T 4) 1.75

v> 1.50

g 1.25

he 2 1.00 •{ ro 3 A LabVIEW, o-ring before, Run 1 1 0.75 -I —•— LabVIEW, o-ring before. Run 2

0.50 -] mm mm Average ofva I ues • *•-•• Answer 0.25 5 rpm 10 rpm 15 rpm 20 rpm

132 UNH Ring Shear Test Summary Sheet

Date: 07/14/09 Test: T05 Material: F-75 sand e max: 0.805 Wdry(gr): 1659 Total vol: 0.0364 ft3 e min: 0.486 Wdry(lb): 3.66 Dry UW: 100.49 pcf Dr (%) = 50.0 Void ratio: 0.646 Sample Height Sample Sample Dummy # Dial reading reading Difference height 1 0.319 0.685 0.366 0.948 2 0.320 0.706 0.386 0.968 3 0.319 0.716 0.397 0.979 Bag pressure (psi): 17.5 Average H (in) = 0.965 H dummy (in): 0.582

Total load kg lb B Value Test LCI 425 937 a lange in pore presure (psi): 7.3 LC2 471 1039 Changie in external pressure (psi): 10.1 896 1976 Angel of Friction 40.8 Coeffent of external pressure 0.347 Total stress: 31.4 psi B Value 0.82 Initial pore U: 15.0 30.2 : Final U Initial effective: 16.4 1.2 : Final effective stress

Residual Strength LabView Ruin l Test (a) O-ring (b) (a-b)/5.08' 762.83" Speed Speed Torque Before After Before After (rpm) (in/sec) (in-#) (ln-#) (in-#) (psi) (psi) 5 2.6 1483 1271 839 0.66 2.02 10 5.2 1721 2025 1600 -0.95 0.38 15 7.9 1911 2306 Test -1.24 - 20 10.5 2243 2540 Stopped -0.93 » Run 2 5 2.6 1968 1941 0.09 - 10 5.2 2416 2377 Not 0.12 - 15 7.9 2922 2607 Measured 0.99 - 20 10.5 3131 2716 1.30 -

Motion Planner Run 1 5 2.6 ~ 10 5.2 Not - - 15 7.9 Available -- ~ 20 10.5 ~ - Run 2 5 2.6 ~ - 10 5.2 Not ~ ~ 15 7.9 Available - ~ 20 10.5 — —

133 UNH Ring Shear Test Summary Sheet

3500 Torque vs. Time (Test 5) 20 rpm

3250 15 rpm 3000

2750 4

2500 20 rpm ^ 2250

15 rpm yi 5 rpm o 2000 • LabView Q. 10 rpm y a> 1750 3 vr 1500

Run 2 1250 Run 1

1000 o CM CO Time (seconds)

2.00 -I Potential Sur values (T 5) 1.75 J

*3» 1.50 \ a.

1.25 J c

CO 1.00 H TO LabVIEW, o ring before. Run i 3 0.75 J LabVIEW, o=nng before, Run a: 2 Ct) LabVIEW, o-rtng after, Run 1 0.50 J Answer

0.25 -I 5 rpm 10 rpm 15 rpm 20 rpm

134 UNH Ring Shear Test Summary Sheet

Date: 07/22/09 Test: Til Material: F-75 sand e max: 0.805 Wdry(gr): 1529.3 Tota 1 vol: 0.0330 ft3 e min: 0.486 Wdry(lb): 3.37 Dry UW: 102.07 pcf Dr (%) = 58.0 Void ratio: 0.620 Sample Height Sample Sample Dummy # Dial reading reading Difference height 1 0.246 0.528 0.282 0.862 2 0.259 0.556 0.297 0.877 3 0.250 0.556 0.306 0.886 Bag pressure (psi): 18.7 Average H (in) = 0.875 H dummy (in): 0.580

Total load kg lb B Value Test LCI 473 1043 a lange in pore presure (psi): 11.7 LC2 471 1039 Change in external pressure (psi): 11.8 944 2082 Angel of Friction 42.6 Coeffent of external pressure 0.323 Total stress: 33.1 psi B Value 1.00 Initial pore U: 15.0 32.3 : Final U Initial effective: 18.1 0.8 : Final effective stress

