Geo Congress 2020 University of PDF

GEOTECHNICAL SPECIAL PUBLICATION NO.
321
GEO-CONGRESS 2020
UNIVERSITY OF MINNESOTA 68TH ANNUAL
Downloaded from ascelibrary.org by 130.193.199.31 on 02/29/20. Copyright ASCE. For personal use only; all rights reserved.
GEOTECHNICAL ENGINEERING CONFERENCE

SELECTED PAPERS FROM SESSIONS OF
GEO-CONGRESS 2020
February 25–28, 2020

Minneapolis, Minnesota
SPONSORED BY
The Geo-Institute of the
American Society of Civil Engineers
EDITED BY
Joseph F. Labuz, Ph.D., P.E.
Brent A. Theroux, P.E.
James P. Hambleton, Ph.D.
Roman Makhnenko, Ph.D.
Aaron S. Budge, Ph.D., P.E.
Published by the American Society of Civil Engineers
Geo-Congress 2020
Published by American Society of Civil Engineers

1801 Alexander Bell Drive
Reston, Virginia, 20191-4382
www.asce.org/publications | ascelibrary.org
Any statements expressed in these materials are those of the individual authors and do not
necessarily represent the views of ASCE, which takes no responsibility for any statement
made herein. No reference made in this publication to any specific method, product, process,
or service constitutes or implies an endorsement, recommendation, or warranty thereof by
ASCE. The materials are for general information only and do not represent a standard of
ASCE, nor are they intended as a reference in purchase specifications, contracts, regulations,
statutes, or any other legal document. ASCE makes no representation or warranty of any
kind, whether express or implied, concerning the accuracy, completeness, suitability, or
utility of any information, apparatus, product, or process discussed in this publication, and
assumes no liability therefor. The information contained in these materials should not be used
without first securing competent advice with respect to its suitability for any general or
specific application. Anyone utilizing such information assumes all liability arising from such
use, including but not limited to infringement of any patent or patents.
ASCE and American Society of Civil Engineers—Registered in U.S. Patent and Trademark
Office.
Photocopies and permissions. Permission to photocopy or reproduce material from ASCE

publications can be requested by sending an e-mail to permissions@asce.org or by locating a
title in ASCE's Civil Engineering Database (http://cedb.asce.org) or ASCE Library
(http://ascelibrary.org) and using the “Permissions” link.
Errata: Errata, if any, can be found at https://doi.org/10.1061/9780784482841
Copyright © 2020 by the American Society of Civil Engineers.

All Rights Reserved.
ISBN 978-0-7844-8284-1 (PDF)
Manufactured in the United States of America.
Geo-Congress 2020
Geo-Congress 2020 GSP 321 iii
Preface
This is the 68th offering of the University of Minnesota Annual Geotechnical
Engineering Conference, which was started in 1953, largely through the efforts of
Miles Kersten, University of Minnesota; Charles Britzius, Twin City Testing; C.K.
Preus, Minnesota Highway Department; and Wilfred Darling, Corps of Engineers. The
speakers at the 1st conference included Kersten, Britzius, and Preus, as well as Herbert
Wright, University of Minnesota; Rockwell Smith, Association of American Railroads;
and Ralph Peck, University of Illinois. In 1979, at the 27th conference, the Kersten
Lecture was established and George Sowers, Georgia Institute of Technology,
presented the first lecture.
The tradition continues through the Planning Committee, whose members represent the
contracting industry, government agencies, consulting engineers, and the University,
with a program offering technical information and discussion on recent projects for the
geoengineering community. Topics at the 68th conference cover energy
geotechnology, numerical modeling in geomechanics, lessons learned from failures,
waste on-site disposal, and case histories. The conference provides a forum to interact
with peers, meet specialty contractors, and hear researchers and practitioners discuss
theory and application of geomechanics.
This volume is one of seven containing the full collection of papers presented at Geo-
Congress 2020, numbering well of 400 in total. Each of the seven volumes were
reviewed in accordance with ASCE GSP standards. Each paper was subjected to
technical review by two or more independent peer reviewers, and publication required
concurrence by at least two peer reviewers. These publications and the conference itself
would not have been possible without the diligent effort of the individuals recognized
in the acknowledgements.
The Editors,
Joseph Labuz, Ph.D., P.E., F.ASCE, University of Minnesota

Brent Theroux, P.E., Barr Engineering
James P. Hambleton, Ph.D., M.ASCE, Northwestern University
Roman Makhnenko, Ph.D., A.M.ASCE, University of Illinois at Urbana-Champaign
Aaron S. Budge, Ph.D., P.E., M.ASCE, Minnesota State University, Mankato
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 321 iv
Acknowledgments
Geo-Congress 2020 and its published proceedings would not have been possible without the
diligent, coordinated effort of numerous people, especially the reviewers who provided detailed
feedback on papers and the session chairs who made decisions and provided editorial assistance
on selected topics. The reviewers recognized below provided assistance specifically with the
papers in this special volume. The editorial team is also profoundly grateful to Mr. Derrick
Dasenbrock (Conference Chair), Dr. Roman Hryciw (Technical Program Co-Chair), Dr. Nick
Hudyma (Technical Program Co-Chair), Dr. Nima Goudarzi, Mr. Brad Keelor, Ms. Erin Ryan,
and the ASCE publications team for their crucial roles in the production of these proceedings.
Geo-Congress 2020 Conference Program Committee
Conference Chair
 Derrick D. Dasenbrock, P.E., D.GE, F.ASCE, Minnesota Department of Transportation
Technical Program Committee
 Roman D. Hryciw, Ph.D., M.ASCE, University of Michigan

 Nick W. Hudyma, Ph.D., P.E., M.ASCE, Boise State University
 Menzer Pehlivan, Ph.D., P.E., M.ASCE, Jacobs
 Lizan N. Gilbert, P.E., M.ASCE, Guy F. Atkinson Construction, LLC
 Jean Côté, Ph.D., Université Laval Québec (Representative, Canadian Geotechnical
Society)
 Domenic D'Argenzio, P.E., M.ASCE, Mueser Rutledge Consulting Engineers
(Representative, COPRI of ASCE)
 Brent A. Theroux, P.E., Barr Engineering Co. (Representative, Minnesota Geotechnical
Society; Representative, University of Minnesota Annual Conference)
Proceedings Editors
 James P. Hambleton, Ph.D., M.ASCE, Northwestern University

 Roman Makhnenko, Ph.D., A.M.ASCE, University of Illinois at Urbana-Champaign
 Aaron S. Budge, Ph.D., P.E., M.ASCE, Minnesota State University, Mankato
Reviewers
Pouyan Asem Ivan Contreras Megan Hoppe

Chris Behling Ron Farmer Nathan Iverson
Joseph Bentler Bryan Field Joseph Labuz
Ryan Berg Steve Gale Richard Lamb
Aaron Budge Michael Haggerty Dan Mahrt
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 321 v
Steve Olson David Saftner Brent Theroux

Ryan Petersen Brian Sanchez Joseph Westphal
Greg Reuter Joel Swenson
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 321 vi
Contents
The Art of Numerical Modeling in Geomechanics .................................................................. 1
Peter A. Cundall
Lessons (Re)learned from Geotechnical Failures .................................................................. 14

Richard J. Finno
Twenty-Year Performance of a Mixed LLRW/RCRA Waste Disposal Facility................... 29

Rudolph Bonaparte, Beth A. Gross, Ranjiv Gupta, John F. Beech, Leslie M. Griffin,
and David K. Phillips
Gas Explosion Analysis ........................................................................................................... 57

Otto D. L. Strack
The Use of Steel Sheet Pile as Permanent Building Foundation Walls: Lessons
Learned over 15 Years of Design in Minneapolis, Minnesota, USA ..................................... 68
Chad A. Underwood
Quo Vadis? Inakeyaa! Inferring Flow Directions Using a Bayesian Approach ................... 78
Randal J. Barnes and Richard Soule
Supporting a Bridge between Countries Case Study: Construction of Baudette

Bridge Drilled Shafts ............................................................................................................... 87
Nathan W. Iverson
Computational Tools for the Analysis of Stability of Embankments in

Frictional-Cohesive Soils....................................................................................................... 100
Carlos Carranza-Torres and David Saftner
Eisenhower Bridge North Abutment and Approach Settlement: A Case History

of Timber Pile Downdrag and Comparative Downdrag Effect on Steel Piles .................... 119
Steven J. Olson
Washington Park Reservoir Improvements: Accommodating Ancient Landslide

Movement with a Compressible Inclusion ........................................................................... 128
Thomas Westover, Dan Hogan, Gerry Heslin, and Andrew Kost
A Retrospective on the Evolution of Geotechnical Sensing and Instrumentation

for Monitoring at MnDOT.................................................................................................... 142
Joel N. Swenson and Derrick D. Dasenbrock
Experimental Study of Forces Induced in Mechanical Excavation of Rock ....................... 156

John Pultorak, Dmitry Drozdov, Jia-Liang Le, and Emmanuel Detournay
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 321 vii
An Overview of Performance Monitoring for Drilled Full Displacement Type

Rigid Inclusions under Highway Embankments ................................................................. 163
Liang Chern Chow, Joseph G. Bentler, Alex Potter-Weight, and Andrew J. Eller
Detecting Pile Length of Sign Structures and High Mast Poles .......................................... 174
Daniel V. Kennedy, Bojan B. Guzina, and Joseph F. Labuz
Sky Harbor Airport Runway Realignment .......................................................................... 187

Hector D. Flores and Brandon J. Twedt
A Failure Mechanism around Axially Loaded Sockets in Weak Rock ............................... 198
Pouyan Asem and Joseph F. Labuz
Kennedy Bridge Instrumentation: A Pier Review ............................................................... 210

James C. Bennett
A Review of LRFD Bridge Foundation Design and Construction in South Dakota........... 221
Brett E. Belzer, Bret N. Lingwall, and Lance A. Roberts
On Solid Ground: Preventative and Responsive Geotechnical and Structural

Mitigation of Geologic Hazards Impacting Oil and Gas Production .................................. 232
Charles D. Hubbard
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 321 1
The Art of Numerical Modeling in Geomechanics

Peter A. Cundall1
1
Minneapolis, MN, USA. E-mail: pacundall@aol.com
ABSTRACT
Numerical modeling is used extensively in the design and evaluation of projects that involve
Downloaded from ascelibrary.org by University College London on 02/29/20. Copyright ASCE. For personal use only; all rights reserved.
the behavior of rock, soil, and groundwater. There are many potential pitfalls for the modeler,
particularly if inexperienced, some of which are discussed. The importance of understanding
mechanisms is stressed, and this point is illustrated with examples that show the advantages of
building simplified models in addition to detailed models for site analysis.
INTRODUCTION
Numerical modeling is used widely to understand and predict the behavior of soil, rock and
coupled fluids in the fields of mining, geothermal, petroleum and civil engineering. Typical
methods are finite elements, finite volume, boundary elements and discrete elements.
Although these calculation schemes are based on mathematical formulations, and thus may
be expected to be subject to scientific methodology, there is an art involved in the effective
application of numerical models in engineering.
The author was involved in helping users of Itasca codes (FLAC, UDEC, PFC, etc.) for many
years and has seen many incorrect usages and inappropriate approaches to modeling, as well as
opportunities that have been missed for understanding mechanisms that characterize the behavior
of rock, soil and coupled fluid in geomechanical applications. Drawing on these experiences, an
attempt is made to formulate some suggestions for successful and informative modeling.
UNDERSTANDING MECHANISMS
We believe that the most important aspect of any engineering project is to understand the
dominant mechanisms that determine the system’s behavior. If a model is too complicated, the
mechanisms may not be obvious, so it is worth constructing simple models first, driven by the
questions that the models are supposed to answer. This approach was recommended by Cundall
and Starfield (1988) and reinforced by Detournay et al. (1993) and Hammah and Curran (2009).
As Albert Einstein said, “Everything should be made as simple as possible, but not simpler.” A
hierarchy of models of increasing complexity is also useful for showing how each added feature
contributes to the behavior. Two examples are presented that use simple models to explain
mechanisms that occur in complex models.
Displacements arising from fluid impoundment behind an arch dam: Consider the
question of water seepage from the filling of a reservoir behind a large arch dam. The question
we wish to answer is: Will the river banks experience convergence, and over what time scales
will they occur? Most large dams are constructed in complex geological environments with
irregular topography for the dam placement and reservoir. A model with all this complexity in
place is not only time-consuming to run, but the dominant mechanisms are likely to be obscured.
The simplest model is two-dimensional elastic, showing how water emplacement in an idealized
valley leads to short- and long-term displacements. Figure 1 shows an upstream cross-section of
an idealized valley with impounded water, where the colors denote layers of different rock types.
In the short term (without fluid migration into the rock), the mechanical response of the saturated
© ASCE
Geo-Congress 2020
conduct fluid. By injecting fluid into a “cluster” of nodes, a hydraulic fracture may propagate in
a direction that is determined by the local stress tensor (which is related to the forces carried by
the lattice springs).
Figure 3. Incremental displacement vectors and contours of horizontal displacement, due

only to fluid flow. The maximum outward displacement is 6.2 mm.
Figure 4. Contours of excess pore pressure at steady-state fluid conditions.

The simulation consists of propagating a hydraulic fracture from one injection point, and
then—some time later—propagating a second fracture from an injection point at some distance
from the first. Figures 5, 6 and 7 show the two fractures that propagate from injection points that
are progressively farther apart, where the green dots correspond to the nodes of the initial
fracture and the blue dots to those of the second fracture. The view is of a cross-section of a
three-dimensional model, where the minimum principal stress is in the vertical direction and the
intermediate principal stress is in the horizontal direction. The simulations show that the second
fracture curves towards the first if the injection points are closely spaced, but away from the first
if the spacing is larger.
Figure 5. Two injection points close together. Green dots: initial fracture; Blue dots: second
fracture.
© ASCE
Geo-Congress 2020
upwards for large d, in accordance with the observations in the more complex lattice model with
XSite.
Figure 9. Angle of vertical principal stress at the tip of the upper crack of Figure 8 versus
d, for various lengths, L, of the upper crack.
DIAGNOSTICS
All models are simplifications of reality, even if they appear to capture all the essential
elements of a physical system. Modelers should probe their models by imposing changes in
conditions and assumptions with the aim of letting the models themselves inform us of their
limitations. As an example, we return to the river valley considered earlier. The excess pore
pressures shown in Figure 4 are for conditions of zero pressure (no resistance to flow) at the
lateral boundaries. At the opposite extreme, the lateral boundaries may be regarded as
impermeable. When this condition is substituted in the model, the results are quite different, even
if the boundaries are quite remote from the river valley. Clearly, the real conditions in the far
field must lie between these two extremes, but we learn a couple of valuable lessons from this
test. First, the fluid boundaries are important to the response, even though they are not part of the
river and valley system. Second, we realize that we need more field data to characterize
conditions remote from the valley, such as geologic features that act to restrict or channel water
flow.
MISCONCEPTIONS, MISTAKES AND MISUNDERSTANDINGS

Boundary conditions: A common mistake is to place boundaries too close to the region
being investigated. A simulation of slope failure could show active plastic flow at one of the
model boundaries (not a symmetry line). This is clearly unacceptable, because the boundaries in
© ASCE
Geo-Congress 2020
such a model are artificial and must not be part of the failure mechanism. Apart from increasing
the boundary distance, another test for erroneous boundary proximity is to change the nature of
the boundary condition (e.g., from a roller boundary to a fixed boundary, or from fixed
displacement to stress boundary) and observe if the mechanism changes.
2D versus 3D: In many cases, a 2D model is a satisfactory approximation to a 3D system.
However, sometimes the nature of the solution is qualitatively different in 2D compared to 3D.
For example, it might be imagined that a point load applied to the free surface of a uniform
elastic half-space could be approximated by a model with a lower boundary that is some distance
away from the surface, and that the displacement response would converge to a fixed value as
the distance to the boundary is increased. However, for a line load in 2D, the surface
displacement continues to increase as the boundary distance is increased. This is not true if a 3D
model is made: the surface displacement converges to a fixed value as the boundary distance is
increased. Similar considerations apply to fluid flow in 2D and 3D.
Loading rate: Software that uses a time-marching solution scheme requires special care to
ensure that the simulation correctly represents the intended physics. For example, if a laboratory
test on a sample is being modeled, the slow loading rate employed in the laboratory is often
impractical to impose on the numerical sample because of the excessive computer time needed.
The loading rate may be increased as long as inertial forces are small compared with quasi-static
forces arising from the sample stresses. By doing a series of tests with different loading rates, the
maximum loading rate may be found that is consistent with results from simulations done at
lower rates.
The preceding remarks assume that the loading rate is imposed as a velocity condition on the
system being modeled and can be changed by the modeler. However, in some cases, stresses may
be applied, or the “loading” consists of elements that are carrying stress being removed from the
model. In the latter case, dynamic excitation of the system may be minimized by substituting an
applied stress for the element to be removed, and then gradually reducing this stress. As before,
the effect of the procedure may be tested by varying the size of the increments of stress and
reducing the increments if dynamic effects are noted.
Some systems are physically unstable; reducing the loading rates has no effect on the
response because the system determines the path that is followed. This must be accepted as the
way the system behaves. In such cases, an explicit, time-marching solution is better than an
implicit, static solution method, because the former follows the response in a physically realistic
way by accounting for accelerations correctly.
Time-dependence: Time dependence in rock and soil systems is widely misunderstood. A
site engineer may observe changes in displacement over time, and it is tempting to try to
“explain” these with a numerical model (particularly one that employs a time-marching solution
scheme, such as FLAC or UDEC) that also exhibits non-steady behavior. There are many causes
of time-dependent behavior, such as material creep, unsteady fluid flow and progressive
fracturing. If none of these is included explicitly in a model, then ongoing movement observed
numerically may simply reflect the fact that the model is still in the process of converging to
equilibrium. The term “plastic flow” often leads to confusion, because there is no time associated
with the standard formulations of plasticity; specifically, there are no time derivatives in the
equations. Plastic flow cannot be invoked to explain the ongoing “flow” of material observed
at—say—a rock slope in the field. An elastic-plastic solution of the slope either converges to
equilibrium or it indicates acceleration if the slope is unstable. Neither case corresponds to time-
dependent observations on site, which may be due to the factors mentioned earlier.
© ASCE
Geo-Congress 2020
When fluid flow is combined with plasticity, there can be confusion about the role each plays
in apparent time-dependence. At a particular stage in a simulation, the displacements (for
example) may be changing. To confirm whether these indicate lack of mechanical convergence
or the effects of fluid migration, a short trial probe may be made in which fluid flow is switched
off. If the displacement quickly becomes constant, then the solution is converging correctly, and
the observed time-dependence is solely due to changing pore pressures. If the displacement
continues to change, then the solution is not in mechanical equilibrium at each time in the
evolution of the fluid field.
Localization: Any system that embodies an elastic-plastic material can exhibit spontaneous
localizations, such as shear bands. Even a perfectly plastic, non-associated material will localize,
and the tendency is enhanced if the material can soften with strain. The challenge for the modeler
is to judge to what extent the numerical results represent reality. A good analysis of the
conditions that lead to shear bands and their characteristics (spacing and length) is given by
Poliakov et al. (1994). One of the difficulties of modeling localization with continuum codes is
that there is generally a dependence of results on grid size. Physically, in any material there is an
internal length scale, such as grain size, that will limit the thickness of the shear band. In a
conventional continuum model, the only length scale is the element size, so the thickness of a
shear band collapses to some multiple of this size. If a strain-softening material is used, the
softening rate should be scaled in order to affect an approximate independence of element size
on the results (see Crook et al. 2003 for a discussion on this topic).
JOB TITLE : (*10^3)
FLAC (Version 7.00)
LEGEND
3.500
12-Jun-15 10 33
step 20842
-2.778E+02 <x< 5.278E+03
-1.028E+03 <y< 4 528E+03
Boundary plot 2.500
0 1E 3
Max. shear strain-rate
0.00E+00
2.50E-06 1.500
5.00E-06
7.50E-06
1.00E-05
1.25E-05
1.50E-05
0.500
1.75E-05
2.00E-05
Contour interval= 2.50E-06
-0.500
0.500 1.500 2.500 3.500 4.500

(*10^3)
Figure 10. Contours of shear strain rate at failure: the primary localization consists of a
circular slip surface. Secondary localizations at the base of the slope correspond to a
toppling mechanism.
As an example of the effect of localizations on slope stability, consider a 45° slope
constructed in an elastic-plastic material with constant friction angle and cohesion. With an
element size in FLAC of 1/100 of the slope height and zero dilation angle, the slope fails with a
© ASCE
Geo-Congress 2020
characteristic circular slip surface: see Figure 10. Other localizations near the toe are resolved
with the fine discretization. These are vertical shear bands that correspond to toppling failure
(reverse slip). Adding a 20° dilation angle to the material inhibits these secondary shear bands,
but not the primary one: see Figure 11. The factor of safety of the slope is decreased slightly
when the secondary shear bands are active. In real slopes, the stability may be decreased if such
toe mechanisms are possible, but they may be inhibited by roughness on potential slip planes.
The numerical model has provided us with the knowledge of what to look for in the field.
JOB TITLE : (*10^3)
FLAC (Version 7.00)
LEGEND
3.500
12-Jun-15 10:40
step 23102
-2.778E+02 <x< 5.278E+03
-1 028E+03 <y< 4.528E+03
Boundary plot 2.500
0 1E 3
Max. shear strain-rate
0.00E+00
5.00E-07 1.500
1.00E-06
1.50E-06
2.00E-06
2.50E-06
3.00E-06
0.500
3.50E-06
4.00E-06
4.50E-06
Contour interval= 5.00E-07

-0.500
0.500 1.500 2.500 3.500 4.500

(*10^3)
Figure 11. Conditions as Figure 10 but with dilation angle of 20°.