Residual Strength LabView Ruin l Test (a) O-ring (b) (a-b)/5.08' 762.83" Speed Speed Torque Before After Before After (rpm) (in/sec) (in-#) (in-#) (in-#) (psi) (psi) 5 2.6 810 680 825 0.41 -0.05 10 5.2 1096 740 936 1.11 0.50 15 7.9 1240 744 1085 1.55 0.48 20 10.5 1200 741 1329 1.44 -0.41 Run 2 5 2.6 838 657 874 0.57 -0.11 10 5.2 1189 674 1202 1.61 -0.04 15 7.9 1211 695 1170 1.62 0.13 20 10.5 1420 705 1297 2.24 0.38

Motion Planner Run 1 5 2.6 688 585 711 0.32 -0.07 10 5.2 1004 624 843 1.19 0.50 15 7.9 1090 727 1005 1.14 0.27 20 10.5 1127 677 1218 1.41 -0.29 Run 2 5 2.6 733~ 540 764 0.61 -0.10 10 5.2 1071 574 1086 1.56 -0.05 15 7.9 1065 676 1033 1.22 0.10 20 10.5 1318 643 1178 2.11 0.44

135 UNH Ring Shear Test Summary Sheet

1500 Torque vs. Time (Test 11) 20 rpm j» i$r 1400 J 1300 15 rpm ?] 120 rpm 15 rpm 10 rpm | i, 1200

~ 1100 c o 1000 a, 5 rpm | 900 vl 5 rpm r/vJ * LabView 800 Motion Planner 700

600

Run 2 500 Run 1

400 0 30 60 90 120 150 180 210 240 270 300 330 360 390 420 450 Time (seconds)

LabVIEW, o-nng before. Run 1

LabVIEW, o-nng before, Run 2

LabVIEW, o-nng after Run 1

LabVIEW, o-nng after, Run 2

Motion Planner, o-ring before. Run 1

3— Motion Planner, o-nng before. Run 2

•—Motion Planner, o-ring after, Run 1

^— Motion Planner, o-nng after, Run 2

— Average of values

Answer

5 rpm 10 rpm 15 rpm 20 rpm

136 UNH Ring Shear Test Summary Sheet

Date: 07/23/09 Test: T12 Material: F-75 sand e max: 0.805 Wdry(gr): 149373 Total vol: 0.0330 ft3 e min: 0.486 Wdry(lb): 3.29 Dry UW: 99 68 pcf Dr (%) = 45.8 Void ratio: 0.659 Sample Height Sample Sample Dummy # Dial reading reading Difference height 1 0.248 0.533 0 285 0.873 2 0.259 0.539 0.28 0.868 3 0.253 0.551 0.298 0.886 Bag pressure (psi): 11.5 Aveirag e H (in) = 0.875 H dummy (fn): 0.588

Total load kg lb B Value Test LCI 266 587 anang e in pore presure (psi): 20.0 LC2 265 584 Change in external pressure (psi): 11.3 531 1171 Angel of Friction 39.8 Coeffent of external pressure 0.360 Total stress: 18.6 psi B Value 1.33 Initial pore U: 15.0 18.4 : Final U Initial effective: 3.6 0.2 : Final effective stress

Residual Strength LabView Ruin l Test (a) O-ring (b) (a-b)/5.08''762.83 " Speed Speed Torque Before After Before After (rpm) (in/sec) (in-#) (in-#) (in-#) (psi) (psi) 5 2.6 663 863 718 -0.62 -0.17 10 5.2 734 804 913 -0.22 -0.56 15 7.9 803 808 1163 -0.02 -1.13 20 10.5 929 823 1368 0.33 -1.38 Run 2 5 2.6 674 728 899 -0.17 -0.70 10 5.2 895 751 1125 0.45 -0.72 15 7.9 1025 755 1375 0.85 -1.10 20 10.5 1142 769 1483 1.17 -1.07

Motion Planner Runl 5 2.6 546 714 617 -0.52 -0.22 10 5.2 655 712 847 -0.18 -0.60 15 7.9 739 739 1044 0.00 -0.95 20 10.5 875 791 1265 0.26 -1.22 Run 2 5 2.6 569 605 786 -0.11 -0.68 10 5.2 806 661 1048 0.46 -0.76 15 7.9 947 697^ 1209 0.78 -0.82 20 10.5 1066 745 1366 1.01 -0.94