Fluid time scales and approximations: The former sections showed how the behavior of a
complex system can often be understood by exercising a simple numerical model. A similar line
of thought can be followed by using the results of simple analytical solutions to provide guidance
on the expected behavior of more realistic models.
The advantage of an analytical solution is that it gives a relationship between model variables
that may not be trivial or intuitive. The superiority of the simple numerical model approach is
that it allows a degree of sophistication (or complexity) that cannot be matched by the analytical
method. We consider the topic of fluid-mechanical interaction to show how the use of simple
analytical relation can support the modeling of more complex problems.
The prediction of soil and rock behavior under a natural or engineered change is often
challenging, especially when the material pores or voids are filled with fluid. There are several
levels of possible interaction between the solid matrix and the fluid that add richness, but also
complexity to the problem. For example, fluid pressure may increase in the short time after
application of a load on a rock mass; also, fluid pressure in the pores may induce yielding in a
plastic rock. But what is possibly the most interesting feature is that the rock or soil can display a
delayed (or expanded) reaction in time if fluid pressure can change or diffuse in the medium. The
increase or decrease of pore pressure in the porous space creates an expansion or contraction in
the medium, much like the one that would be produced by a temperature change. Manifestations
© ASCE
Geo-Congress 2020
of this mechanism are the delayed settlement typically observed after the application of a load on
a saturated soil mass and further dissipation of excess pore pressure (in a consolidation process),
or associated reduction in reservoir fluid pressure after oil extraction from a production well.
The time it takes for a mechanism to develop is relevant to the geomechanical engineer and
the modeler alike. The quantity can be approximated based on the expression of characteristic
time derived from analytical source solutions (Theis 1935). The fluid time scale for the diffusion
process is given by the relation tc  L2 / c , where L is the average length of the flow path
through the medium, and c is the diffusivity, defined as the ratio of hydraulic conductivity, k H
(in units of length over time), and specific storage coefficient, S s (units of 1/length). For a
confined or fully saturated medium, S s is the elastic specific storage coefficient, which has
contributions from the fluid and the solid matrix compressibility:
 1 2 
Ss   f g    (1)
 M K  4 / 3G 
where  f is fluid density, g is gravity acceleration, M is Biot modulus,  is Biot coefficient,
and K and G are material bulk and shear modulus, respectively. For a phreatic aquifer, in
addition to elastic specific storage, there is an often dominant contribution that comes from the
drainage of pores:
 1 2  n
Ss   f g    (2)
 M K  4 / 3G  Lp
where n is effective porosity and L p is average thickness of the aquifer. Typically, the phreatic
storage is two to four orders of magnitudes larger than the elastic storage.
Figure 12. Elastic block being pushed with a constant end velocity, with a frictional
interface between the block and the elastic base.
The evaluation, even approximate, of fluid time scale from material properties gives valuable
information on how long it would take for the model to reach steady-state flow (or for a fluid
pressure ‘front’ to travel a specified distance, L ). If the required time scale of the analysis, ta , is
very small compared to tc , the influence of fluid flow on the simulation results will probably be
negligible, and an undrained simulation can be performed. No real time will be involved in the
numerical simulation, but the pore pressure will change due to volumetric straining. The short-
term simulation of a footing load is an example of this approach. On the other hand, if ta is
© ASCE
Geo-Congress 2020
much larger than tc , then steady-state conditions prevail at t  ta and the pore pressure
calculation can, for most engineering purposes, be uncoupled from the mechanical computation
at a large saving of computational time.
Also, the simple analytical relationship can be used to approximate the distance, L  ct that
the pressure ‘front’ has travelled in a given amount of time. This information helps with the
assessment of maximum grid extent (to avoid boundary effects), explicit time step magnitude
and general numerical simulation feasibility.
*****
*** **************************
************************ **********************
********************
Figure 13. Snapshot of velocity vectors at a particular point in time, corresponding to slip-
stick waves traveling along the block. Black crosses denote slipping segments.
Figure 14. Depiction of instantaneous normal stresses (red) and shear stresses (green) on
the frictional interface, corresponding to the state shown in Figure 11.
The evaluation of time scales can help devise computationally efficient numerical schemes.
We take the example of transient phreatic surface evolution. The calculation time of transient
phreatic surface can be excessive. It is possible, though, to speed up the transient calculation,
which can be accomplished by first recognizing that two distinct time scales are involved. For
given boundary conditions (including a given saturation distribution), there is a characteristic
time for the flow field to adjust to an imposed disturbance (the estimate is found using the elastic
specific storage, equation (1)). For a fluid modulus of 2 GPa, this time scale is usually quite
rapid—of the order of minutes. The second time scale involves the bulk transport of fluid, when
the saturation of a region is changing (see specific storage coefficient estimate in Equation (2)).
This process is related to the actual elapsed time (the volume transport is proportional to flow
rate and time). Typical time scales to alter phreatic surfaces are of the order of weeks. Advantage
© ASCE
Geo-Congress 2020
can be taken of these different time scales to speed a FLAC transient analysis by making a
calculation that alternates between fast and slow processes in stages (see Itasca 2011).
Chaos and instability: Many systems exhibit chaotic behavior, in which the result depends
on trivial details of the initial conditions. This is not a limitation of numerical models. Chaos is
an inherent response of many nonlinear physical systems, and the model simply exhibits similar
behavior. An example of a chaotic system is a model of a jointed rock mass in which several
joints meet at the same point in space, such as two orthogonal, continuous joint sets. Joints of
either set can slip at an intersection, but when one slips, then the other set is inhibited from
slipping because of the offset induced by the first set. This is classical chaotic behavior, in which
a tiny geometrical perturbation causes a large change in system response (i.e., slip on one joint
set rather than the other). Such behavior happens physically, and the numerical model responds
in a similar way. This is not an “error” as some modelers may believe.
1.000
0.900
0.800
0.700
0.600
0.500
0.400
0.300
0.200
0.100
10 20 30 40 50 60
-03
(10 )
Figure 15. Time history of normalized driving force ( Fs / W tan  ) on sliding block. Note
that the mean value is below the force required for slip initiation.
Some systems may be physically unstable. It is essential to recognize this and ensure that the
numerical model can handle it without numerical instability. Time-marching solution schemes
can do this. As an example of an unexpected instability, consider an elastic block being pushed at
constant speed on a base block, with a perfect frictional (non-softening) interface between them
(see Figure 12). We might expect the behavior to be trivial, namely that the upper block would
start steady sliding when the driving force equals its weight multiplied by the friction coefficient.
However, under certain conditions, slip-stick waves are seen to propagate spontaneously within
the upper block, as illustrated in Figure 13, which shows a snapshot of velocity vectors at a
particular point in time. The source of the instability is that small spatial variations in normal
stress lead to either a stuck segment (slightly higher stress) or a slipping segment (slightly lower
stress). These variations are then amplified and reinforced by waves that embody large-scale
variations in shear and normal stresses that propagate along the block (see Figure 14). The
© ASCE
Geo-Congress 2020
driving force (i.e., the reaction force on the left-hand end of the top block) exhibits oscillatory
fluctuations (see Figure 15), but—interestingly—its mean value is below the static friction force
of W tan  where W is the weight of the top block and  is the friction angle of the interface.
The system exhibits strain-rate weakening (see Figure 16) even though the interface friction is
constant. It is not known if this behavior occurs in—say—a geological setting, but it is an
example of new knowledge being gained by exercising a numerical model.
Figure 16. Mean normalized driving force versus driving velocity for three values of
interface friction angle.
CONCLUSIONS
Some common misconceptions and topics for misunderstandings that occur in the application
of numerical modeling in geomechanics have been discussed, and suggestions made for good
modeling practice, including the building of simplified models to understand mechanisms. A
numerical model may also provide new knowledge of system behavior even though data is
lacking for a detailed analysis.
ACKNOWLEDGEMENTS
Itasca colleagues are thanked for their support and help over many years. They are
exceptional engineers, dedicated to ensuring that correct physics is embodied in numerical
models.
REFERENCES
Crook, T., Willson, S., Yu, J.G., and Owen, R. (2003) “Computational modelling of the localized
deformation associated with borehole breakout in quasi-brittle materials.” Journal of
© ASCE
Geo-Congress 2020
Petroleum Science and Engineering, 38, 177-186. Elsevier.

Cundall, P.A. (2011) “Lattice method for brittle, jointed rock.” Continuum and Distinct Element
Numerical Modeling in Geomechanics, Proceedings of the 2nd International FLAC/DEM
Symposium, February 2011, Melbourne, Australia. pp. 11-19. Minneapolis: Itasca.
Detournay, E., Pearson, J.R.A., and Thiercelin, M. (1993) “Modelling Rock Mechanical
Processes in Petroleum Exploration and Production.” ISRM News Journal, 1(2) 17-20.
Hammah, R.E., and Curran, J.H. (2009) “It is Better to be Approximately Right than Precisely
Wrong: Why Simple Models Work in Mining Geomechanics.” ARMA 09, Proceedings 43rd
US Rock Mechanics Symposium and 4th US-Canada Rock Mechanics Symposium, Asheville,
NC, June 28–July 1.
Itasca Consulting Group, Inc. (2011) FLAC – Fast Lagrangian Analysis of Continua Version,
Ver. 7.0. Minneapolis: Itasca.
Itasca Consulting Group, Inc. (1019) XSite Ver 3.0. Minneapolis: Itasca.
Poliakov, A.B., Herrmann, H.J., Podladchikov, Y.Y., and Roux, S. (1994) “Fractal Plastic Shear
Bands.” Fractals, 2(4), 567-581.
Starfield, A.M., and Cundall, P.A. (1988) “Towards a methodology for rock mechanics
modeling.” Int. J. Rock Mech. & Min. Sci. 25(3), 99-106.
Theis, C.V. (1935) “The relation between the lowering of the piezometric surface and the rate
and duration of discharge of a well using groundwater storage.”, Am. Geophys. Union Trans.,
16, 519-524.
© ASCE
Geo-Congress 2020
Lessons (Re)learned from Geotechnical Failures

Richard J. Finno, P.E., Ph.D., D.GE, M.ASCE1
1
Professor Emeritus, Dept. of Civil and Environmental Engineering, Northwestern Univ.,
Evanston, IL. E-mail: r-finno@northwestern.edu
ABSTRACT
Despite the advances in knowledge in geotechnical engineering over the last decades, failures
during construction of structures still occur. This paper describes four such failures related to
excavation support, and discusses their causes, from both technical and non-technical
standpoints. The failures described include both limit and serviceability state failures. One can
expand the notion of redundancy to include non-technical issues and it is in this context that the
failures are discussed. The case histories show that if redundancy had been properly employed
during design and construction, the failures probably would not have occurred.
INTRODUCTION
Despite the advances in knowledge in geotechnical engineering over the last decades, failures
during construction of structures still occur. It is not the lack of knowledge that creates these
situations, but rather the inappropriate application of it, or in some cases, a lack of
communication among vested parties. This problem has been long recognized in geotechnical
engineering. Indeed, Osterberg (1989) discussed this subject in the twenty-first Terzaghi lecture.
Therein, he defined redundancy as “designing, incorporating, and including physical and human
processes into analysis, design, and construction in such a way that if one element, whether
physical or human, fails to function or fails to function in the way intended, other elements take
over in such a way that the structure will still function essentially as intended.” He categorized
the various phases of work geotechnical engineers normally perform in a project and, by the use
of eight case studies, pointed out how redundancy should be used for each. More than 30 years
have passed, and failures still occur. It seems useful to revisit Osterberg’s methodology and
apply it to more recent failures.
This paper will focus several aspects of redundancy as defined by Osterberg: redundancy in
analysis, design, peer review and people problems. These concepts are illustrated herein by four
case studies that include failures of a self-sinking caisson, a cut slope and retention system and
two retention systems. Three of the failures are limit state failures while one of the retention
systems was a serviceability failure.
SELF-SINKING CAISSON
The reinforced concrete cylinder, shown during construction in Figure 1a, was designed as a
self-sinking caisson. It was intended as both temporary and permanent walls for the structure. It
had an inner diameter of 136 ft and a wall thickness of 7.5 ft. It was designed to be 105 ft deep
and constructed in eight stages with concrete rings varying between 12 and 16 ft tall. The first
stage contained the more heavily reinforced steel-plated cutting shoe that formed the base of the
caisson during sinking.
Such a caisson is built by constructing individual rings on the ground and excavating the soil
inside the ring such that the caisson sinks under its own weight. As indicated in Figure 1b, this
occurs when the caisson weight exceeds the side and tip resistances as well as overcoming the
© ASCE
Geo-Congress 2020
effects of buoyancy. If a slurry-filled annulus is included in the system, the side resistance can be
restricted to that which develops along the cutting shoe and the inside surface of the concrete
ring. In this case, the annulus was not provided.
Figure 1. Self-sinking caisson

Subsurface conditions: The subsurface conditions at the site are shown in Figure 2 and
consist of fill overlying glacially-derived soils atop limestone. The surficial fill was 10 to 20 ft
thick and consisted of silty sands of variable consistency and stiff to very stiff clays.
Glaciolacustrine clays lie beneath the fill, with an upper desiccated crust which transitioned to
medium stiff clays near the bottom of the stratum. Hard glacial tills and dense outwash are
encountered beneath the clays and extend to the limestone. As noted in Figure 2, the conditions
below the medium clays varied significantly at opposite ends of the caisson. The piezometric
surface within the dense soils above the rock and within the upper portions of the limestone was
close to the ground surface.
Figure 2. Subsurface Conditions
© ASCE
Geo-Congress 2020
Design considerations: The general structural design philosophy was to compare the
stresses induced by the applied loads to the cracking moment capacity of selected cross sections
of the completed caisson. The thickness of the walls were selected primarily to achieve the
weight needed to sink the caisson. The designer checked for buckling with cracked section
properties at the bottom of the caisson using uniform lateral soil and water pressures. Thus a key
aspect of the design of a self-sinking caisson is the selection of the lateral earth pressures that act
against its side. These stresses are directly related to the frictional resistance between the soil and
the caisson, as well as the lateral pressures against the wall. The designer used at-rest pressures
based on Jaky’s formula and applied them uniformly around the caisson when checking for
buckling. An out-of-roundness criteria of 5 inches for formwork tolerance was used to compute
flexural stresses. This design approach resulted in successful sinking of caissons in the area,
albeit for smaller diameter caissons constructed in different subsurface conditions.
Construction: There were a number of challenges that the contractor faced as construction
proceeded. Among them were starting the sinking without tilting, potential bearing capacity
failure in the softer clays, and the high water pressures at the top of rock.
Figure 3. Conditions in caisson after failure

After grouting the rock, the caisson construction began by making an approximately 15 ft
deep pre-cut. Launching the caisson from this lower elevation mitigated potential problems of
obstructions in the surficial fill and helped with maintaining verticality as the caisson began to
sink. The first two stages of the caisson were constructed above ground in the pre-cut before the
contractor began excavating the soil inside the caisson to allow sinking. At first, backhoes
excavated the soil in the dry until the weight of the unexcavated soil above the water bearing
sands approached the water pressure within the sands above the top of rock. At this point, the
© ASCE
Geo-Congress 2020
caisson was flooded to preclude uplift and potential piping problems. Thereafter excavation
proceeded in the wet with clam shell buckets. The caisson was advanced steadily until the tip
reached the harder soils at 60 ft, after which progress significantly slowed.
Failure: Small cracks in the concrete first occurred in the rings exposed at the ground
surface when the caisson first encountered the hard soils at 60 ft. When 7 of the 8 intended stages
had been constructed and the tip of the caisson reached a depth of about 90 ft, damage to the
structure was observed, as summarized in Figure 3. Large circumferential delaminations were
noted about 6 inches from the inside face of the caisson at both the 6 and 12 o’clock positions;
these were as large as 11 inches wide. At about the 3 o’clock position, a large radial crack
extended across the entire wall thickness and was 3 inches wide at the exterior wall. The overall
cross section of the caisson had deformed from its original circular shape to a more elliptical
shape. The top of the caisson was 26 inches too narrow between the 12 and 6 o’clock position
and 21 inches too wide between the 3 and 9 o’clock position. The out-of-roundness extended to
the bottom of the caisson and increased about 6 inches in each direction. The depression close to
the caisson shown in the figure was a surface reflection of one of several localized soils failures
that developed during attempts to sink the caisson through the hard soils. After the cracks and
movements occurred, the caisson was partially backfilled to stabilize it.
So why did this happen? Looking at the different soil conditions on each side of the caisson
in Figure 2, it is clear that the at-rest pressures were different in the harder soils at depth because
the glacial till was heavily overconsolidated in contrast to the outwash which was not overlain by
glacial ice. For the caisson to be subjected to uniform pressures, it would have to behave as a
flexible ring that would deform until the lateral pressures became uniform, as illustrated in
Figure 4.
Figure 4. Relative Stiffness of Wished-in-place Caisson

This concept was elucidated by Peck (1969) when discussing tunnel liners. In contrast to a
flexible ring, a rigid ring would not deform and thus would be subjected to variable stresses that
would exist in situ, and thus lead to significant bending moments in the structure. Once one
realizes that non-uniform lateral pressures will act against the caisson, the only way to design for
uniform pressures is to implicitly assume that the caisson is flexible. Clearly the caisson was
© ASCE
Geo-Congress 2020
neither rigid nor flexible, but somewhere in between, and thus should have been designed for
non-uniform lateral stresses.
The non-uniformity in loading caused by the variable soil conditions at depth was
exacerbated by non-uniformities that develop during sinking. When one considers that the
allowable tilt of the caisson during sinking was 1°, a tight restriction in itself, additional
variations in lateral stresses would be expected. Once the caisson extended into the harder soils,
and because no annulus was maintained, a 1° tilt would imply that the caisson would push into
the soil as much as 1 ft in some locations and generate increases in lateral soil pressure and move
away from it in others to cause corresponding decreases in pressure. Furthermore, development

of the failure zones in the soil adjacent to the wall in some locations would cause further non-
uniformities in lateral pressures. The structural damage shown in Figure 3 was the result of a
design that did not account for in situ and construction-induced lateral pressures. No shear
reinforcing was included at all.
How could have this failure been prevented? There was no peer review of the design and thus
no chance for one to change the incorrect conceptual model of the shaft. When the cracks first
appeared in the concrete at the ground surface, no review of the design was undertaken, rather
the cracks were thought to be caused by shrinkage. Presumably, a peer review for the original
design could have identified the errors and change the loadings to non-uniform radial pressures
that reflected the in situ and expected variations caused by construction. There was an over-
reliance on precedent based on smaller diameter shafts sunken mainly through softer clays, like
the upper 60 ft of the soil profile. Using precedence as a basis of design for conditions which are
materially different from those on which the precedence was based is a problem that still plagues
overconfident designers.
Figure 5. Conditions during construction
SLOPE AND RETENTION SYSTEM FAILURE

A failure of an earthen embankment supporting a rail line occurred when approximately 200
feet of the slope and an adjacent earth retention structure failed during ongoing construction
activities. This work included construction of an internally-braced excavation and an open-cut to
provide a tie-in for conduits to an existing facility. A photo of conditions prior to failure at the
© ASCE
Geo-Congress 2020
location of the failure are shown in Figure 5.

Subsurface conditions: Figure 6 shows the subsurface conditions near the location shown in
the photo. The surficial layer consists of a very stiff to stiff clay that gets weaker with depth.
Below this layer, four distinct strata are encountered that become progressively stronger with
depth. All these clay layers are glacial in origin and consist primarily of clays of low to medium
plasticity. The hard clay stratum is a basal till deposited directly from beneath the ice. The
bottom of the open cut and internally braced excavation was located at elevation -8 ft.
Figure 6. Subsurface Conditions
Figure 7. Cross Section at Open Cut
© ASCE
Geo-Congress 2020
Design considerations: The earth retention system was designed for an apparent earth
pressure diagram with a uniform pressures and a surcharge of 546 psf. Because of the variation
in slope behind parts of the wall and different depths of excavation, different wall sections were
designed for between 27H to 29H of lateral pressure, including surcharge. Slope stability
calculations were made during design for the cut slope shown on the right side of Figure 7, and
acceptable margins of safety were computed.
Failure: A failure occurred and photos taken soon afterwards are shown in Figure 8. The
photo in (8a) of the top of the slide area was taken before any regrading was done in the area.
The failure mass moved laterally towards the open cut, in the direction of the red arrow, resulting
in a graben at the back of the slide. Beneath the suspended track, the top of whitish ballast
material with stains from the railroad ties is visible on the mass of soil that fell from beneath the
rails after the slope failed and the mass translated towards the cut.
Figure 8. Failure of Open Cut and Retention System

The sliding mass impacted the retention system with the result shown in Figure 8b. Because
of the observed failures of both the open cut excavation and the retention system, a structural
review of the retention system was carried out to evaluate its role in the failure. The analyses
examined its capacity to support the estimated loads on the structure just prior to the collapse.
The results of the study indicated that the as-constructed retention system was somewhat under-
designed, but was stable under the conditions which existed immediately prior to the failure, and
would have remained so without additional demands being placed upon it when the adjacent
slope became unstable.
The cause of the embankment failure was a translational type slope stability failure of the
open-cut excavation. The failure surface largely developed near the bottom of the medium clay
stratum. The computed factor of safety for conditions within the slope at the time of the collapse
using the strengths shown in Figure 6 is slightly less than one, indicating that failure of the slope
was likely. The sliding soil mass created by the slope failure overwhelmed the earth retention
system causing it to partially collapse.
But why did this happen? Design of retention system and slope stability analyses are well-
established methodologies in geotechnical engineering. The geotechnical engineer provided
slope excavation recommendations in its geotechnical report that did not take the loading of the
embankment into consideration. These conditions were appropriate for the slope shown in the
right side of Figure 7. However, the location of the conduit was moved much closer to the
embankment during the latter stages of the design process, and thus the left hand side of Figure 7
© ASCE
Geo-Congress 2020
was applicable to parts of the open cut. The location of the cut was moved without informing the
geotechnical engineer responsible for the cut slope design. The recommendations for the slopes
in Figure 7 were subsequently replicated, without validation, in the structural drawing set even
though the location of the new conduit and retention system were moved putting the excavation
within the influence of the embankment. Engineers working for the contractor tasked with
demonstrating to the owner that the proposed excavations were stable to support the
embankment failed to include the embankment in their initial submittals. When duly noted by the
owner’s reviewers, the initial calculations were retracted and the original open-cut design was
cited as a demonstration of adequacy. Despite acknowledging the lack of supporting calculations,

the owner permitted the open-cut excavation to proceed. The peer review process failed, and the
factors leading to the slope failure involved lapses in oversight and communication processes.
SERVICEABLITY FAILURE OF RETENTION SYSTEM

A reinforced concrete parking structure with above and below grade parking was constructed
across an alley from a row of brick and masonry structures. The 20-ft-deep excavation for its
basement was supported by a soldier pile and lagging wall supported by one row of ground
anchors, as shown in the photo in Figure 9a. The parking structure was supported on auger-cast
piles constructed at excavated grade. Hairline to 1-inch-wide vertical and diagonal cracks
developed in the interior masonry walls in the adjacent buildings during construction, as
illustrated in Figure 9b. The damage extended about 40 ft from the face of the buildings adjacent
to the cut. While no optical survey or inclinometer data were collected during excavation,
vibration measurements were obtained during soldier pile installation. Optical surveys were
initiated after damage to the buildings occurred.
Figure 9. Excavation Support and Typical Damage in Adjacent Building

Subsurface conditions: The subsurface conditions and the relative locations of the soldier
pile wall and buildings are shown in Figure 10. The stratigraphy consisted of a surficial layer of
poorly graded sand fill with occasional construction debris underlain by very loose to medium
dense alluvial sand. Standard penetration test N values varied from 2 to 30 blows/ft with the
higher values occurring at greater depths. CPT tip resistances as low as 10 ksf were encountered
in probes located near the soldier pile wall. The alluvial sands extended to 75 ft below grade, the
© ASCE
Geo-Congress 2020
maximum depth explored in site investigations. Ground water was located several ft below
excavated grade. The buildings were about 20 ft from edge of the excavation. The ground
anchors extended beneath wall footings. The auger cast piles extended into the dense sand.
Figure 10. Subsurface conditions

Construction history: The construction of the foundations for the parking facility began by
driving the HP soldier piles, excavating and installing lagging until the excavated grade had
reached several ft below the level of the tiebacks, installing and stressing the ground anchors,
excavating and installing lagging for the full depth of the excavation and then constructing the
auger cast piles from the excavated grade. The occupants of the adjacent buildings complained
about damage to the buildings during the time when the auger cast piles were being installed
along the column line closest to the buildings. After the damage was noted, tieback liftoff tests
were conducted and several tiebacks were found to have lost 30% of their lockoff loads.
Failure: Various construction activities could have contributed to the damage noted in the
buildings, including vibrations caused by driving the soldier piles, installing the ground anchors
beneath the buildings, stress relief caused by the excavation and installing the auger piles.
Vibration data collected in front of the buildings during pile driving indicated maximum
particle velocities between 0.1 and 0.7 inch/sec at frequencies of about 20 Hz. According to the
US Office of Surface Mines, at 20 Hz, the recommended safe limit of particle velocity is 1.25
inch/sec, indicating that it was unlikely that pile-driving induced vibrations caused any damage
to the buildings.
Excessive ground loss can occur when installing tiebacks through loose sands. But if the
tiebacks are installed properly, the movements would be negligible. In this case, the holes for the
tiebacks were made by driving casing to the required length, inserting a bar and knocking off the
drive point, and injecting grout under pressures of at least 100 psi as the casing was withdrawn.
This procedure provides full support for the sands as the tiebacks were installed, and thus
movements associated with this procedure were likely negligible.
© ASCE
Geo-Congress 2020
The magnitude and extent of ground movements associated with stress relief from the
excavation can be estimated based on performance of excavations made under similar
conditions. Clough and O’Rourke (1990) suggested that the maximum normalized settlement
(settlement divided by the depth of excavation) would be about 0.15%. For the 20 ft deep cut, the
maximum settlement is estimated in this way as 0.4 inch and would occur adjacent to the soldier
pile wall. The extent of the settlement would be limited to about twice the excavated depth, or 40
ft. The settlement at the front end of the building would be 0.2 inch and the distortion under the
building can be estimated as 1/1200. This estimated amount of distortion is small enough that
damage would be expected to be negligible, based on performance data compiled by Son and
Cording (2006).
Ground loss occurs during auger cast pile installation when one removes more soil than the
theoretical volume of the auger cast pile. The ground loss results in movements of the surround
ground that can extend significant distance from the pile. The risk of settlement is greatest when
working through loose, cohesionless sands, such as existed at this site (e.g., Van Impe 1988 and
Van Weele 1988). Thorburn et al (1993) presented data to show that the ground movements from
installing a single pile could extend to a radius of two-thirds of the depth of the pile. They
reported case studies where movements up to 1.5 inch were observed, and attributed this large
ground loss to liquefaction of the loose sands, as illustrated in Figure 11. This is particularly
exacerbated when the auger transitions into the dense soils and the rate of advance slows so that
the number of cycles that occurs in the loose sands above increase significantly.
Figure 11. Cyclic Stresses Induced by Installing Auger Cast Piles (adapted from Thorburn
et al. 1993)
Thorburn et al. (1993) also indicated that significant drops in CPT resistance were noted
within 3.3 ft of auger cast piles after installing them in loose, saturated sands. Presumably this
drop in resistance arises from a reduction the lateral stress in the soil surrounding the pile. This
© ASCE
Geo-Congress 2020
response suggests that when the auger cast piles were installed close to and to depths below the
soldier piles, the toe of the soldier pile would move into the excavation to generate the necessary
reaction to maintain stability of the wall. This phenomenon would contribute to movements
behind the wall, as well as cause a drop in tieback load, as was seen in several of the lift-off tests.
The sands at the site respond to changes in stress with very little time lag, there is essentially
no hydrodynamic effect because of the permeable nature of the sands, and sands are not very
creep susceptible. So the soils are the site are essentially material rate-independent, and thus one
would expect little time delay between the causal event and the resultant damage. Hence, given
the other possible causes are unlikely to have contributed to the damage, it is likely movements
arising during installation of the auger cast piles caused the damage to the structure because they
were being installed when the complaints arose.
So how did this situation develop? It was apparent by the lack of ground or building
instrumentation installed prior to construction that the designer of the parking structure did not
consider movements during construction to be an issue. Vibrations were monitored as the HP
sections were driven, suggesting that the structural engineer was at least concerned about the
effects of vibration. There was a lack of redundancy in the analysis and design of the excavation
support and foundations. There was no apparent communication between the geotechnical
engineer, the excavation support subcontractor and the auger cast subcontractor. There was no
peer review for this project which could have identified the potential problems with installing
auger cast piles in loose, saturated sand.
Figure 12. Collapse of Type M3 excavation adjacent to Nichol Highway
RETENTION SYSTEM FAILURE

The failure of the retention system at the Nichol Highway has been well documented by the
Committee of Inquiry (COI 2005). In April 2004, a collapse (Figure 12) occurred at part of the
excavation site known as “Type M3”, which was directly adjacent to the Nicoll Highway. The
cut-and-cover excavation at that location was 109 ft deep and 66 ft wide. The lateral support
system included 2.6 ft thick diaphragm walls that extended through deep layers of soft estuarine
© ASCE
Geo-Congress 2020
and marine clays and were keyed (mostly) into the underlying dense old alluvium. The walls
were laterally braced by ten levels of pre-loaded, cross-lot bracing and by two strata of
continuous jet grouted piles constructed between the walls. The uppermost jet grouted stratum
was excavated after installing the 9th level of bracing. Collapse occurred following excavation of
this upper jet grouted stratum. At the time of the failure, none of the lowest level of struts had
been installed.
The Committee of Inquiry identified the critical design and construction errors that led to the
failure of the earth retention system. They included use of an inappropriate soil simulation model
which overestimated the soil strength at the site and underestimated the forces on the support
system, an error in the design of the strut-waler support system with the connections being
under-designed, and deviations in actual construction from design plans, which further
aggravated the under-designed conditions. The net effect of the latter two factors resulted in the
strut-waler system being about 50% weaker than it should have been.
The Committee also found deficiencies in the project management that perpetuated and
aggravated the design errors. The human and systemic failures included inadequate
instrumentation and monitoring of works, improper management of instrumentation data, lack of
competency of persons carrying out specialized work, inability of the project management team
and supervisory personnel to identify adverse trends in the construction process and implement
corrective measures, and lack of clarity in the reporting structure for decision-making among the
different parties involved in the project.
Analysis and impacts on design: I only will discuss the “inappropriate soil simulations
model” because there is clearly a trend that results of finite element analyses are being used more
often in design to compute axial loads in internal bracing and bending moments in walls.
Figure 13. Implied Su in undrained plane strain compression for MC model

Whittle and Davies (2006) summarized the site characterization, selection of soil parameters
and analyses of soil-structure interaction that affected the design of the earth retention system.
They noted that a linearly-elastic, perfectly plastic Mohr-Coulomb (MC) model represented the
soil responses in the finite element analyses. The designers used drained effective stress
parameters for Young’s modulus, E', Poisson’s ratio, ', friction angle, φ', and cohesion, c', in an
undrained plane strain simulation of the construction of the excavation. Use of the effective
stress strength parameters leads to undrained shear strengths, Su, for lightly to normally
consolidated clays that are greater than actual values. This is illustrated in Figure 13 which
© ASCE
Geo-Congress 2020
shows total and effective stress paths on a Lambe stress space diagram for an isotropically
consolidated, undrained plane strain compression test on such a clay. The stress path dictated by
the MC model is vertical implying that the excess pore water pressure during shearing is equal to
the change in total mean normal stress, and correspond to a Skempton A parameter of 0.5 for
plane strain conditions. This is a result of the isotropic elastic response until failure for the
elastic, perfectly plastic assumption inherent in the MC model used in the finite element
simulation. Actual behavior of such clays are shown in the curved path where more excess pore
water pressure is generated, and hence the actual Su value will be lower than that dictated by the
MC model. Thus, the MC model is unconservative with respect to the shear strength of normally
to lightly overconsolidated clays.
Conversely, in a heavily overconsolidated clay, the MC model will be conservative with
respect to Su. While not shown in the figure, the effective stress path will bend to the right and
will intersect the failure line to the right of the vertical path dictated by the model assumptions,
and hence the model will predict lower Su values than actual results.
Whittle and Davies (2006) indicated that these assumptions greatly overestimated Su of the
normally consolidated marine clay in the original design leading to a serious underestimation of
computed wall deflections, bending moments and the mobilization of forces in the jet grouted
strata.
Implications for practice: In the context of designing earth retention support systems, it is
important to realize how the profession got to this point. Since the 1960s, flexible support
systems in the US were oft times designed based on apparent earth pressure diagrams (Terzaghi
and Peck, 1967). These diagrams are envelopes of maximum apparent pressures that were based
on axial loads in braces that were observed in the field for excavations constructed in similar soil
conditions. The measured axial loads reflected more than lateral earth pressures. They also are
affected by details of construction, what Peck elegantly referred to as workmanship, and
temperature effects. One can think of them as factored loading diagrams, in that the values are
larger than what will exist at any particular time during the process of excavation and bracing.
When one uses the results of finite elements to size the structural elements, one computes the
forces and moments in the structural elements based solely on the material properties used to
represent the soil behavior and the construction sequence input to the analyses. Typically this
analyses assumes that the excavation proceeds to a point several feet below the brace elevation
before the brace is added. This represents very good workmanship, from the point of view of
lack of overexcavation by a zealous excavator is not considered and the connections are perfectly
made in the analyses. Furthermore, the effects of temperature variations on the axial forces and
bending moments in the braces are not normally considered in the simulation. Consequently
some of the conservatism that was inherent in an apparent earth pressure design, is not included
in a design based on the finite element simulations. Thus the precedent that is associated with the
design and construction of excavation support systems is not directly applicable to finite element
based designs, and consequently, such designs warrant more detailed field observations than
typically associated with such projects to verify the predicted results than a design based on
apparent earth pressure envelopes.
Understanding the implications of a particular soil model with respect to computed response
is of fundamental importance when using finite element simulations to design excavation support
systems, as was dramatically illustrated in the collapse of the Nichol Highway. In general,
effective stress based soil models are more versatile than total stress models, but the undrained
strength implied by a more complex effective stress-based model is not as apparent as for models
© ASCE
Geo-Congress 2020
with total stress strength parameters or the relatively simple effective stress MC model discussed
earlier. When considering excavations through clays for undrained conditions, one should
simulate undrained compression and extension triaxial tests using the model parameters and
applicable effective stress conditions and compare them to site specific field vane or CPT-based
values and published correlations (e.g. Ladd 1991) to assure that the effective stress failure
parameters yield reasonable values of Su.
CONCLUSION
The four case histories presented herein show that it is not necessarily a lack of technical
knowledge that create failures. Rather, the process of design and construction are interrelated,
and peer review and communications throughout the life of a project are key ingredients to a
successful project. If one expands the notion of redundancy to include non-technical issues, then
if redundancy had been properly employed during design and construction, the failures probably
would not have occurred.
ACKNOWLEDGMENTS
The author collaborated with Wiss, Janney Elstner & Associates on the first three cases
presented in this paper. The contributions of Messrs. Mark Krueger, Jonathon McGormley and
Ernest Rogalla are greatly appreciated.
REFERENCES
COI (2005) “Report of the Committee of Inquiry into the Incident at the MRT Circle Line
Worksite that Led to Collapse of Nicoll Highway on 20 April 2004,” Ministry of Manpower,
Singapore.
Son, M. and Cording, E.J. (2005). “Estimation of Building Damage Due to Excavation-Induced
Ground Movements.” J. Geotech. and Geoenv. Eng., 10.1061/(ASCE)1090-
0241(2005)131:2(162)
Clough, G.W. and O’Rourke, T.D. (1990). “Construction Induced Movements of Insitu Walls.”
Proceedings, Design and Performance of Earth Retaining Structures, Lambe, P.C. and
Hansen L.A. (eds). ASCE, 439-470.
Ladd, C.C. (1991). “Stability Evaluation during Staged Construction,” J. Geotech. Eng., ASCE,
117(4), 540-615.
Osterberg, J.O. (1989). “Necessary Redundancy in Geotechnical Engineering,” J. Geotech. Eng.,
ASCE, 115(2), 1513-1531.
Peck, R.B. (1969). “Deep Excavations and Tunneling in Soft Ground.” Proc. 7th International
Conference of Soil Mechanics and Foundation Engineering, State-of-the-Art Volume, 225-
290.
Terzaghi, K, and Peck R.B. (1967). Soil Mechanics in Engineering Practice, 2nd edition, John
Wiley & Sons, New York, NY,
Thorburn, S., Greenwood, D.A., and Fleming, W.G.K. (1993). “The Response of Sands to the
Construction of Continuous Flight Auger Piles,” Proc. Deep Foundations on Bored and
Auger Piles, ed. Van Impe, Balkema, Rotterdam, 429-443
Van Impe, W.F. (1988). “Considerations on the auger pile design,” Proc. First International
Seminar on Deep Foundation on Bored and Augered Piles, Balkema, Rotterdam, 193-201
© ASCE
Geo-Congress 2020
Van Weele, A.F. (1988). “Cast-in-insitu Piles – Installation Methods, Soil Disturbance and
Resulting Pile Behavior,” Proc. First International Seminar on Deep Foundation on Bored
and Augered Piles, Balkema, Rotterdam, 219-217.
Whittle, A.J. and Davies, R.V. (2006). “Nicoll Highway Collapse: Evaluation of Geotechnical
Factors Affecting Design of Excavation Support System,” Proceedings, International
Conference on Deep Excavations, Singapore
© ASCE
Geo-Congress 2020
Twenty-Year Performance of a Mixed LLRW/RCRA Waste Disposal Facility