137 UNH Ring Shear Test Summary Sheet

1500 Torque vs. Time (Test 12) 20 rpm

1400 +

1300 15 rpm

1200 20 rpm 10 rpm 1100 •o c VvW*' c 1000 15 rpm o. c -msr' 10 rpm 900 *^s 3 cr 800

700 LabView 600 Motion Planner

500 Run 2

400 —i r- 0 30 60 90 120 150 180 210 240 270 300 330 360 390 420 450 480 Time (seconds)

2.00

Potential Sur values (T 12) 1.75 H

* 1.50 Q. g> 1.25 H 0»

LabVIEW, o-nng be for e. Run 1 15 3 LabVIEW, o-nng before, Run 2 "5 •A— Motion Planner, o-ring before. Run 1

Motion Planner, o-ring before, Run 2

Answer

5 rpm 10 rpm 15 rpm 20 rpm

138 UNH RingShearTest Summary Sheet

Date: 07/29/09 Test: T18 Material: F-75 sand e max: 0.805 Wdry(gr): 1527.5 Tota 1 vol: 0.0333 ft3 e min: 0.486 Wdry(lb): 3.37 Dry UW: 101.14 pcf Dr (%) = 53.3 Void ratio: 0.635 Sample Height Sample Sample Dummy # Dial reading reading Difference height 1 0.250 0.567 0.317 0.897 2 0.255 0.561 0.306 0.886 3 0.251 0.535 _ 0.284 0.864 Bag pressure (psi): 19.0 Average H (in) = 0 882 H dummy (in): 0.580

Total load kg lb B Value Test LCI 488 1076 Chang e in pore presure (psi): 6.7 LC2 505 1114 Chan§; e in external pressure (psi): 9.6 993 2190 Angel of Friction 41.5 Coeffent of external pressure 0.337 Total stress: 34.8 psi B Value 0.81 Initial pore U: ~~ 14.9 35.0 : Final U Initial effective: 19.9 -0.2 : Final effective stress

Residual Strength LabView Ru n 1 Test (a) O-ring (b) (a-b)/5.08' 762.83" Speed Speed Torque Before After Before After (rpm) (in/sec) (in-#) (in-#) (in-#) (psi) (psi) 5 2.6 958 750 827 0.65 0.41 10 5.2 1374 — 827 905 1.71 1.47 15 7.9 1432 871 972 1.76 1.44 20 10.5 1771 929 1028 __ 2.64 2.33 Run 2 5 2.6 1113 757 825 1.12 0.90 10 5.2 1261 872 910 1.22 1.10 15 7.9 1321 1029 935 0.91 1.21 20 10.5 1411 1115 988 0.93 1.32

Motion Planner Runl 5 2.6 792 623 657 0.53 0.42 10 5.2 1133 721 743 1.29 1.22 15 7.9 1171 801 839 1.16 1.04 20 10.5 1431 860 878 1.79 1.73 Run 2 5 2.6 868 ~~ 644 651 0.70 0.68 10 5.2 1041 756 777 0.89 0.83 15 7.9 1067 996 825 0.22 0.76 20 10.5 1222 1031 829 ~~ 0.60 1.23

139 UNH Ring Shear Test Summary Sheet

s ?. Torque vs. Time (Test 18)

LabView Motion Planner

20 rpm 10 rpm 10 rpm 15 rpm v\* V1 I % ^ sS4

Run 2 Run 1

30 60 90 120 150 180 210 240 270 300 330 360 390 420 450 Time (seconds)

2.UU - Potential Sur values (T 18) 1.75 -

—A— LabVIEW, o-nng before, Run 1 tfl 1.50 i Q, • LabVIEW. o-nng before, Run 2 SZ u> 1.25 -J —•— LabVIEW, o-ring after, Run 1 0c) h. # LabVIEW, o-nng after. Run 2 •* t/> 1.00 A —A— Motion Planner, o-ring before, Run 1 «3 —B— Motion Planner, o-f ing before, Run 2 x> 0.75 \ 0 Motion Planner, o-ring after, Run 1 £ —O— Motion Planner, o-f ing after. Run 2 0.50 — — Average of values • Answer 0.25 5 rpm 10 rpm 15 rpm 20 rpm

140 UNH Ring Shear Test Summary Sheet

Date: 07/30/09 Test: T19 Material: F-75 sand e max: 0.805 Wdry(gr): 1509.2 Tota 1 vol: 0.0323 ft3 e min: 0.486 Wdry(lb): 3.33 Dry UW: 103.05 pcf Dr (%) = 62.8 Void ratio: 0.605 Sample Hei§;h t Sample Sample Dummy # Dial reading reading Difference height 1 0.248 1X514 0.266 0.845 2 0.253 0.537 0.284 0.863 3 0.245 0.525 0.28 0.859 Bag pressur •e(psi): 20.1 Average H (in) = 0.856 H dummy (in): 0.579

Total load kg lb B Value Test LCI 490 1080 Chlang e in pore presure (psi): 8.1 LC2 511 1127 Change in external pressure (psi): 12.6 1001 2207 Angel of Friction 43.7 Coeffent of external pressure 0.309 Total stress: 35.1 psi B Value 0.77 Initial |oor e U: 15.0 35.8 : Final U Initial effective: 20.1 -0.7 : Final effective stress

Residual Strength Lab>/ie w Run 1 Test (a) O-ring (b) (a-b)/5.08' 762.83" Speed Speed Torque Before After Before After (rpm) (in/sec) (in-#) (in-#) (in-#) (psi) (psi) 5 2.6 856 779 ~ 0.24 10 5.2 1034 Not 888 -- 0.46 15 7.9 1500 Measured 917 ~ 1.83 20 10.5 2147 882 ~ 3.97 Run 2 5 2.6 1136 754 ~ 1.20 10 5.2 1265 Not 805 - 1.44 15 7.9 1405 Measured 784 ~ 1.94 20 10.5 _1482 872 ~ 1.91

Motion Planner Run 1 5 2.6 —718 614 0.32 10 5.2 911 Not 730 ~ 0.57 15 7.9 1299 Measured 792 ~ 1.59 20 10.5 1787 757 ~ 3.23 Run 2 5 2.6 954 589 ~" ~ 1.14 10 5.2 1080 Not 655 ~ 1.33 15 7.9 1258 Measured 718 ~ 1.69 20 10.5 _ 1319 739 — 1.82

141 UNH Ring Shear Test Summary Sheet

2000 Torque vs. Time (Test 19) 1900 20 rpm 1800

1700

1600 15 rpm/ 20 rpm 1500 15 rpm *v J / rV4 | 1400 10}rp m ? §. 1300 iii ^ fyy • c 't 1200 3 §" 1100 10 rpm 1000 5 rpm LabView 900 L/S * Motion Planner 800

700 Run 2 Run 1 600

500 —,— i—^ i——=-i— i——i —i— t i —i— ^T^= r= 0 30 60 90 120 150 180 210 240 270 300 330 360 390 420 450 Time (seconds)

2.00 -I Potential Sur values (T 19) 1.75 J

A LabVIEW, o-ring before, Run i 'w 1.50 j e, —•— LabVIEW, o-f ing before, Run 2 £ • LabVIEW, o-ring after, Run 1 1.25 \ c % LabVIEW, o-rfngafter, Run 2 a •fc*- —A— Motion Planner, o-ring before, Run 1 tfi 1.00 J ou —E3— Motion Planner, o-ring before, Run 2 55 "O O Motion Planner, o-ring after, Run 1 ifi 0.75 J Of —6)— Motion Planner, o-ring after, Run 2 — — Average of values 0.50 \ ...... Answer 0.25 J 5 rpm 10 rpm 15 rpm 20 rpm

142 UNH Ring Shear Test Summary Sheet

Date: 08/07/10 Test: T103 Material: F-75 sand e max: 0.805 Wdry(gr): 1500 Total vol: 0.0327 ft3 e min: 0.486 Wdry(lb): 3.31 Dry UW: 100.99 pcf Dr (%) = 52.5 Void ratio: 0.637 Sample Height Sample Sample Dummy # Dial reading reading Difference height 1 0189 0.466 0.277 0.857 2 0.191 0.486 0.295 0.875 3 0.195 0.487 0.292 0.872 Bag pressure (psi): 19.3 Average H (in]) = 0.868 H dummy (in): 0.580

Total load kg lb B Value Test LCI 448 988 Change in pore presure (psi): 10.3 LC2 471 1039 Charige in external pressure (psi): 12 919 2026 Ang