Rudolph Bonaparte, F.ASCE1; Beth A. Gross, F.ASCE2; Ranjiv Gupta, M.ASCE3;
John F. Beech, M.ASCE4; Leslie M. Griffin, M.ASCE5; and David K. Phillips6
1
Geosyntec Consultants, Atlanta, GA; Professor of the Practice, School of Civil and
Environmental Engineering, Georgia Tech, Atlanta, GA
2
Geosyntec Consultants, Austin, TX
3
Freeport-McMoRan, Phoenix, AZ
4
Geosyntec Consultants, Kennesaw, GA
5
6
ABSTRACT
A mixed low-level radioactive waste (LLRW) and Resource Conservation and Recovery Act
(RCRA) waste on-site disposal facility (OSDF) was constructed as part of the remediation of the
U.S. Department of Energy Feed Material Production Center in Fernald, Ohio. The 56-acre
(lined area) OSDF is fully constructed, filled with waste, and closed. The facility was operated
from November 1997 until October 2006, with operational data available from January 2000 to
the present, a period of nearly 20 years. Post-closure monitoring is ongoing. This paper presents
the design, construction, and performance of the OSDF, with a primary focus on its operational
performance. Available performance data for the OSDF’s leachate collection system, leakage
detection system, and horizontal-well vadose zone monitoring system are presented and
evaluated. Observations from post-closure monitoring of the OSDF are described. The data and
information indicate that the OSDF is performing well and is not impacting groundwater or
surface water at the site.
INTRODUCTION
This paper describes the project background and design, construction, and performance
monitoring of a mixed low-level radioactive waste and Resource Conservation and Recovery Act
(LLRW/RCRA) on-site disposal facility (OSDF) at the U.S. Department of Energy (DOE)
former Feed Material Production Center (FMPC) in Fernald, Ohio. The OSDF was designed to
achieve the DOE design life criterion of “1,000 years, to the extent reasonable and in any case
for 200 years.” Conventional RCRA land disposal facilities for both municipal and hazardous
waste typically have a design life goal in the range of 50 to 100 years. This case study has been
previously discussed in Bonaparte et al. (2008, 2011), where the design of the OSDF was
described in greater detail than herein, and in Bonaparte et al. (2016a), which briefly summarized
available performance data for the OSDF’s leachate collection and removal system (LCS) and
leakage detection system (LDS) through 2014. This paper expands on the performance
information in the 2016 paper by presenting additional data for the LCS and LCRS, plus
information obtained from the monitoring of a horizontal-well, vadose-zone monitoring system
(Horizontal Till Well [HTW]) installed in the glacial till beneath the OSDF liner system and
information on the performance of the final cover system. This case study has also been
discussed by Kumthekar and Chiou (2006), Powell et al. (2011) and Hooten et al. (2016).
© ASCE
Geo-Congress 2020
FERNALD SITE AND OSDF BACKGROUND INFORMATION

The DOE FMPC was located on a 1,050-acre site in Fernald, Ohio, approximately 18 miles
northwest of downtown Cincinnati. From 1951 to 1989, the facility was used for the processing
of uranium ore to produce uranium intermediate products (i.e., uranium trioxide [UO3] and
uranium tetrafluoride [UF4]) and high-purity uranium metals (99.3% U238 and 0.7% U235) for
shipment to other DOE facilities where the intermediate products and high-purity metals were
further processed to produce nuclear weapons. In 1989, DOE ceased uranium processing
operations at Fernald as the Cold War came to a close and the government’s need for Fernald’s
uranium intermediates and products fell sharply. Fernald’s environmental legacy in 1989
included 31 million pounds of nuclear metals, nearly 260,000 cubic yards of low-level
radioactive solid waste, 1 million tons of waste pit sludges, 2.5 million cubic yards of soils
impacted by LLRW and RCRA hazardous constituents, building debris, non-radiological solid
waste, and contaminated groundwater. Clean-up and restoration of the Fernald site was carried
out under the remedial process detailed in Title 40 of the U.S. Code of Federal Regulations
(CFR), Section 300, which codifies the requirements of the Comprehensive Environmental
Response, Compensation, and Liability Act (CERCLA). Under this process, the site’s
environmental legacy was divided into several large remediation projects to be implemented
under five CERCLA Operable Units (OUs). The total reported cost for remediating the site is
$4.4 billion.
The U.S. EPA Record of Decision (ROD) for OU2 (DOE, 1995b) at the Fernald site
addressed decommissioning and demolition (D&D) of buildings and excavation of soils
impacted by LLRW and RCRA hazardous constituents at concentrations above clean-up criteria,
and placement of those materials in the OSDF. The OU2 ROD allowed wastes from these
sources to be placed in an OSDF if the wastes satisfied certain waste acceptance criteria (WAC).
DOE estimated the quantity of such materials at 2.5 million bank cubic yards.
Conceptual design of the OSDF was developed through a CERCLA feasibility study (FS)
and ROD (DOE, 1995a,b). DOE next performed a site pre-design investigation (DOE, 1995c)
and established functional requirements for the OSDF (Fernald Environmental Restoration
Management Company [FERMCO], 1995). Design criteria were then developed (Geosyntec,
1997a), and the detailed design completed (Geosyntec, 1997b). Construction of the OSDF
commenced in May 1997, first waste was placed in November 1997, and the facility was filled
and closed in October 2006. Upon completion, the OSDF contained 2.96 million in-place cubic
yards of waste.
The DOE-reported actual cost for construction, filling, and closure of the OSDF is $224
million in 2006 dollars. This cost includes engineering, design, construction management and
quality control/quality assurance (QC/QA), waste placement and compaction, and construction
of the OSDF subgrade, liner system, and final cover system as well as the leachate collection and
transmission (including valve houses and pump station), stormwater management, and
environmental monitoring systems. The cost excludes waste pre-processing, waste transport from
the source to the OSDF, operation of the LCS, LDS, HTW, and groundwater monitoring system,
and site administration and management. The reported cost equates to: $4 million per acre for the
eight-cell, 56-acre lined footprint of the OSDF; $3 million per acre for the 74-acre footprint of
the final cover system; $76 per in-place cubic yard of waste; and $45 per estimated ton of waste.
Figure 1 presents an aerial photo of the Fernald site in 1996, prior to the start of OSDF
construction in the field on the right side of the photograph (east side of facility). Figure 2 shows
© ASCE
Geo-Congress 2020
the OSDF in 2002 with Cells 1 and 2 closed, Cell 3 being closed, Cells 4 and 5 being operated,
and Cells 6 and 7 in construction. Figure 3 shows the site in October 2006 with site remediation
complete, the OSDF filled, and Cell 8 in the final stages of closure. Post-remediation land use at
the Fernald site includes 400 acres of woodlands, 390 acres of prairie, 140 acres of wetlands and
surface waters, and 97 acres for the OSDF and various infrastructure. The site is now formally
the Fernald Preserve (www.lm.doe.gov/Fernald/Sites.aspx).
Fig. 1 Aerial photo of Fernald site, June 1996.

Future OSDF location is the open field on right side of photo.
Fig. 2. Aerial photo of OSDF in 2002 with various cells (numbered in yellow) being
constructed, operated, and closed.
© ASCE
Geo-Congress 2020
Fig. 3. Aerial photo of OSDF site in final stages of closure (with Cells 6 to 8 numbered in
yellow), October 2006.
Subsurface conditions at the OSDF site are illustrated in Figure 4. Preconstruction ground
elevations ranged from El. 618 feet (ft.) National Geodetic Vertical Datum (NGVD) on the
northeast corner of the site to El. 586 ft. in the southwest corner. Brown and gray glacial till form
the surficial stratigraphic unit at the OSDF site. Brown till, covered by a thin topsoil veneer, had
typical pre-construction thicknesses of 10 to 15 ft. within the OSDF footprint. As shown on
Figure 5, a portion of this material was removed to achieve the OSDF design base grades. The
thickness of the underlying gray till ranges from about 45 ft. at the OSDF north end to 15 ft. at
the south end. The till is underlain by sand and gravel of the Great Miami aquifer, an important
source of drinking water for the region. This sand and gravel unit is approximately 200 ft. thick
beneath the OSDF and is, in turn, underlain by shale and fossiliferous limestone with essentially
horizontal bedding. Information on the geotechnical and hydrogeological characteristics of the
soil units underlying the OSDF are given in Parsons (1995) and Bonaparte et al. (2011).
Radiological fate and transport modeling performed as part of the FS resulted in a
requirement that at least 12 ft. of undisturbed gray till be left in place below the OSDF to
function as both a hydraulic barrier and geochemical barrier to potential downward migration of
radiological waste constituents. The gray till was not penetrated in construction of the OSDF,
thereby meeting this requirement (Figure 5).
WASTE CHARACTERISTICS
Materials disposed in the OSDF consist of about 85 percent soil and soil-like materials
(SLMs) excavated as part of the remediation of the Fernald site and about 15 percent building
demolition debris, structural members, mass concrete, decommissioned equipment, lime sludge,
coal flyash, municipal solid waste, asbestos waste, and small quantities of other materials.
OSDF waste acceptance criteria (WAC) for radiological and hazardous constituents in soil
are given in Table 1. These criteria were established during the FS through fate and transport
modeling of leaching and leakage scenarios from the OSDF to groundwater. If waste materials at
© ASCE
Geo-Congress 2020
the Fernald site exceeded any of these criteria, the waste could not be placed in the OSDF.
Fig. 4. Idealized subsurface profile. Vertical exaggeration = 20x.
Fig. 5. OSDF north-south cross section. Vertical exaggeration = 10x.

The OSDF also had a large number of physical WAC, including for example: (i) concrete
structural members could not be more than 10 ft. long nor more than 18 in. thick and 4 ft. wide;
(ii) reinforcing bars protruding from concrete debris were cut to within 12 in. of the concrete;
(iii) metal structural members could not be more than 10 ft. long nor more than 18 in. thick and
© ASCE
Geo-Congress 2020
TABLE 2. Select functional requirements for Fernald OSDF.

Functional Requirements
Location (exclusions)
 within 200 ft. laterally of stream, lake, or wetland
 within 15 ft. vertically of the uppermost aquifer
 within a regulatory floodplain
 within an area of potential subsidence
 within 200 ft. laterally of a Holocene fault
Layout
 locate on east side of site, between main facility and power transmission lines
 achieve capacity of 2.5 million bank cubic yards (ultimately 2.95 million in-place cubic
yards)
 maximum height should be less than 70 ft. above original ground (visual impact)
 final cover system slopes must be between 5 and 25 percent
 liner system must overlie at least 12 ft. thickness of undisturbed gray till
 LCS drainage slope must be at least 2 percent
Engineering
 design life of 1,000 years to the extent reasonably achievable, and, in any case, at least
200 years
 long-term static slope stability factors of safety (FS) must exceed 1.5
 pseudo-static FS for 2,300-year recurrence interval earthquake must exceed 1.0
 double-liner system with secondary composite liner must be installed beneath waste
 secondary composite liner must include 3-ft. thick CCL with maximum hydraulic
conductivity of 1 × 10-7 cm/s overlain by HDPE geomembrane at least 60 mil thick
 final cover system must have a composite cap consisting of a 24-in. thick CCL with
maximum hydraulic conductivity of 1 × 10-7 cm/s overlain by HDPE geomembrane at
least 60 mil thick
 cover system must include a biointrusion barrier at least 3 ft. thick
 final cover system topsoil must have a predicted erosion rate of less than 5 tons/acre/year
and must resist gully initiation under the anticipated runoff tractive stresses
 final stormwater management system must accommodate 2,000-year, 24-hour storm flow
Conceptual Design Approach

The function of the OSDF is to isolate impacted material from the environment “for up to
1,000 years to the extent reasonably achievable, and, in any case, for 200 years.” This
performance criterion was adopted by DOE from 40 CFR §192.02(a), which provides minimum
federal disposal criteria for uranium and thorium mill tailings. The design was also developed
using the radiation protection goal of DOE Order 5400.5, which requires application of “As Low
As Reasonably Achievable (ALARA)” principles to activities involving the excavation,
transportation, and disposal of LLRW (DOE, 1989). These criteria were achieved in design by
addressing five potential mechanisms for OSDF performance degradation:
 internal hydrologic control – provide leachate containment and collection within the
© ASCE
Geo-Congress 2020
OSDF to prevent OSDF leachate from entering the environment;

 external hydrologic control – provide resistance to external hydrologic impacts, including
infiltration through the cover system and damage by surface-water run-on and runoff;
 geotechnical stability – provide adequate OSDF slope and foundation stability during
construction, filling, closure, and post-closure, including conditions associated with
potential long recurrence-interval earthquake events;
 erosional stability – provide resistance to erosion of OSDF soil layers to achieve minimal
erosional impacts throughout the performance period; and
 biointrusion resistance – provide resistance to OSDF intrusion by plant roots and

burrowing animals.
For the OSDF, natural (i.e., geological) materials were used in preference to manufactured
(i.e., geosynthetic) materials for certain functions (e.g., internal drainage layers). In addition,
relatively thick compacted-clay liners (CCLs) were incorporated into the design of both the liner
and cover systems in preference to liner systems constructed completely of geosynthetics, or
with thinner CCLs.
High density polyethylene (HDPE) geomembranes were specified in preference to other
types of geomembranes based, in part, on their durability characteristics. Studies available in
1995 indicated that the HDPE service life would be on the order of hundreds of years (Koerner et
al., 1992; Bonaparte, 1995). Later studies (e.g., Bonaparte et al., 2002; Rowe, 2005) indicated
that at an ambient ground temperature of about 55°F (12°C), the design life for buried HDPE
geomembranes would be on the order of 1,000 years. More recent studies (Bonaparte et al.,
2016b) have shown that radiological doses to which the HDPE geomembranes are anticipated to
be exposed in the OSDF are well below the threshold doses likely to induce geomembrane
physical and mechanical property changes. Regulations required the HDPE geomembranes to be
at least 60-mil thick. However, the design specified that the geomembrane be at least 80-mil
thick as another measure to increase the service life of this material. The rationale for the thicker
material is that the primary degradation processes for HDPE geomembranes involve polymer
chain oxidation that starts at the surface and works inward. In the event of surface oxidation, a
thicker material will retain its properties longer than a thinner material, all other factors being the
same. The specifications also required the HDPE formulation to contain 2 to 3 percent
antioxidant-containing carbon black (ASTM D1603, which has since been replaced by ASTM
D4218) and to have a minimum environmental stress crack resistance (ESCR) of 500 hours when
tested in accordance with the notched constant tensile load (NCTL) method of ASTM D5397. In
1995, HDPE geomembrane specifications typically required ESCR of 100 to 200 hours; the more
stringent specification for the OSDF provides a material with better aging potential and less
potential for long-term brittle rupture under stress.
All hydraulic barriers in the liner and final cover systems (primary liner, secondary liner, and
cover barrier layer) were designed as soil-geosynthetic composite barriers in preference to single
component barrier layers. Composite barriers provide superior hydraulic containment compared
to single-component barriers (Giroud and Bonaparte 1989a,b). The individual components of
composite barriers also help to protect the other component. For example, CCLs provide good
bedding layers for geomembranes and geomembranes help to prevent desiccation cracking of
CCLs after installation.
For design, the OSDF performance period was divided into three operating timeframes that
are described in detail in Bonaparte et al. (2008, 2011) but not repeated here for brevity. The
three periods are: (i) initial period; (ii) intermediate period; and (iii) final period. Responsibility
© ASCE
Geo-Congress 2020
for maintenance and stewardship for the OSDF rests in perpetuity with the U.S. government. The
OSDF design allows government decision makers at the time of the final period to select an
appropriate continuing post-closure management strategy for the facility.
Detailed Design Development
Figures 5 and 6 show, respectively, north-south and east-west cross sections of the OSDF as
designed to meet the functional requirements and conceptual design described above. The design
includes eight cells constructed sequentially from north (Cell 1) to south (Cell 8) over the active
life of the facility. The photograph in Figure 2 illustrates this sequential cell development. The
OSDF was designed as an essentially above-ground facility. The bottom of each cell is graded in
a herringbone pattern at a 2 percent slope to drain leachate by gravity to the west side of the cell.
This grading configuration was designed to follow the pre-construction natural grades in the
project area, thereby limiting foundation excavation requirements. This configuration also allows
the LCS and LDS drainage pipes to exit the cell at or near the original ground surface elevation.
Fig. 6. OSDF east-west cross section.

Figure 7 shows the double-composite liner system configuration for the OSDF. The system
consists of a combination of hydraulic barrier layers, drainage layers, and filter and cushion
geotextiles. Gravel was specified for the LCS and LDS drainage layers in preference to
geosynthetics due to design life considerations. The gravel has a maximum particle size of 0.75
in. and less than 2 percent fines. Durability considerations also drove selection of the HDPE
geomembranes. The durability characteristics of the geosynthetic clay liner (GCL) hydraulic
barrier are less well defined than either the HDPE geomembrane or compacted clay. However,
the GCL is intended to function principally during the active life when most leachate is
produced, so, it was not critical to define a long-term design life for this component.
Each of the eight OSDF cells was designed with intercell berms so that the LCS and LDS for
a cell captured only the liquids produced in that cell. Accordingly, each LCS and LDS has its
own liquids removal pipe (Figures 8 and 9). All LCS/LDS pipes gravity drain from the cells to a
double-walled collection forcemain located in valve houses outside each cell (Figure 6). Gravity
drainage was judged to be a more reliable long-term liquids removal strategy than one using
© ASCE
Geo-Congress 2020
Redundant features were incorporated in the liner system design, including a second, back-up
LCS liquids removal pipe (Figure 9), rather than the one pipe that is customarily used in landfill
applications.
Fig. 9. Liner and leachate collection system cross-sectional details at downgradient cell
outlet.
Fig. 10. OSDF final cover system.

One unique aspect of the OSDF design is management of precipitation that falls on an active
cell. Standard practice for municipal solid waste (MSW) landfills is to cover the waste with daily
© ASCE
Geo-Congress 2020
Figure 10 shows the configuration of the final cover system. Final cover slopes are 6
horizontal:1 vertical (6H:1V) on the sides of the OSDF and 6 percent on the top deck.
Construction of the CCL in the final cover system is shown in Figure 11. The final cover side
slope inclination is flatter than for most MSW landfills and was chosen based on the results of
slope stability and erosion gullying analyses to achieve the functional requirements described
previously. To maintain the long-term functionality of the final cover system, the biointrusion
layer is designed to arrest plant root and/or burrowing animal intrusion. Placement of the
biointrusion layer is shown in Figure 12. The thickness of the final cover system was also driven
by radon attenuation considerations (U.S. Nuclear Regulatory Commission [NRC], 1984 a,b)
As part of the design process, extensive laboratory and field evaluations were conducted.
These pre-design evaluations are described in detail in Bonaparte et al. (2008, 2011) and
included:
 Hydraulic conductivity of CCLs – This evaluation involved laboratory hydraulic
conductivity testing of both laboratory and field compacted samples of the brown till
material used to construct the OSDF CCLs. It also involved the installation and
monitoring of two field sealed double-ring infiltrometers (SDRI). From this evaluation,
an acceptable zone of compaction moisture content and dry unit weight (i.e., zone of
compaction conditions that would result in a CCL hydraulic conductivity not greater than
1 x 10-7 cm/s) was established for the brown till, along with pre-processing requirements
(e.g., over-size particle removal, moisture conditioning) and compaction equipment
requirements.
 Compatibility of liner system with OSDF leachate – A leachate-geomembrane
compatibility testing program was developed to evaluate the compatibility of five
different commercial 80-mil thick HDPE geomembranes with OSDF leachate. The
geomembranes were obtained from four major U.S. manufacturers and were evaluated
using U.S. EPA Method 9090, “Compatibility Test for Waste and Membrane Liners.”
This method involved submersion of geomembrane samples in test leachate at 73°F
(23°C) and 122°F (50°C) for 120 days. Specimens of the various samples were retrieved
at 30, 60, 90, and 120 days and evaluated against controls for changes in physical,
mechanical, and chemical properties. A detailed presentation of the leachate-
geomembrane compatibility testing program is contained in Geosyntec (1997c). Based on
the testing results, all five HDPE geomembranes were qualified as acceptable based on
radiological compatibility; however, only four were found acceptable based on a separate
ESCR evaluation.
 Soil-geosynthetic shear testing program – An extensive soil-geosynthetic interface
testing program was conducted to support slope stability analyses performed as part of
the detailed design. Twenty-seven direct shear tests were performed in a 12 in. × 12 in.
shear box in accordance with ASTM D5321. The program also included testing for soil
moisture content, compaction characteristics, particle size distribution, plasticity, and
classification. Tested materials included HDPE geomembranes, brown till used for CCL
construction, and several different internally reinforced GCLs. From the testing program,
minimum strength requirements as a function of normal stress were defined for the
various soil-geosynthetic interfaces.
CONSTRUCTION
As already noted, the Fernald OSDF was constructed sequentially as eight contiguous cells.
© ASCE
Geo-Congress 2020
Construction of the first cell (Cell 1) began in May 1997, and closure of the last cell (Cell 8) was
completed in October 2006. The sequence of activity for each cell essentially involved
excavation of topsoil and brown till to the design base grades; construction of earthen perimeter
and inter-cell berms; installation of the liner system, LCS/LDS piping, and liner penetration
boxes; placement of protective layer soil over the liner system; construction of the valve houses
and leachate transmission piping; construction of a truck haul road into the cell; installation of
interim stormwater management controls; placement of a 3-ft. thick select layer of waste (soil
and SLM); placement of wastes in accordance with an Impacted Materials Placement Plan to
designated final grades; placement of another 3-ft. thick select layer of waste and a 12-in. thick
soil contouring layer; installation of the final cover system; topsoil seeding; and construction of
final stormwater management controls.
A critical aspect of the project was to always provide constructed disposal capacity in time to
prevent delays in overall site remediation. During the nine-year active life of the OSDF, cell
construction and cell closure occurred concurrently with impacted material placement. The
preferred construction season in Ohio starts in late April to early May and continues until mid-
November. A full season was required to construct the multiple layers of a cell’s liner or final
cover system. Key to maintaining the construction schedule was the advanced procurement and
processing of materials, including the necessary QC/QA testing.
PERFORMANCE TO DATE
Overview
Filling of Cell 1 of the OSDF began in November 1997, so this cell has been functioning for
22 years. Cells 2 through 8 have been functioning for successively shorter periods of time, with
Cell 8 being the shortest at 15 years. OSDF operational data include LCS, LDS, and HTW liquid
removal rates, LCS, LDS, and HTW liquid chemical constituent concentrations, and results from
inspections of the final cover system (DOE, 2019). Overall, performance of the OSDF to date is
consistent with design expectations and the project is considered by DOE to be a critical success.
In August 2016, DOE issued a report titled “Fourth Five-Year Review Report for the Fernald
Preserve” (DOE, 2016). The report summarizes DOE’s findings with respect to OSDF
performance as follows: “The cap and liner systems of the OSDF are functioning as designed
and are successfully isolating the waste materials. The volume of leachate generated from the
OSDF is continuing to decline, and the leachate is being effectively collected and treated to
minimize impacts to human health and the environment.” A report presenting the findings of
DOE’s fifth five-year review is expected in 2021.
Available Data
Available operational data for the OSDF include monthly combined LCS flow volumes for
all cells from January 2000 (when Cells 1 to 3 were active) through December 2008 (when all
cells had been closed), monthly LCS flow volumes for each cell from January 2009 through
December 2018, monthly LDS flow volumes for each cells from January 2000 through
December 2018, and quarterly to annual chemical analysis data for select constituents present in
LCS flows and, when sufficient flow volumes could be collected, LDS flows. Chemical analysis
data are also available for samples of HTW liquids.
© ASCE
Geo-Congress 2020
LCS Flow Rates

The primary sources of water contributing to leachate generation in the OSDF were
precipitation and water used to suppress dust generation during cell operations. The volume of
dust suppression water was very significant for this facility as DOE had strict air quality criteria
related to airborne particulates and radionuclides. The significance of the dust suppression
activities is evidenced by the fact that in 2006, 10.6 million gallons of precipitation fell onto the
active cells of the OSDF while 25.9 million gallons of dust suppression water were applied in
that same year.
Fig. 13. Monthly Average OSDF LCS Flow Rates and Comparison to Design Estimates.
Monthly average LCS flow rates for the entire OSDF are presented in Figure 13. Flows are
presented using a normalized metric, gallons per acre of liner system per day (gpad). From
March 2000 through December 2001, when only Cells 1 to 3 were active, monthly LCS flow
rates ranged from 400 to 2,500 gpad, or 12 to 80% of average annual precipitation, with a mean
of 1,300 gpad, or 42% of average annual precipitation. From January 2002 through October
2006, when additional cells were being constructed and filled, monthly LCS flow rates ranged
from 30 to 2,600 gpad, or 1 to 85% of average annual precipitation, with respective averages of
1,000 gpad and 33%. It can be seen that the monthly-average LCS flow rate decreased
substantially late in the operating life of the OSDF, after all cells were constructed and contained
waste and the initial cells were closed. The LCS flow rate dropped by approximately one-order
© ASCE
Geo-Congress 2020
of magnitude within one month of closure of the last operating cells (i.e., Cells 7 and 8) and it
has been stable at a monthly average value of 10 gpad or less since then. From November 2006
to December 2014, combined LCS flow rates for the closed cells continued to slowly decrease
from about 0.6 to less than 0.2% of average annual precipitation. The rates have remained below
that level through the most recent reporting update at year-end 2018.
From Figure 13, monthly average leachate flows had considerable month to month variation
during active OSDF operations. This variability was associated with the occurrence of rainfall
events and the operational status of the various OSDF cells at any point in time. The very low
and high “peakiness” of the measured LCS flow rates was also affected by the operating
schedule for the leachate transmission system that conveyed leachate from the OSDF, through
the valve houses, to the on-site advanced wastewater treatment plant.
The measured OSDF LCS flow rates can be compared to the original design estimates made
in 1995 using the U.S. EPA Hydrologic Evaluation of Landfill Performance (HELP) model. The
estimated average annual leachate generation rates obtained using the HELP model were 1,150
gpad for a cell in the initial stage (10 ft thickness of waste), 700 gpad for a cell at the operational
stage (30 ft. thickness of waste plus soil cover), less than 1 gpad shortly after closure, and 0.002
gpad at the end of the post-closure period. These flow rates were combined on a monthly basis
based on the operational status of each cell to obtain the overall facility design LCS flow rate
shown in Figure 13. From Figure 13, it can be seen that the leachate generation rates calculated
using actual LCS flow data are significantly higher than the original design estimates based on
the HELP model. Two project-specific factors have been identified that likely account for the
differences: (i) the presence of the impacted stormwater catchment area in each cell (previously
described); and (ii) the large volume of dust suppression water applied to active cells. The
stormwater catchment areas were introduced to the design after the HELP modeling was
completed, and the decision was made at the time not to revise the analyses to include the greater
leachate generation volumes caused by these areas. The application of dust suppression water
was also not included in the original analyses.
The measured OSDF LCS flow rates can also be compared to the measured LCS flow rates
in a U.S. EPA database reported by Bonaparte et al. (2002). The U.S. EPA database includes
data for 73 MSW landfill cells and 32 hazardous waste (HW) landfill cells. The LCS data are
evaluated in several ways including as a function of average annual precipitation. The historical
average annual precipitation in the Fernald vicinity is approximately 40 in. This rainfall average
is closest to the northeast (NE) and southeast (SE) facility categories in the database. The
average LCS flow rates in the database for the MSW cells in the NE and SE during the
operational stage were 378 and 314 gpad, respectively, and for the HW cells, 576 and 524 gpad
respectively. These average LCS flow rate results from the database are somewhat lower than the
design values calculated in 1995 for the OSDF project using the HELP model, and significantly
lower than the measured OSDF operational stage LCS flow rates.
Leachate generation rates after OSDF closure are compared to rates for other closed landfills
in the U.S. EPA database, cited above, in Figure 14, taken from Bonaparte et al. (2002) and
expanded with data from U.S. EPA (2017) for 35 cells at 8 HW landfills, including data for 3
cells at 1 HW landfill included in the Bonaparte et al. (2002) report. This figure includes data for
11 MSW and 57 HW cells from the databases, plus the 8 OSDF cells. The OSDF data are for
2009 to 2018, which represent 3 to 17 years after closure. From the figure, it is seen that the post-
closure LCS flow rates for the OSDF are consistent across cells, generally consistent with
median post-closure rates for the 11 MSW and 57 HW cells, on the order of 5 to 10 gpad, and
© ASCE
Geo-Congress 2020
showing a slowly decreasing trend during the post-closure period.

Fig. 14. Average LCS flow rates (gpad) versus years after closure for eleven MSW cells
(blue circles), 22 HW cells (green squares), and the 8 OSDF cells at Fernald (red
diamonds). MSW and HW data are from Bonaparte et al. (2002) and U.S. EPA (2017).
LDS Flow Rates
The liquid of concern in monitoring the LDSs of the OSDF cells is leakage through the
composite primary liner. However, leakage is not the only potential source of LDS liquids in a
typical landfill, as indicated by Figure 15. Potential other sources of LDS liquids are drainage of
water (mostly rainwater) that infiltrates the LDS during construction and drains after the start of
facility operation (construction water), water expelled from the LDS during waste placement as a
result of LDS compression under the weight of the impacted material (compression water), and
infiltration of groundwater perched in the brown till that finds a pathway through the secondary
composite liner (infiltration water). Gross et al. (1990) provides a general discussion of each of
these potential sources along with methods to estimate liquid quantities from each source given
site conditions and facility design details. At the start of cell operation, much of the LDS flow is
often attributable to construction water. Continuing flow in the LDS after the construction water
has fully drained can be due to one or several of the sources in Figure 15. Site-specific analyses
using the methods described in Gross et al., coupled with comparisons of the chemical
compositions of the LDS and LCS flows, allow evaluation of the most likely source(s) for the
LDS flows.
© ASCE
Geo-Congress 2020
Fig. 15. Potential sources of LDS liquids (modified from Bonaparte and Gross, 1990).
Fig. 16. Monthly Average LDS Flow Rates for OSDF Cells.
Monthly average LDS flow rates for the OSDF are presented in Figure 16. LDS flow rates
from all cells have been low and less than the LDS action leakage rate of 20 gpad. The action
leakage rate is the LDS flow rate that, if exceeded, triggers response actions. If exceeded, the
© ASCE
Geo-Congress 2020
facility owner/operator is required to evaluate the source(s) of the LDS flow and the potential for
a release from the facility, and then to implement corrective actions if found necessary. LDS
flow rates for the Fernald OSDF were expected to be highest shortly after the cell was placed
into operation (i.e., initial period of operation) and then decrease over time. However, this was
not the case for many of the OSDF cells. Instead, these cells exhibited their highest LDS flow
rates later in their development – sometimes near the time of cell closure. During their active
period of operation, average LDS flow rates for the cells ranged from 0.2 to 6 gpad, with five of
the eight cells having average LDS flow rates exceeding 1 gpad. These flow rates, while very
low in absolute terms, are higher on average than was expected.

The possible sources of the active-period LDS flows shown in Figure 16 are those illustrated
in Figure 15. Because the LDS material in the Fernald OSDF is clean gravel, water that
infiltrated the LDS during construction should have drained prior to or shortly after each cell
became operational, and the gravel should not have retained capillary water that could be
squeezed out as the gravel is loaded by overlying waste. Therefore, construction water and
compression water are not likely sources of the flow. Given that the primary liner for the OSDF
contains a GCL, some of the active-period LDS flow (up to approximately 2,000 gallons/acre
assuming a compression ratio of 0.27) could have been due to consolidation of the GCL
bentonite as the OSDF cells were filled with waste. This reflects a relatively small contribution
to LDS flow. As shown subsequently, comparison of LCS and LDS chemical constituent data
indicates that only a relatively small fraction of the LDS flow could have been leachate leakage
through the composite primary liner. Perched groundwater was also a potential source. However,
if perched groundwater was a significant contributor, then the correlation between flow rate
reduction and cell closure seen in Figure 16 would not be expected and LDS flows would not
have decreased to near zero levels over time. Hence, the source(s) of the LDS flows during the
active period remain uncertain. In any case, the post-closure LDS flow rates are extremely low
and several cells are generating insufficient flow for LDS tank pump out or semi-annual water
quality sampling. In 2018, 17 years after final closure of Cell 1 and 12 years after final closure of
Cell 8, the LDS tanks for Cells 1 to 3, 5, and 7 were too dry for liquid removal. Measured
maximum monthly LDS accumulation rates for Cells 4, 6, and 8 were 0.08, 0.19, and 0.005 gpad
respectively, and total accumulation rates were insufficient to trigger automatic pump out of the
LDS tanks.
Performance of Composite Primary Liner
The performance of the primary liner for the OSDF was evaluated by calculating the monthly
apparent hydraulic efficiency (Ea) of the liner for each cell. The parameter, Ea, introduced by
Bonaparte et al. (1996), is defined as:
 LDS Flow Rate 
Ea  %   1   100 (1)
 LCS Flow Rate 
The higher the Ea, the smaller the flow rate from an LDS compared to the flow rate from an
LCS. The parameter Ea is referred to as an “apparent” liner hydraulic efficiency because, as
previously noted, flow into the LDS sump of a landfill may be due to sources other than primary
liner leakage (Figure 15). Ea was calculated using total flow from the LDS, regardless of source.
If the only source of flow from the LDS is primary liner leakage, then Eq. 1 provides the “true”
liner hydraulic efficiency, Et. True liner efficiency provides a measure of the effectiveness of a
particular liner in limiting or preventing advective transport across it. For example, if a liner has
© ASCE
Geo-Congress 2020
an Et of 99%, the rate of leakage through the primary liner would be 1% of the LCS flow rate.
The true efficiency of a liner is not a constant, but rather a function of the hydraulic head in the
LCS and size of the area over which LCS flow is occurring (the area is larger at high flow rates
compared to low flow rates).
LCS and LDS flow rates and Ea for Cell 4, are shown in Figure 17. For this cell, LDS flows
were highest during the cell’s active period and decreased rapidly at the time of cell closure.
LDS flow rates were less than 0.5 gpad within two years after closure and less than 0.2 gpad
(monthly Ea generally greater than 99%) within three years after closure. LDS flows have
continued to decrease and since 2011, approximately 6 years after closure, have only been
recorded intermittently (e.g., several months a year), with monthly LDS flow rates ranging from
0.005 to 0.08 gpad. After cell closure, monthly Ea values were initially greater than 99.9%.
However, as LCS flow rates decreased while LDS flows, when they were recorded, remained
approximately of the same magnitude, monthly Ea decreased somewhat for the months in which
LDS flow were recorded. During 2017 and 2018, Ea values for Cell 4 ranged from 98.5% to
100% on a monthly basis and averaged approximately 99.5% on an annual basis. All other cells
have also been producing very low flow rates (Figure 16) and exhibiting high Ea values,
indicating the GM/GCL composite primary liner is effective in all cells. During 2017 and 2018,
the average monthly Ea value for Cell 6 was approximately 99.3% and the Ea value was above
99.9% for all other cells.
Fig. 17. Monthly Average LCS and LDS Flow Rates and Apparent Liner Hydraulic
Efficiencies for Cell 4.
Figure 18 shows the post-closure LDS flow rates for all eight OSDF cells (i.e., the same post-
© ASCE
Geo-Congress 2020
closure data shown in Figure 16), with the data plotted as months since final closure instead of
by date as was done in Figure 16. The rate of reduction in LDS flow rate with increasing time
after closure can be clearly seen in the figure. All eight LDSs are producing very low flows,
indicating effective performance of the composite primary liner in all cells.
Fig. 18. Monthly-average post-closure LDS flow rates for the eight OSDF cells.
HTW Vadose Zone Monitoring System
The OSDF design includes an HTW vadose zone monitoring system that utilizes horizontal
monitoring wells installed in the clay till soil beneath the OSDF. An HTW is installed at the
location where the LCS and LDS pipes penetrate the liner system in each of the eight OSDF cells
(Figures 6 and 7). The purpose of the HTW monitoring system is to provide for the rapid
detection of any leachate release from the OSDF cells due to failure at the most critical locations
in the liner system (i.e., the liner penetrations and low points in the cells). The wells were
installed in trenches excavated into the till at a depth of approximately 4 ft. below the elevation
of the bottom of the LDS pipe penetration. Each HTW well includes a 62 ft. length of 6-in.
diameter HDPE perforated pipe embedded in a 2ft. × 2 ft. gravel filled trench. Vadose zone
water that enters each HTW flows by gravity through a solid-wall section of pipe to a monitoring
point at the downgradient western boundary of the OSDF.
The HTW system has been monitored on a roughly quarterly basis starting in 1999 for Cells
1 to 3 and in subsequent years for the latter cells. Monitoring parameters have included the
annual volume of water removed, and total uranium (mg/L), boron (mg/L), sodium (mg/L), and
sulfate concentrations.
© ASCE
Geo-Congress 2020
Available monitoring data show that the highest HTW flow rates occurred at the start of the
monitoring period with decreasing flow rates over time. For Cells 1 to 3, total annual HTW flows
were 5,000 to 6,000 gallons in 1999, progressively decreasing to 800 to 1,600 gallons in 2017.
For Cells 4 to 6, annual HTW flows decreased from 10,000 to 35,000 gallons at the start of
monitoring in 2002/2003 to 700 to 2,100 gallons in 2017. For Cells 7 and 8, total annual flows
were 2,800 to 4,000 gallons at the start of monitoring in 2004 and had dropped to 0 and 700
gallons in 2017. The decreasing flow rates with time are not surprising given that construction of
the OSDF liner system in a cell area cuts off vertical recharge to the till layer. The reasons for the
much higher initial flow rates for the Cells 4 to 6 HTWs compared to those for the other cells has
not been investigated by the authors.
The till in which the HTWs are located has low level contamination as a result of historical
operations at the FMPC. The mean concentrations from pre-construction samples of total
uranium for 17 certification units (i.e., sample areas) that correspond to the locations of the
HTWs ranged from 6.0 to 57.2 milligram/kilogram (mg/kg). These concentrations can be
compared to a site background value of 3.7 mg/kg and site remediation goal of 82 mg/kg
(Hooten et al., 2016). Average uranium concentrations in HTW liquid samples, based on
quarterly monitoring for time periods ranging up to 20 years, have been in the range of 3.7 to
19.5 micrograms per liter (ug/L). Samples from many of the wells exhibited increasing
concentration trends in the first few years of monitoring followed by relatively constant
concentrations to the present. To benchmark these uranium concentrations, the U.S. EPA
drinking water maximum containment level (MCL) for total uranium is 30 μg/L.
Chemical Constituent Data
Quarterly total uranium concentrations in LCS and LDS flows from each cell are available
from the start of operation of each cell through the end of 2018, a time period of 21 years for
Cell 1 and progressively shorter time periods for each succeeding cell, with Cell 8 having a
monitoring period of 14 years. In addition, more limited data are available for a few other
chemical constituents. These same types of data are available for the HTW beneath each cell for
the same period (discussed above).
Average total uranium concentrations in the LCS, LDS, and HTW flows for each of the eight
OSDF cells over the entire monitoring period are shown in in Table 3 (DOE, 2019). Also shown
are the ratios of the average LDS uranium concentrations to average LCS concentrations,
expressed as percentages.
TABLE 3. Average Total Uranium Concentrations (µg/L) by Cell Since the Start of Cell
Operations
Cell LCS LDS HTW LDS/LCS (%)
1 84.3 10.8 8.4 12.8
2 124 14.5 10.8 11.7
3 76.8 17.0 19.2 22.1
4 88.2 14.6 5.60 16.5
5 122 15.6 9.2 12.8
6 131 26.6 12.0 20.3
7 165 25.7 3.9 15.6
8 172 23.0 5.0 13.4
© ASCE
Geo-Congress 2020
As can be seen, average uranium concentrations in the LCS of each cell range from 76.8 to
172 μg/L, while average uranium concentrations in the LDS range from 10.8 to 26.6 μg/L. The
ratios of uranium concentrations in LDS to LCS liquids are in the range of 11.7 to 22.1%. Stated
differently, on average, uranium concentrations in LCS flows are 4.5 to 8.5 times higher than the
concentrations in LDS flows. Inspection of the LCS and LDS monitoring data shows that in most
cases, the measured quarterly concentration in both systems are relatively constant over time,
with a slight trend of increasing concentrations in some cases. The variability in average uranium
concentrations between cells is attributed to the composition of the waste placed in the cells and
the operating practices used in the cells (e.g., the moisture content of the waste).
Fig. 19. Uranium-Sodium Bivariate Plot for 4 Cell (DOE, 2019).