Residual Strength Lab View Ruii 1 Test (a) O-•ring(b) (a-b)/5.08' 762.83" Speed Speed Torque Before After Before After (rpm) (in/sec) (in-#) (in-#) (in-#) (psi) (psi) 5 2.6 625 349 431 0.86 0.61 20 10.5 935 590 448 1.08 1.53 Run 2 5 2.6 551 412 386 0.44 0.52 20 10.5 933 717 417 0.68 1.62

Motion Planner Run 1 5 2.6 902 550 631 1.10 0.85 20 10.5 1280 898 710 1.20 1.79 Run 2 5 2.6 818 655 580 0.51 0.75 20 10.5 1284 1036- 682 0.78 1.89

143 UNH Ring Shear Test Summary Sheet

1500 j Torque vs. Time (Test 103) 1400 J

1300 {

1200 J

1100 { *D c 20 rpm 20 rpm o3 1000 + a. c XAv N 900 { ^'\/ - 3 CT 800

700 5 rpm LabView Motion Planner 5 rpm 600 -f

500 Run 1 Run 2

400 60 120 180 240 300 360 420 480 Time (seconds)

2.00 i Potential Sur values (T 103) 1.75 -

*5» 1.50 - *wa* LabVIEW, o-ring before, Run 1 £ O) 1.25 LabVIEW, o-ring before, Run 2 c LabVIEW, o-ring after, Run 1 ak>_ +•* c/> 1.00 - LabVIEW, o-ring after, Run 2 flj 3 Motion Planner, o-ring before, Run 1 *o (A 0.75 - Motion Planner, o-rsng before, Run 2 Q: Motion Planner, o-ring after, Run 1 0.50 O Motion Planner, o-ring after, Run 2 —• — Average of values 0.25 - 5 rpm 20 rpm

144 UNH Ring Shear Test Summary Sheet

Date: 08/08/10 Test: T106 - Material: F-75 sand e max: 0.805 Wdry (gr): 1500 Tota 1 vol: 0.0327 ft3 e min: 0.486 Wdry(lb): 3.31 Dry UW: 101.10 pcf Dr (%) = 53.1 Void ratio: 0.636 Sample Height Sample Sample Dummy # Dial reading reading Difference height 1 0.192 0.468 0.276 0.856 2 0.194 0.490 0.296 0.876 3 , 0.198 0.485 0.287 0.867 Bag pressure (psi): 18.6 Average H (In) = 0.867 H dummy (In): 0.580

Total load kg lb B Value Test LCI 445 981 o-lange In pore presure (psi): 12.2 LC2 489 1078 Change In external pressure (psf): 9.1 934 2059 Angel of Friction 41.5 Coeffent of external pressure 0.338 Total stress: 32.8 psi B Value 1.17 Initial pore U: 15.3 32.8 • Final U Initial effective: 17.5 0.0 : Final effective stress

Residual Strength Lab View Ruin l Test (a) O-rlng (b) (a-b)/5.08 1762.83" Speed Speed Torque Before After Before After (rpm) (In/sec) (in-#) (ln-#) (in-#) (PS?) (psi) 5 2.6 640 410 - 342 0.72 0.93 20 10.5 966 570 549 1.24 1.31 Run 2 5 2.6 572 370 356 0.63 0.68 20 10.5 946 544 629 1.26 0.99

Motion Planner Run 1 5 2.6 866 600 506 0.83 1.13 20 10.5 1309 823 787 1.52 1.64 Run 2 5 2.6 798 552 522 0.77 0.86 20 10.5 1253 798™ 871 1.43 1.20

145 UNH Ring Shear Test Summary Sheet

1500 Torque vs. Time (Test 106) 1400

1300

1200

1100 20 rpm T3 C 20 rpm 3 O 1000 Q.

C - , *3» ** 1»*" *->"D 0) 900 3 c 800 * LabView 700 5 rpm Motion Planner

5 rpm 600 --*-v~-* ^*J 500 Run 1 Run 2

400 0 30 60 90 120 150 180 210 240 270 300 330 360 390 420 Time (seconds)

2.00 Potential Sur values (T 106) 1.75

' —B— Motion Planner, o-nng before. Run 2 01 0.75 O Motion Planner, o-nng after. Run 1 0.50 H —©— Motion Planner, o-nng after, Run 2 mm «•» Average of values 0.25 5 rpm 20 rpm

146