Figure 19, taken from DOE (2019), shows a uranium-sodium bivariate plot for the LCS,
LDS, and HTW liquids for Cell 4. The data, collected from 2002 to 2018, indicate that the LCS
and LDS flows for Cell 4 have distinctly different chemical fingerprints, and the water draining
from the cell LDS is not principally leachate. The low levels of uranium found in the LDS flows
may be indicative of several possible sources, including minor amounts of primary liner leakage,
water entering the LDS during construction (e.g., dust suppression water), and/or infiltration of
perched groundwater. If a source is primary liner leakage, the uranium concentrations suggest
that the leachate is contributing only 15-20% of the flow in the LDS (assuming the uranium is
not being attenuated as it migrates through the GCL). Thus, only a fraction of the very low flows
from the LDS in Cell 4 could potentially be due to primary liner leakage. High sodium
concentrations in the LDS versus LCS are likely due to leaching of sodium ions from the GCL
and/or cation exchange in the GCL between sodium and calcium ions in the LDS liquid.
Comparison of the LDS and HTW bivariate plots, coupled with the measured pre-construction
© ASCE
Geo-Congress 2020
concentrations of uranium in the till layer and the extremely low levels of post-closure liquid
accumulation in the LDS, are indicative of the till being the source of uranium concentrations in
HTW liquid samples and not leakage from the LDS through the secondary composite liner.
Cover System Erosion and Maintenance
The final cover system for the Fernald OSDF was designed to support a stand of native
prairie grasses and with a maximum calculated topsoil erosion rate of 4.3 tons/acre/year based on
the Modified Universal Soil Loss Equation (MUSLE). If erosion were to occur at this maximum
calculated rate, the thickness of the cover system topsoil and vegetative soil layer would be
reduced from 27 in. to 22 in. after 200 years, and to 3.5 in. after 1,000 years. The cover system
was also designed to resist rill and gully formation during the 2,000-year storm event. Analyses
conducted using the Temple (Temple et al., 1987), Horton/NRC (NRC, 1990), and permissibly
velocity (NRC, 1990) methods indicated during design that the formation of rills and gullies
would not occur during the design storm as long as the final cover system was vegetated with a
well-established stand of native grasses. In addition, the biointrusion barrier was designed to
resist erosion in the 2000-year storm and probable maximum precipitation (PMP) event. The
Stephenson (Abt et al., 1988) and Hartung and Scheuerlein (1970) methods were used to
evaluate the stability of the biointrusion barrier under these design storm events. The analyses
indicate that the biointrusion barrier layer would stop the advancement of gullies in the 2,000
year storm, although this scenario is not expected given the analysis results for the vegetation
and topsoil layer described above.
Quarterly post-closure inspections of the Fernald OSDF cover system and surface-water
management features have occurred since December 2006, a period of 13 years. The cover
system is inspected for evidence of erosion rills/channels, depressions, cracks, ponded water, and
seepage; vegetation coverage; presence of woody and/or invasive vegetation; evidence of
biointrusion; and evidence of vehicle traffic. Figure 20 shows photographs of cover vegetation
from inspections in September 2007, 2011, and 2018.
Based on a review of the quarterly inspection reports, relatively modest levels of final cover
system maintenance and repair have been required since OSDF closure. These include repair
of erosion rills, repair of small mammal (e.g., mole to groundhog) burrows, management of non-
native and noxious plant species, removal of woody vegetation, removal of rocks that surface as
topsoil settles, and other general site maintenance (e.g., perimeter fence repair and litter
removal). Inspection of the quarterly monitoring reports between December 2006 and June 2011
shows that 70 erosion rills were observed during that period, some requiring repair by backfilling
and reseeding. This equates to, on average, 1 rill per 600 feet of OSDF cover crest length per
year. Rills were repaired if they reached a depth of 6 in. or more. The frequency of erosion rills
observed in the final cover system has decreased substantially in the time since facility closure as
vegetation coverage has increased and stabilized. Comparison of the inspection reports shows
that the frequency of rilling is decreasing on an annual basis in parallel with the establishment of
denser and more stable cover vegetation. For 2018, only two locations of “minor erosion” were
documented in the quarterly inspections.
Over the years, animal burrows have occurred in the final cover system of every cell, have
been documented each year in the final cover system of at least one or two of the eight OSDF
cells, and have been found more frequently during winter months. Burrow holes are repaired by
backfilling, re-seeding, mulching, and applying temporary irrigation. Due to the continuing
presence of animal burrows, the biointrusion barrier component of the final cover system has
© ASCE
Geo-Congress 2020
proven integral to preventing vertical migration of the burrow holes.

Fig. 20. East face of Cell 1 showing the progressive evolution in cover vegetation from
September 2007 (A) to September 2011 (B) to September 2018 (C).
Starting in September 2010, annual surveys of the final cover system have been conducted to
quantify the condition of the vegetative cover. Since the last survey, all cells have had at least
90% vegetative cover with the vegetation being composed of 50-70% native species. Due to the
progressive increase in vegetative coverage and growth since OSDF closure, erosion issues due
to exposure of bare or poorly vegetated topsoil have largely been eliminated, as noted
above. Conversely, the frequency of undesired vegetation, such as woody materials, and invasive
plant species, has increased. By far, the most common conditions requiring attention in recent
© ASCE
Geo-Congress 2020
years have been the occurrence of woody vegetation (e.g., cedar, pear, willow, dogwood, and
honeysuckle) and invasive species (e.g., thistle, teasel, and garlic mustard). Current maintenance
procedures require removal of the woody vegetation at the root level, backfilling the root holes,
and re-seeding. Spot applications of herbicides are routinely applied to control invasive species
and noxious weeds. In addition, beginning in March 2016, controlled burning of the vegetation
covering the OSDF has been conducted as a management practice to control the presence of
invasive species and maximize the prevalence of prairie grasses. In addition, controlled burning
allows the inspection of the burned portion of the OSDF final cover system prior to regrowth of
vegetation. Controlled burning of the cover vegetation presently occurs on a biennial basis (i.e.,
half of the OSDF undergoes a controlled burn one year and the other half the next).
ACKNOWLEDGEMENTS
The S. M. Stoller Corporation holds a contract with the DOE Office of Legacy Management
which includes environmental monitoring and reporting at Fernald. Data and inspection results
collected by Stoller can be found at www.lm.doe.gov/Fernald/. The views expressed in this paper
are those of the authors and do not necessarily represent those of either DOE or the United
States.
REFERENCES
Abt, S.R., Wittler, R.J., Ruff, J.F., LaGrone, D.L., Khattak, M.S., Nelson, J.D., Hinkle, N.E., and
Lee, D.W., (1988). “Development of Riprap Design Criteria by Riprap Testing in Flumes:
Phase II – Followup Investigations”, NUREG/CR-4651-V2 ORNL/TM-10100/V2, U.S.
Nuclear Regulatory Commission, Washington, D.C., 84p. (plus appendices).
Bonaparte, R. and Gross, B.A., (1990).“Field Behavior of Double-Liner Systems,” Proceedings
of the Symposium on Waste Containment Systems, ASCE Geotechnical Special Publication
No. 26, San Francisco, CA, pp. 52-83.
Bonaparte, R., (1995). “Long-Term Performance of Landfills,” Proceedings of the ASCE
Specialty Conference Geoenvironment 2000, ASCE Geotechnical Special Publication No. 46,
Vol. 1, pp. 515-553.
Bonaparte, R., Beech, J.F., Griffin, L.M., Phillips, D., Kumthekar, U., and Reising, J., (2008).
“Design and Performance of a Low-Level Radioactive Waste Disposal Facility,”
Proceedings, 6th International Conference on Case Histories in Geotechnical Engineering,
Washington, D.C., 22p.
Bonaparte, R., Beech, J.F., Griffin, L.M., Phillips, D.K., Gross, B.A., Klenzendorf, B., and
O’Leary, L., (2011). “Cold War Legacy – Design, Construction, and Performance of a Land-
Based Radioactive Waste Disposal Facility,” 19th Spencer J. Buchanan Lecture, Texas A&M
University, College Station, TX, 27p.
Bonaparte, R., Gross, B.A., Brown, A., Goulding, L., and Gupta, R., (2016a). “Evaluation of
Liquids Management Data for Double-Lined Mixed Low-Level Radioactive Waste
Landfills”, Geo-Chicago 2016, ASCE Geotechnical Special Publication No. 274,
Geoenvironmental Engineering: Honoring David E. Daniel, pp. 1-22.
Bonaparte, R., Islam, Z.M., Damasceno, V.M., Fountain, S.A., Othman, M.A., and Beech, J.F.,
(2016b). “Geomembrane-Leachate Compatibility for U.S. Department of Energy CERCLA
Waste Disposal Facilities”, Geo-Chicago 2016, ASCE Geotechnical Special Publication No.
275, Geosynthetics, Forging a Path to Bona Fide Engineering Materials: Honoring Robert M.
Koerner, pp. 183-207.
© ASCE
Geo-Congress 2020
Bonaparte, R., Koerner, R.M., and Daniel, D.E., (2002). Assessment and Recommendations for
Improving the Performance of Waste Containment Systems, research report published by the
U.S. Environmental Protection Agency, National Risk Management Research Laboratory,
EPA/600/R-02/099.
Bonaparte, R., Othman, M.A., Rad, N.R., Swan, R.H., and Vander Linde, D.L. (1996).
“Evaluation of Various Aspects of GCL Performance,” Appendix F of Report of 1995
Workshop on Geosynthetic Clay Liners, U.S. Environmental Protection Agency, Office of
Research and Development, Daniel, D.E., and Scranton, H.B., editors, EPA/600/R-96/149,
Jun, F1-F34.
Fernald Environmental Restoration Management Corporation (FERMCO), (1995). “On-Site
Disposal Facility, Detailed Facility Description/Functional Requirements” (draft), Fernald,
OH.
Geosyntec, (1997a). “Design Criteria Package, On-Site Disposal Facility, Rev. 0,” prepared for
U.S. DOE Fernald Environmental Management Project, Fernald, OH.
Geosyntec, (1997b). “Final Design Calculation Package, On-Site Disposal Facility, Rev. 0,”
prepared for U.S. DOE Fernald Environmental Management Project, Fernald, OH.
Geosyntec, (1997c). “Leachate/Liner Compatibility Study Final Report, On-Site Disposal
Facility, Rev. C,” prepared for U.S. DOE Fernald Environmental Management Project,
Fernald, OH.
Giroud, J.P. and Bonaparte, R., (1989a). “Leakage Through Liners Constructed with
Geomembranes. Part I: Geomembrane Liners,” Geotextiles and Geomembranes, Vol. 8, No.
1, pp. 27-67.
Giroud, J.P. and Bonaparte, R., (1989b). “Leakage Through Liners Constructed with
Geomembranes. Part II: Composite Liners,” Geotextiles and Geomembranes, Vol. 8, No. 2,
pp. 77-111.
Gross, B.A., Bonaparte, R., and Giroud, J.P., (1990). “Evaluation of Flow from Landfill Leakage
Detection Layers,” Proceedings, Fourth International Conference on Geotextiles,
Geomembranes and Related Products, Vol. 2, The Hague, pp. 481-486.
Hartung, F. and Scheuerlein, H., (1970). “Design of Overflow Rockfill Dams”, Proceedings of
the Tenth International Conference on Large Dams.
Hooten, G., Hertel, W.A., Glassmeyer, C., and Broberg, K. (2016). “Lessons Learned
Concerning the Onsite Disposal Facility at the Fernald Preserve, Harrison, Ohio.”
Proceedings WM2016 Conference, Phoenix, AZ, 13p.
Koerner, R.M., Lord, A.E., and Hsuan, Y.H., (1992). “Arrhenius Modelling to Predict
Geosynthetic Degradation,” Geotextiles and Geomembranes, Vol. 11, No. 2, pp. 151-184.
Kumthekar, U.A. and Chiou, J.D. (2006). “Lessons Learned from the On-Site Disposal Facility
at Fernald Closure Project,” Waste Management 2011 Conference, Phoenix, AZ.
Parsons, (1995). “Geotechnical Investigation Report, On-Site Disposal Facility, Operable Unit
2,” prepared for Fernald Environmental Management Project, Fernald, OH.
Powell, J.P., Abitz, R.J., Broberg, K.A., Hertel, W.A., and Johnston, F., (2011). “Status and
Performance of the On-Site Disposal Facility Fernald Preserve, Cincinnati, Ohio,” Waste
Management 2011 Conference, Phoenix, AZ.
Rowe, R.K., (2005). “Long-Term Performance of Contaminant Barrier Systems,” 45th Rankine
Lecture, Geotechnique, Vol. 55, No. 9, pp. 631-678.
Temple, D.M., Robinson, K.M., Ahring, R.M., and Davis, A.G., (1987). “Stability Design of
Grass-Lined Open Channels,” Agriculture Handbook Number 667, PB88-116629, U.S.
© ASCE
Geo-Congress 2020
Department of Agriculture, Agriculture Research Service, Beltsville, MD, 167p.

U.S. Department of Energy (DOE), (1989). “Uranium Mill Tailings Remedial Action Project,
Technical Approach Document, Revision II,” UMTRA-DOE/AL 050425.0002, DOE
UMTRA Project Office, Albuquerque, NM.
U.S. Department of Energy (DOE), (1995a) “Final Feasibility Study Report for Operable Unit
2,” Fernald Environmental Management Project, DOE Fernald Area Office, Fernald, OH.
U.S. Department of Energy (DOE), (1995b). “Final Record of Decision for Remedial Actions at
Operable Unit 2,” Fernald Environmental Management Project, DOE Fernald Area Office,
Fernald, OH.
U.S. Department of Energy (DOE), (1995c). “Pre-Design Investigation and Site Selection Report
for the On-Site Disposal Facility,” Fernald Environmental Management Project, DOE
Fernald Area Office, Fernald, OH.
U.S. Department of Energy (DOE), (2016). “Fourth Five-Year Review Report for the Fernald
Preserve,” LMS/FER/S13683, Office of Legacy Management.
U.S. Department of Energy (DOE), (2019). “Fernald Preserve 2018 Site Environmental Report,”
DOE Office of Legacy Management, LMS/FER/S23329.
U.S. Environmental Protection Agency (EPA), (2017). “Post-Closure Performance of Liner
Systems at RCRA Subtitle C Landfills,” EPA/600/R-17/205.
U.S. Nuclear Regulatory Commission (NRC), (1984a) “Radon Attenuation Effectiveness and
Cover Optimization with Moisture Effects (RAECOM),” Computer Program prepared by
Rogers and Associates Engineering, Salt Lake City, Utah, for the U.S. Nuclear Regulatory
Commission, Washington, D.C.
U.S. Nuclear Regulatory Commission (NRC), (1984b). “Radon Attenuation Handbook for
Uranium Mill Tailings Cover System,” NUREG/CR-3533, prepared by Rogers and
Associates Engineering, Salt Lake City, Utah, for the U.S. Nuclear Regulatory Commission,
Washington, D.C.
U.S. Nuclear Regulatory Commission (NRC), (1990). Staff Technical Position: “Design of
Erosion Protection Covers for Stabilization of Uranium Mill Tailings Sites,”.
Vander Linde, D.L., and Beech, J.F., (2000). “Innovative Geomembrane Liner Penetration for
Long-Term Waste Containment,” Proceedings, Waste Tech 2000, Orlando, FL, 14p.
© ASCE
Geo-Congress 2020
Gas Explosion Analysis

Otto D. L. Strack1
1
Dept. of Civil, Environmental, and Geo-Engineering, Univ. of Minnesota, MN. E-mail:
strac001@umn.edu
ABSTRACT
We present the analysis carried out to investigate the cause of a gas explosion that occurred
in Minneapolis. A gas pipeline leaked, resulting in the explosion, and the objective of the
analysis is to determine whether the leak resulted from failure of a water main existing below the
pipeline, or from differential settlements causing the failure of couplers that connected sections
of the pipeline. An analysis based on the theory for a beam on an elastic foundation makes the
differential settlements the most likely cause of the explosion.
INTRODUCTION
A gas explosion occurred on March 17, 2011, on 60th Street and vacated 1st Avenue South
in Minneapolis, MN. The objective is to investigate possible reasons for the explosion. Accidents
such as this one are potentially very dangerous, and it is imperative that an objective analysis be
carried out and a serious attempt be made to investigate all possible causes for the accident. This
paper is based on an analysis carried out by Strack [2014] at the request of the City of
Minneapolis. The gas pipeline was constructed to pass at right angles over an existing storm
sewer, constructed in 1941, as described in of Minneapolis [2012], and was only inches above
the storm sewer. A water main also existed near the gas pipeline, a somewhat poor black and
white photograph reproduction shows the gas pipeline after the explosion and the existing water
main, see Figure 1. Photographs of the gas pipeline are shown in Figure 2.
Figure 1: Proximity of the water main to the gas pipeline, shown after the explosion.
An explanation put forward by a consultant in Minneapolis is that the explosion was caused
by a washout below the gas pipeline, due to a failing water main, see Consultants [2011a,b,
© ASCE
Geo-Congress 2020
Figure 4: Reconstruction of the bend as it was prior to the explosion
Figure 5: Sketch of bended gas pipeline over the City sewer

The presence of organic soils below the gas pipeline and the construction of a bend over an
existing storm sewer is likely to lead to uneven settlements. Since it is not possible to compact
properly the soil above and around an existing sewer and near a water main, the soil supporting
the gas pipeline will be loose, and the soil between the gas pipeline and the sewer and the soil
next to the sewer is unlikely to provide any support at all.
The storm sewer is in the middle of the upward bend and we expect the problem to be sym-
metrical with respect to the storm sewer. It is thus likely that much, or all, of the weight of the
section between centerline of the sewer and the Dresser coupling labelled DC1, combined with
the weight of the soil above, results in a downward force that acts on the rightmost section of the
pipe, the force labeled P in Figure 5. The Dresser couplings in the gas pipeline will not be
capable of carrying the same bending moment as the intact sections of pipe; possible change in
angle between the sections connected at the coupler will have to be withstood by the coupler.
We analyze settlements of the gas pipeline using the theory of a beam on a Winkler
foundation, in order to estimate the possible deformation that the couplers would have to
withstand, based on the assumption that the resistance of the soil on which the beam, in this case
the pipe, is placed, is proportional to the vertical displacement. This is a commonly accepted
approach for the design of beams on a soil foundation, and is taught in foundation engineering
courses at universities. A full description of the approach is found in texts on soil and rock
mechanics, for example, Verruijt [1995].
In order to estimate the differential settlements of the gas pipeline as a result of the
construction of the bend over the sewer, combined with the lack of support in that area, we carry
out two different analyses, both concerning the rightmost section of pipe. The first one considers
the pipe as a single intact beam, loaded by the force P at the leftmost end. The analysis makes it
possible to compute the bending moments that the Dresser coupling DC2 in Figure 5 has to
withstand. The second analysis is based on the assumption that the couplers act as hinges that
cannot support any bending moment at all, and allow the central portion of the bend to press
© ASCE
Geo-Congress 2020
down on the lower section of the pipe, i.e., the section to the right of the force P in Figure 5. The
downward displacement at the leftmost point of the short section between the Dresser couplings,
DC1 and DC2 in Figure 5 predicted by the analysis will give an indication of the movement that
the coupler will have to withstand. The two analyses are based on two extreme viewpoints: one
that the coupler is fully intact, and the other that the coupler acts as a hinge, thereby bracketing
the predictions -we expect the truth to be between the two extremes. This is a common approach
in engineering practice and gives much insight in the possible explanation of what truly
happened.
The simplified problem is sketched in Figure 6 and shows the short section of pipe of length
L, see Figure 5, attached by Dresser coupling DC2 to the rightmost section of pipe, which we
approximate as being semi-infinite. The distance from the center of the storm sewer to the center
of the upward bend of the pipe is b, the distance from the center of the bend to the first coupler,
DCI is a, and the distance from the first to the second coupler is L.
Figure 6: Beam consisting of two sections on a Winkler foundation
Table 1: Typical values of modulus of subgrade reaction (ks) for different types of soils
Table taken from the Structural Engineering Forum of India, www.sefindia.org
Type of Soil ks(kN/m3)
Loose sand 4,800-16,000
Medium dense sand 9,600-80,000
Dense sand 64,000-
128,000
Clayey medium dense 32,000-80,000
sand
Silty medium dense sand 24,000-48,000
The various lengths and data required for the computations are given below. The gas pipeline
weight is taken from the API 5L specifications (pages 50 and 51) with the wall thickness given
as inches. The weight from the table in API 5L is 52.8 pounds per foot. The length L shown in
Figure 6 is about 4 times the diameter of the pipe, measured from Figure 2. About the same
length is dimension b in Figure 5. The soil is dry, and we estimate the specific weight of such
soil at 1800 N/m3. We took a value for the modulus of sub grade reaction, k, from the table
shown in Table1, taken from the Structural Engineering Forum of India. We have chosen a value
© ASCE
Geo-Congress 2020
of k = 5, 000 kN/m3 or k = 5 ∗ 106 N/m3, reflecting the weak and poorly compacted soil. We
further reduced the estimated modulus by a factor of 2, taking account of the reduction in
effective surface; the pipe is cylindrical, rather than square, in section and the soil will only
provide effective resistance over the portion of the cylinder where the angle between the normal
of the cylinder and the vertical is less than 30 degrees, the estimated angle of internal friction of
the soil. The results of the computation are as follows:
Approach 1: intact beam

 Settlement at leftmost point: 3.6 cm

 Settlement at Dresser coupling number 2: 1.4 cm
 Moment at Dresser coupling number 2 in magnitude: 5.9 104 Nm=59 kNm
Approach 2: two-part beam
 Settlement at leftmost point: 8.3 cm

 Settlement at Dresser coupling number 2: -1.25 cm
Figure 7: Settlements of the gas pipeline as predicted by the analysis based on the Winkler
foundation. Note the upward movement of the joint at the second Dresser coupling.
Note that the analysis predicts an upward movement of the second Dresser coupling. This is
remarkable, because the photographs taken after the accident indeed suggest that such an upward
movement occurred. It is also worth noting that the analysis does not apply to cases where
upward movement occurs; upward movement results in a downward reactive force according to
the theory, which the soil cannot exert. Thus, the actual upward movement is likely to be larger:
the solution provides an estimate of the settlements that is slightly on the low side. It should be
noted that it is not possible, after the explosion, to obtain the data necessary to state without
doubt that the construction of the gas pipeline has caused the accident. However, the analysis
© ASCE
Geo-Congress 2020
does show that it is a likely explanation.

The geotechnical analysis shows that differential settlements as a cause of the gas explosion
are likely; predictions of settlements are in a range of values that support the assumption that the
construction of a bend in the gas pipeline over the storm sewer is the most likely explanation of
the cause of the accident. This explanation fits well with the failure of the Dresser coupling,
which cannot be expected to withstand the forces it was subjected to, it also fits with the upward
movement of the second Dresser coupling that is observed from the photographs. Furthermore, a
gas leak was reported on October 23, 2008 (see City of Minneapolis Public Works Utilities
Infrastructure Impact, March 17, 2011 Gas Explosion 60th E at Vacated 1st Ave S., page 4)
which fits with a gradual deformation of the gas pipeline as a result of the design. The change in
angle between the two sections of the gas pipeline resulting from the differential settlement is
about 2.7 degrees. We expect that this change in angle is sufficient to cause leakage of gas into
the ambient soil in sufficient quantities that an explosion may result.
CONCLUSIONS
The gas pipeline is constructed near the City water main, which was installed in 1948. The
soils that form the foundation of the gas pipeline are soft, contain peats, and are sufficiently
prone to settlements that the City of Minneapolis supported its sewer on piles, see Figure 8,
which shows a quote from of Minneapolis [2012] that makes this quite clear. The close
proximity of the gas pipeline to the water main and other utilities makes it unlikely that the soil
below the gas pipeline, soft to begin with, was compacted properly. The decision to run the gas
pipeline over a large sewer (44 inches in outer diameter) resulted in a construction that is likely
to cause settlements of the gas pipeline. The analysis supports the theory that the design caused
differential settlements and bending moments in the gas pipeline which led to a gap opening in
the second Dresser coupling that enabled gas to escape. A reported gas leak supports this
assumption. The movement of gas through a porous medium in partially saturated condition is
extremely difficult to predict and the appearance of such a leak can be erratic. The combined set
of conditions and observations make it highly unlikely that the explosion was caused by the
water main, either by a spontaneous burst, or by slow deterioration. The latter would be expected
to cause increasingly frequent failures and complaints from water users, which are not in
evidence.
Figure 8: Quote from of Minneapolis [2012] indicating the weak soils in the area of the
accident
APPENDIX: COMPUTATION OF THE SETTLEMENTS

The differential equations that govern the settlement of a beam on a Winkler foundation are
as follows:
d 4w
EI 4  kw  q (1)
dx
© ASCE
Geo-Congress 2020
d 2w
M   EI (2)
dx 2
d 3w
V   EI 3 (3)
dx
where
k  k0 * b  N / m2  (4)
3
k0 is the modulus of sub-grade reaction [N/m ] and b is the width of the beam. M is the bending
moment in the beam [Nm], V is the shear force [N], and q is the unit weight of the beam per unit
length [N/m]. The theory is valid for a square beam of constant cross-sectional area, but is
applied in this case to a cylindrical pipe. Since the resistance to settlement will be reduced due to
the shape, the effective width of the beam, i.e., the width that will generate full resistance of the
soil, is assumed to be half the cross-section of the beam, or equal to the radius.
The signs are as indicated in Figure 9. The term q/k represents displacement due to a constant
weight q per unit area [N/m2]. The general solution to the differential equation (1) is
Figure 9: Sign convention for Winkler foundation analysis and boundary conditions
w  e x C1 sin   x   C2 cos   x   e  x  B1 sin   x   B2 cos   x  (5)

The solution
q
w (6)
k
is a particular solution to the differential equation; for constant q and k, (6) satisfies the
differential equation. We introduce a constant β with the dimension [1/m] as
1
 k 4
  (7)
 4 EI 
where EI is the bending resistance of the beam; E [N/m2] is the modulus of elasticity and I the
area moment [m4]. We differentiate (5) three times to obtain expressions for M and V . This
gives the following results
dw
  e  x  C1  C2  sin   x    C1  C2  cos   x  
dx (8)
  e  B1  B2  sin   x     B1  B2  cos   x  
 x
© ASCE
Geo-Congress 2020
The second derivative is

d 2w
dx 2  
 2 2 e x  C2 sin   x   C1 cos   x   e  x  B2 sin   x   B1 cos   x  (9)
The third derivative is
d 3w
dx 3 
 2 3 e  x    C1  C2  sin   x    C1  C2  cos   x  
(10)
e  x

 B1  B2  sin   x    B1  B2  cos   x  
Verification that the solution meets the differential equation

Differentiation of (10) will give an expression for the fourth derivative of w
d 4w
 2 4 {e  x [(C1  C2 ) sin(  x)  (C1  C2 ) cos(  x)
dx 4
(C1  C2 ) cos(  x)  (C1  C2 ) sin(  x)] (11)
 x
e [( B1  B2 ) sin(  x)  ( B1  B2 ) cos(  x)]
( B1  B2 ) cos(  x)  ( B1  B2 ) sin(  x)]}
Combining terms gives
d 4w
 4 4{e  x [C1 sin(  x)  C2 cos(  x)]
dx (12)
 x
e [ B1 sin(  x)  B2 cos(  x)]}
It follows from (7) that
k
4 4  (13)
EI
so that we can write (12) with (5) so that
d 4w k
4
 w (14)
dx EI
or
d 4w
EI 4  kw  0 (15)
dx
We can add the particular solution ω = qo/k, which is a constant and clearly satisfies (1) so
that
d 4w
EI 4  kw  q0 (16)
dx
The solution presented above thus satisfies the differential equation.
Initial Conditions for semi-infinite beam

The initial conditions for the case of the semi-infinite beam are as follows
M  0,V   P x  0 (17)
The condition at infinity is that the displacement is finite; the term with the positive
exponential in (5) must vanish, i.e.,
C1  C2  0 (18)
© ASCE
Geo-Congress 2020
Application of the first condition in (17) to (2), with (9) and (18) gives
B1  0 (19)
The final constant, B2, is obtained from the second condition in (17) with (3) and (10)
P  2 3 EIB2 (20)
It follows from (7) that
1
EI  k  4 (21)
4
So that we obtain from (20)

1k 
P B2  B2  2 P (22)
2 k
The statement for this case are given by
 q
w  2 Pe  x cos   x   0 (23)
k k
The expression for moment M is
1 P
M  4 3 PEIe  x cos   x    e  x cos   x  (24)
k 
Initial Conditions for the case of the two-piece beam

The initial conditions are as follows
M  0 x  L (25)
V  P x  L (26)
V  V  x  0 (27)
w  w x  0 (28)
M  0 x 0 (29)
q
w x (30)
k
The first two conditions express that there will be zero moment and a load P at the left end of
the section of beam of length L. The third and fourth conditions express that the shear force must
be continuous as there is no load at the connection at x = 0 and that the displacement must be
continuous. The next to last condition expresses that no moment is transferred at the hinge. The
final condition expresses that the displacement far away is due only to the weight of the beam.
We solve for the four constants in the solution as follows. The solution for the semi-infinite
portion of the beam, running from x = 0 to x = cannot contain the positive exponentials, so that
this solution can be written as
w  e  x b1 sin   x   b2 cos   x  x  0 (31)
The expression for the moment is obtained from (9) and (2)
M  2 2 EI e  x b2 sin   x   b1 cos   x  x0 (32)
The condition that M = 0 at x = 0 gives
b1  0 (33)
The expression for the shear force V is, with (10) and (3)
V  2 3 EIe  xb2 cos   x  (34)
© ASCE
Geo-Congress 2020
Or, with (7)

k
V  2b2 e  x cos   x  (35)

so that the force P0 = -V at x = 0 becomes
k
V   P0  2b2 (36)

So that
2 P0
b2  (37)
k
The displacement at point x = 0 is
2 P0
w  w0  b2  (38)
k
This displacement and the force P0 must match the values obtained from the section of beam
between x = -L and x = 0. The condition that the moment is zero at x = 0 gives, with (9)
B1  C1 (39)
The moment must also be zero at x = -L which gives, again with (9) and (2)
e  L C2 sin   L   C1 cos   L   e L  B2 sin   L   B1 cos   L   0 (40)
Or, with C1 = B1
B1 cos   L  e L  e  L   e L B2 sin   L   C2e  L sin   L  (41)
So that
tan   L 
B1  C1  C2e  L  B2e  L  (42)
2sinh   L 
The shear force at x = 0 is
d 3w k k
V   P0   EI  C1  C2  B1  B2     2C1  C2  B2  (43)
dx 3
2 2
We obtain the latter two equations to obtain
k  tan   L  
P0   [C2e  L  B2e  L ]  B2  C2  (44)
2  sinh   L  
or
k   tan   L    L   tan   L   L  
P0   C2  e  1  B2  e  1  (45)
2   sinh   L    sinh   L  

The displacement at x = 0 given by (38) must match that obtained from the short beam at x =
0, i.e., expression (5) evaluated at x = 0
2 P0  tan   L    L   tan   L    L 
w  C2  e  1  B2  e  1  C2  B2 (46)
k  sinh   L    sinh   L  
or
 tan   L    L  tan   L   L
C2  e  2  B2 e (47)
 sinh   L   sinh   L 
We express B2 in terms of C2
© ASCE
Geo-Congress 2020
 sinh   L    L 
B2  e2 L  2 e  C2 (48)
 tan   L  
We express the constants C1 and B1 in terms of C2 using (42) which gives
tan   L     L   L sinh   L  
B1  C1  e  e  2  C2  C2 (49)
2sinh   L   tan   L  
We have expressed all constants in terms of the single unknown constant C2. The shear force
is given at x = -L which gives, with B1 = C1 = C2

d 3w k  L
V  P  EI 3  {e 2C2 sin   L 
dx 2 (50)
e  L [C2 [cos   L   sin   L ]  B2 [cos   L   sin   L ]
We use (48) to eliminate B2 from this equation, which gives
P
k
2
2e  L sin   L   e  L [cos   L   sin   L ]
(51)
sinh   L  
e [cos   L   sin   L ]  2
L
[cos   L   sin   L ] C2
tan   L  
This expression can be solved for the final unknown, C2.
REFERENCES
Consultants. 60 th street incident preliminary report. Technical report, Crane Engineering,
August 2011a.
Consultants. 60 th street incident supplemental report. Technical report, Crane Engineering,
December 2011b.
Consultants. Rebuttal to mnops pipeline incident report dated october 12, 2012. Technical report,
Crane Engineering, November 2012.
City of Minneapolis. Public works utilities infrastructure impact march 17, 2011 gas explosion.
Technical report, City of Minneapolis, 2012.
O.D.L. Strack. Settlement of gas pipeline due to uneven loading. Technical report, Strack Con-
sulting, LLC, 2014.
A. Verruijt. Computational Geomechanics. Kluwer Academic Publishers, Dordrecht, The
Nether- lands, 1995.
© ASCE
Geo-Congress 2020
The Use of Steel Sheet Pile as Permanent Building Foundation Walls: Lessons Learned
over 15 Years of Design in Minneapolis, Minnesota, USA
Chad A. Underwood, P.E., P.G., D.GE, M.ASCE 1
1
Engineering Partners International LLC, Richfield, MN. E-mail: chad@epillc.net
ABSTRACT
Engineering Partners International LLC (Engineering Partners) has designed over 30 projects
where steel sheet pile walls have been used as permanent perimeter foundations to support
buildings with one to three levels of below grade parking and up to 16 above grade floor levels.
Similar to conventional foundation walls, the sheet pile walls provide lateral support for soil and
surcharge loads, and vertical support for below grade parking slabs and perimeter walls and
columns of the superstructure. The sheet pile foundations have been designed to support uniform
wall loads up to 12 kips per linear foot and point loads up to 380 kips. When Engineering
Partners completed their first sheet pile foundation design in 2005, the use of sheet pile
foundations together with the construction sequencing described in this paper was among the
first of its kind in the United States. This paper summarizes projects completed in the 15 years
since, including results of several high strain dynamic tests completed on sheet pile using a pile
driving analyzer (PDA). The majority of these projects have been completed in and around
Minneapolis, Minnesota.
INTRODUCTION
Steel sheet pile has been used for temporary and permanent earth retention walls and water
retaining structures since the late 1800’s. Several sheet pile shapes have been developed to serve
different functions for earth and water retention (e.g., flat sheets, U-shaped sheet, and Z-shaped
sheets). Z-shaped sheet pile sections have become the favored sheet pile section for earth
retention walls due to their high bending resistance relative to weight of steel (i.e., a structurally
efficient section). The sheet pile sections discussed in this paper are all of the Z-shape type, most
of which belonging to the AZ series of sheet pile.
Although sheet pile has been historically used to resist horizontal soil and water pressures in
earth and water retaining structures, there is extensive data on the ability for sheet pile to resist
vertical loads. For example, tieback anchors used as lateral support for sheet pile earth retention
walls are typically installed at angles of 10 to 40 degrees from horizontal, which imparts both
horizontal and vertical components of force to the sheet pile wall. For high capacity, steeply
inclined tieback anchors, significant vertical forces are transferred to the sheet pile. Sheet pile
used for earth retention applications often resist this vertical component of tieback force through
skin friction alone. The relatively large surface area of sheet pile in contact with soil below the
bottom of excavation elevation can yield high skin friction resistance.
The ability for sheet pile to resist horizontal soil and water pressures through a structurally
efficient Z-shaped profile, together with the ability to resist vertical loads through soil-to-steel
interface friction over a relatively large surface area, makes sheet pile a candidate for use as
permanent building foundation walls. Although other types of earth retention walls such as
tangent or secant pile walls and slurry diaphragm walls have lateral structural resistance
characteristics and vertical geotechnical resistance characteristics similar to sheet pile, this paper
focuses on the use of steel sheet pile as permanent foundation walls for building applications.
© ASCE
Geo-Congress 2020
When sheet pile is used as a permanent building foundation wall, the sheet pile serves as both
the temporary earth retention system and the permanent foundation wall. Substantial time and
cost savings can be realized by incorporating the temporary earth retention system into the
permanent building foundation wall by eliminating the need for two separate wall systems.
Temporary retention walls are typically offset 4 to 6 feet from the building foundation walls to
allow forming and pouring conventional cast-in-place walls or constructing masonry foundation
walls. Using the permanent building foundation wall as the temporary earth retention system can,
in some cases, allow the building footprint to be expanded outward to maximize the building
footprint on a site. This can increase the revenue generating capacity of the structure, whether
from increased commercial/retail space, residential space, parking stalls, etc.
Installing an earth retention system that also serves as the perimeter building foundation wall
can permit building construction to proceed in a “top-down” (or “up-down”) manner. With top-
down construction, the building floor slabs for the below grade levels are used to resist lateral
earth pressures and can therefore eliminate or reduce the need for temporary tieback anchors for
deep excavations. Top-down construction also allows construction of above grade levels of a
building to start while below grade levels are still being constructed, which can shorten the
construction schedule.
SHEET PILE FOUNDATIONS IN MINNESOTA

The sheet pile foundation projects that Engineering Partners has designed in Minnesota have
included buildings with one to three levels of below grade parking and up to 16 above grade
floor levels. For each project, both the below grade parking slabs and the above grade columns
and/or perimeter walls of the superstructure are supported by the sheet pile foundations. For most
of the projects, the sheet pile foundations were designed as friction piles (predominantly in
sand); a few were designed as end bearing piles (bearing on relatively shallow limestone
bedrock).
The interior structural framing of the buildings (Level 1 and below) for these projects has
either been precast concrete columns, beams, and plank, or cast-in-place concrete columns and
post-tensioned (PT) slabs. The type of structural framing system can dictate the approach to
excavation and construction of footings adjacent to the perimeter sheet pile walls. Projects that
utilize precast concrete framing rely on precast beam and plank support on the perimeter sheet
pile walls, with the first row of interior columns typically located 15 to 20 feet from the
perimeter walls. The location of the interior columns in a precast structure often allows a
temporary soil berm to remain in place against the sheet pile walls during interior footing
construction and installation of precast columns and beams (Figure 1). The use of a temporary
berm against the sheet pile is a form of top-down construction and can eliminate the need for
tieback anchors for excavations of up to one to two levels of below grade (Underwood and
Greenlee, 2010).
Buildings that utilize PT slabs typically require columns located just inside the perimeter
sheet pile walls for slab support. Because the interior columns are located directly adjacent to the
sheet pile, it is not practical to leave a temporary soil berm against the sheet pile as with a precast
structure. For PT structures, temporary tieback anchors have been used for buildings with more
than one level below grade (Figure 2).
The soil conditions on most of the project sites presented in this paper generally consist of
granular outwash and alluvial soils and granular glacial till. The subsurface conditions on each
site were primarily characterized using Standard Penetration Test (SPT) borings. Field or
© ASCE
Geo-Congress 2020
laboratory testing was generally not completed to define soil strength parameters prior to design
for these projects; therefore, soil strength parameters were estimated based on SPT values (N-
values). Groundwater was present at or slightly below the lowest floor slab elevation on nine of
the 30 projects; groundwater was significantly deeper on the remaining 21 projects.
Figure 1. Permanent sheet pile foundation with a temporary soil berm for construction of
below grade precast concrete framing.
Figure 2. Permanent sheet pile foundation wall with temporary tieback anchors for
construction of cast-in-place concrete columns for a post-tensioned (PT) slab structure.
© ASCE
Geo-Congress 2020
AXIAL SHEET PILE CAPACITY EVALUATION

Evaluation of the capacity for sheet pile foundations to resist vertical loads is based on
conventional analyses for piles (Terzaghi and Peck, 1996; USACE, 1991) as shown in Equations
1 through 3.
Qult  Qs  Qt (1)
Qs  fs • As (2)
Qt  q • A t
(3)
where:
Qult = ultimate pile capacity
fs = average unit skin friction
Qs = shaft skin friction resistance
As, At = pile surface area, pile tip area
Qt = end bearing tip resistance
q = unit tip bearing capacity
For evaluation of sheet pile that derive their capacity primarily from shaft resistance, unit
skin friction values (fs) can be estimated from field load tests (i.e., static or dynamic load tests) or
by estimating the wall interface friction angle and in situ horizontal stresses (Equation 4).
fs  h • tan  i  (4)
where:
σh = in situ horizontal stress
ϕi = sheet pile/soil interface friction angle
When incorporating test results of shaft resistance into design, one must recognize that the
horizontal stresses acting on the sheet pile during testing are different than the horizontal stress
conditions present during the service life of the structure (Underwood and Greenlee, 2011).
The relatively large surface area of a continuous sheet pile wall can yield high capacities
from skin friction alone; tip resistance from end bearing is generally neglected for conservatism
unless the pile bears on competent bedrock. Vertical loads acting on the sheet pile are analyzed
on a per lineal foot basis. The load per lineal foot applied over a certain vertical embedment
depth of sheet pile results in a load per square foot of embedded pile that must be resisted by
sheet pile-to-soil interface friction. For uniform loads from walls and slabs, this allows a direct
comparison to the unit skin friction resistance estimates described above. Where heavy point
loads from columns or beams act on the sheet pile, the point loads are distributed to two or more
sheet piles to provide an equivalent uniform loading condition. The load distribution to two or
more sheet piles can occur through load transfer beams and/or by welding interlocks of adjacent
sheet pile together.
FIELD VERIFICATION OF AXIAL SHEET PILE CAPACITY

Axial capacity of sheet pile foundations can be verified in the field using static load testing or
high strain dynamic testing with a Pile Driving Analyzer (PDA). Field verification of axial sheet
pile capacity has been completed using PDA testing for the projects that Engineering Partners
has been involved with. Although sheet pile is typically installed using a vibratory hammer, PDA
testing requires the use of an impact hammer. As a result, test piles are typically driven to a
prescribed depth using a vibratory hammer before switching to an impact hammer for testing.
CAPWAP (Case Pile Wave Analysis Program) analyses are performed on the PDA test data
© ASCE
Geo-Congress 2020
to estimate the total bearing capacity of the pile (shaft and tip resistance), as well as the
distribution of skin friction resistance along the pile shaft. As discussed previously, sheet pile tip
resistance is often neglected in the design of axially loaded sheet pile for conservatism.
Therefore, the discussion of field verification tests will focus on unit skin friction measurements.
It should be noted that performing and interpreting PDA tests on sheet pile sections differs from
conventional testing of H-pile or pipe pile foundations. It is therefore critical to have an
experienced testing agency perform these tests so that reliable data can be collected for
verification of sheet pile capacity.
Approximately 35 PDA tests have been completed on the 30 projects that Engineering
Partners has designed in the past 15 years. Depending on the size of the project and the
variability of soil conditions on the site, one to eight PDA tests have been completed per site.
Sites where sheet pile bears directly on bedrock often did not include a field verification test
program. Most of the project sites have been in sand and silty sand, so the discussion of
measured unit skin friction values focuses on granular soils.
Although the number of PDA tests completed on these projects has resulted in a considerable
database of ultimate unit skin friction values, it has been difficult to draw conclusive
relationships between unit skin friction and commonly available data from Standard Penetration
Test (SPT) borings, such as N-values. It is intuitive that the unit skin friction values would
increase with increased relative density of granular soils, and this general relationship has been
confirmed through PDA testing. Observations during sheet pile installation have suggested that
waterbearing granular soils may provide lower skin friction resistance compared to soils of
similar relative density above the groundwater table. However, test results indicate similar
ultimate skin friction values both above and below the groundwater table for similar soil types.
Table 1. Summary of ultimate skin friction values for various relative densities of granular
soils estimated from PDA test results.
Soil Type Granular
Very
Relative Density Loose Medium Dense Dense
Loose
(N-Values) 0-4 5-10 11-20 21-30 31-40 41-50
Groundwater (1) N Y N Y N Y N Y N Y N Y
Ultimate Min. 0.09 0.15 0.08 0.11 0.10 0.12 0.10 0.24 0.35 - - 0.34 - -
Unit Max. 0.34 0.34 0.57 0.59 0.71 0.74 0.77 0.47 1.07 - - 1.34 - -
Skin
Range 0.25 0.19 0.49 0.48 0.61 0.62 0.67 0.23 0.72 - - 1.00 - -
Friction
(ksf)
Avg. 0.20 0.21 0.24 0.24 0.35 0.36 0.39 0.37 0.59 - - 0.97 - -
Footnotes: (1) "N" = data from soil profile above the estimated groundwater elevation. "Y" = data from soil profile
below the estimated groundwater elevation.
Ultimate unit skin friction values of approximately 0.1 to 1.3 kips per square foot (ksf) have
been estimated based on the PDA test results. Summaries of ultimate skin friction values for
various relative densities of granular soils are provided in Table 1 and Figure 3. These
summaries were produced after eliminating outlier data points from the measured skin friction
values obtained from the PDA tests.
The ultimate skin friction values determined from PDA tests completed on the project sites in
Minnesota are slightly lower than the range of values developed from full-scale static load tests
© ASCE
Geo-Congress 2020
and pull-out tests on instrumented sheet pile wall elements and box piles completed by
Bustamante and Gianeselli (1991). Bustamante and Gianeselli (1991) found that the ultimate
skin friction values measured from static load tests on sheet pile wall elements ranged from
approximately 0 to 0.2 ksf for very loose sand, 0.2 to 0.4 ksf for loose sand, 0.8 to 1.2 ksf for
medium dense sand, and 0.8 to 1.6 ksf for dense sand. The use of higher skin friction values (i.e.,
greater than 0.8 ksf) for medium dense to dense sand are dependent on limit pressures
determined through pressuremeter testing (Bustamante and Gianeselli, 1991). The slightly higher
skin friction values reported by Bustamante and Gianeselli (1991) may be indicative of
differences between dynamic and static load test results, which warrant additional investigation.
Figure 3. Average ultimate skin friction values versus relative density of granular soils
estimated from PDA test results.
Although the PDA test data has not revealed significant differences in skin friction values for
granular soils above and below the groundwater table (Table 1 and Figure 3), there appears to be
some differences when the measured skin friction values are normalized for effective vertical
stress. The data suggests that normalized ultimate skin friction values for waterbearing granular
soils may be slightly lower than soils of the same relative density above the groundwater table.
However, the data is inconclusive and additional data is required before drawing firm
conclusions on the differences in skin friction values for granular soils above and below the
groundwater table.
INSTALLATION TOLERANCES
Sheet pile used as permanent building foundation walls require significantly more stringent
installation tolerances than sheet pile used for temporary applications. Above grade walls and
columns supported at the top of the sheet pile walls can produce significant eccentric loads at the
© ASCE
Geo-Congress 2020
top of sheet pile if walls are installed out of tolerance. An allowable installation tolerance of plus
or minus 2 inches inward or outward is often specified. In some cases, a 0-inch inward tolerance
is required where interior columns are constructed immediately adjacent to the sheet pile walls.
Where the sheet pile walls are installed directly adjacent to a property line, a 0-inch outward
tolerance may be required. The project design team must understand that sheet pile installation
requires some inward or outward tolerance and that a zero inward and outward installation
tolerance is not a practical requirement.
For precast concrete structures, precast-to-sheet pile connections must be designed to
accommodate sheet pile installation tolerances. For example, precast plank is manufactured to a
specified length, often before the sheet pile installation is complete. Sheet pile walls that move
too far inward during installation could require costly field trimming of precast plank. PT
structures often have more forgiveness to sheet pile installation tolerances because the slabs are
cast after the sheet pile is installed.
The use of driving templates is critical to maintain stringent installation tolerances for
permanent sheet pile foundation walls. In addition to taking measures to control inward and
outward sheet pile tolerances, it is important that the contractor monitor in-plane sheet pile
locations during installation. Because of the interlocking nature of the sheet pile, the laying
length of sheet pile can vary as it is installed, sometimes stretching or compressing due to slight
rotations of the interlocks during installation. This stretching or compressing action can be more
prevalent in difficult driving conditions. The plan layout of sheet pile walls used for load bearing
foundation elements often include heavier / longer sheet pile sections at higher point load
conditions. It is critical that the locations of heavier / longer sheet pile sections be installed at the
specified locations in order to accommodate the higher point loads. Failure to monitor in-plane
sheet pile locations can result in mis-location of critical sheet pile sections along the wall
alignment.
VIBRATIONS
Because sheet pile foundation systems are typically used in dense urban environments where
space limitations prohibit the use of a temporary earth retention system for construction of
conventional concrete or masonry foundation walls, there are often nearby vibration sensitive
structures and utilities that have to be considered. Sheet pile is typically installed using vibratory
hammers. Vibratory hammers have traditionally consisted of a cable suspended hammer operated
on a crawler crane. This installation method relies on the weight of the hammer plus vibration to
install the sheet pile. The use of dedicated fixed-mast piling rigs has become more common in
recent years. These rigs allow the application of crowd to further increase the downward force
applied during installation, allowing sheet pile installation in denser/stiffer soils.
The proximity of vibration sensitive structures and utilities often necessitates the use of
variable moment vibratory hammers. These hammers use high frequency vibrations during start
up, driving, and shut down. Higher frequency vibrations result in less soil disturbance and limit
the distance to which vibrations are transmitted. Non-vibratory hydraulic presses are also
available for sheet pile installation but are not commonly available or used in Minnesota.
It is important to complete pre- and post-construction condition surveys of adjacent structures
and utilities and develop a carefully thought out vibration monitoring plan to assess the potential
impacts of vibrations on adjacent structures. It is often beneficial to begin sheet pile installation
farther away from vibration sensitive structures and work towards the structures while
monitoring vibrations. This allows the contractor to react to the vibration monitoring data and
© ASCE
Geo-Congress 2020
modify installation methods as required to reduce vibrations. Some of the projects that
Engineering Partners has been involved with have switched from sheet pile foundation walls to
drilled pile walls, such as soldier pile and lagging walls or secant pile walls, where wall
installation is close to vibration sensitive structures.
ADDITIONAL DESIGN CONSIDERATIONS

The sheet pile foundation projects that Engineering Partners has designed have all been
delegated design components of the building using a performance-based specification. With a

delegated foundation design component, close coordination with both the structural engineer of
record and the project geotechnical engineer is critical. As indicated previously, several of the
projects have utilized a precast concrete framing system for the interior below grade structure.
The design of precast concrete structural elements is also typically a delegated design component
completed by a separate specialty engineer; this requires an additional level of coordination with
the sheet pile foundation design.
Buildings that utilize perimeter sheet pile foundation walls (which is considered a type of
deep foundation) often have interior columns supported on spread footings. The use of both deep
and shallow foundations within the same structure requires consideration of differential
settlement. The expected differential settlement must be within the tolerable limits of the
structural framing system, which is typically on the order of ½ inch. Settlement of the sheet pile
foundation walls is generally minimal, therefore the design of spread footings for interior
columns may require special considerations to reduce footing settlement to maintain tolerable
differential settlement. This requires close design coordination with the project geotechnical
engineer.
Where differential settlement of interior shallow foundations could exceed allowable limits,
delayed connections between the interior floor systems and the perimeter sheet pile foundation
walls may be required. With a delayed connection approach, the upper floor levels of the
building are constructed, adding dead load to the interior column footings and allowing some
settlement to occur before final connections between the structure and the sheet pile foundation
walls are completed. This approach requires close coordination between both the project
structural engineer of record and the geotechnical engineer.
The structural capacity of the sheet pile considers combined axial loading and/or eccentric
loading due to uniform loads and point loads acting on the sheet pile wall, together with bending
moments due to the lateral earth pressure. Additional discussion regarding evaluation of the
structural capacity of sheet pile foundation walls can be found in Underwood and Greenlee
(2011).
As with most basement walls, waterproofing is an important design consideration for sheet
pile foundation walls. Because there is no excavation outside the sheet pile foundation walls,
application of conventional exterior foundation wall waterproofing systems is not feasible. Most
of the projects that Engineering Partners has been involved with have had lowest floor elevations
above seasonal high groundwater elevations. In these situations, the project architect has
typically not specified any special waterproofing of the sheet pile walls. Instead, the relatively
watertight sheet pile interlock (Abbondanza, 2009) is relied on for water seepage control. Where
additional resistance to water seepage is required, a non-structural seal weld can be applied to the
sheet pile interlocks.
Because the sheet pile walls are used to carry vertical building loads, durability and fire
rating are also important aspects of the foundation design. Detailed discussion of these aspects of
© ASCE
Geo-Congress 2020
sheet pile foundation design is beyond the scope of this paper, but additional discussion can be
found in Abbondanza (2009) and Underwood and Greenlee (2010). Sheet pile foundations are
considered both a deep foundation and a structural load carrying element of the building (i.e.,
structural wall and/or column). Therefore, special inspections are required in accordance with the
International Building Code (IBC). These inspections include documentation of pile installation
(pile material, type, and length), inspection of welded and bolted connections, etc.
Sheet pile walls provide a visual profile distinctly different from concrete or masonry walls
and can provide a unique architectural feature to the building. Sheet pile foundation walls are
often painted (Figure 4) for aesthetics and corrosion protection inside the building.
Figure 4. Finished sheet pile foundation wall in a below grade parking garage.
CONCLUSION
Steel sheet pile can be a cost-effective foundation alternative where it is necessary to
construct one or more below grade levels with building foundation walls close to property lines.
Sheet pile foundations allow the use of top-down construction (or a modified version of top-
down construction), which can eliminate or reduce temporary tiebacks or temporary bracing of
the sheet pile walls during construction of below grade levels. This can reduce the overall
construction schedule (and cost) by allowing construction of above grade levels to begin before
below grade levels are completed.
Sheet pile walls can develop relatively high geotechnical capacities to resist axial loads.
Ultimate unit skin friction values of approximately 0.1 to 1.3 kips per square foot (ksf) have been
estimated based on PDA tests completed on multiple projects in Minnesota. Unit skin friction
values increase with increased relative density of granular soils. Observations during pile
installation suggest that waterbearing granular soils may provide lower skin friction resistance
compared to soils of similar relative density above the groundwater table. However, PDA test
results indicate similar ultimate skin friction values both above and below the groundwater table.
There appears to be some differences in skin friction resistance above and below the
© ASCE
Geo-Congress 2020
groundwater table when the measured skin friction values are normalized for effective vertical
stress. Data suggests that normalized ultimate skin friction values for waterbearing granular soils
may be slightly lower than soils of the same relative density soils above the groundwater table.
However, additional data is required before drawing firm conclusions on the differences in skin
friction values for granular soils above and below the groundwater table.
With a deep foundation along the perimeter of the building and shallow spread footings
supporting interior columns, differential settlement between the different foundation types must
be evaluated. The expected differential settlement must be within the tolerable limits of the
structural framing system.

Because sheet pile foundations are designed to support axial loads, stringent installation
tolerances must be met. Sheet pile installed out of plumb or out of position horizontally will
introduce eccentric loads into the foundation which can increase bending stresses in the sheet
pile that are not accounted for in the structural design. The presence of cobbles, boulders, and
dense soils can increase the potential for deviating from specified installation tolerances.
Therefore, thorough characterization of the subsurface conditions is critical to identify potential
difficult pile installation conditions. The impacts of vibrations due to pile installation must also
be investigated prior to construction, and vibrations must be carefully monitored during pile
installation to reduce the potential for negative impacts on adjacent structures.
REFERENCES
Abbondanza, D. (2009). "Steel innovation: changing the economics of below grade foundations."
Contemporary Topics in Ground Modification, Problem Soils, and Geo-Support,
Geotechnical Special Publication No. 187, ASCE, Reston, VA, 57-64
Bustamante, M.G. and Gianeselli, L. (1991). "Predicting the bearing capacity of sheet piles under
vertical load." Proc. 4th Intl. Conference on Piling and Deep Foundations, TESPA, Italy, 9
p.
Terzaghi, K. and Peck, R. B. (1996). Soil Mechanics in Engineering Practice, 3rd ed., John
Wiley & Sons, Inc., New York, NY.
Underwood, C.A., and Greenlee, R.M. (2010). “Steel Sheet Pile Used as Permanent Foundation
and Retention Systems – Design and Construction.” Proceedings of the 2010 Earth Retention
Conference (Earth Retention Conference 3), Bellvue, WA, Geotechnical Special Publication
No. 208, ASCE, Reston, VA, 129-136.
Underwood, C.A., and Greenlee, R.M. (2011). “The Use of Steel Sheet Pile as Permanent
Building Foundations – Case Histories from Minnesota.” Proceedings of the University of
Minnesota 59th Annual Geotechnical Conference, University of Minnesota, St. Paul, MN,
189-196.
USACE (1991). Design of Pile Foundations. US Army Corps of Engineers Engineering Manual
EM 1110-2-2906.
© ASCE
Geo-Congress 2020
Quo Vadis? Inakeyaa! Inferring Flow Directions Using a Bayesian Approach

Randal J. Barnes1 and Richard Soule2
1
Dept. of Civil, Environmental, and Geo-Engineering, Univ. of Minnesota, Minneapolis, MN. E-
mail: barne003@umn.edu
2
Source Water Protection, Minnesota Dept. of Health, Saint Paul, MN. E-mail:
richard.soule@state.mn.us
ABSTRACT
Before we build a predictive groundwater flow model, we first identify candidate conceptual
models. To cull our collection of candidate models, we ask the seemingly simple but key
question: Where is the water coming from? We argue for the use of a Bayesian framework to
address this question. We start with an uninformed prior distribution that says all directions are
equally likely. We then incorporate the available information (usually error-prone, noisy
information of varied quality) and appropriately update the characterization of the uncertain
direction. When the added information is extensive and internally consistent, a clear flow
direction emerges. On the other hand, if the added information is minimal or internally
inconsistent, the uninformed prior is only slightly modified.
INTRODUCTION
Determining the likely groundwater flow direction from water level data is a common
practical problem. Since the advent of the personal computer, this problem domain has been well
trodden (e.g. Kelly and Bogardi, 1989). “Two points determine a line,” and “three points
determine a plane” are common mantras in introductory geometry. Given a plane, we have a dip
and dip direction: a lesson from structural geology. Pinder et al. (1981) furnishes the details of
the three-point problem for computing the hydraulic head gradient based on these principles.
Introductory hydrogeology classes commonly include a method to determine the hydraulic head
gradient from three head measurements (e.g. Fetter, 1994, Section 4.12).
The purely geometric perspective embraced by most authors ignores the inherent uncertainty
in all head measurements. Piezometric data are subject to many sources of perturbations,
deviations, and outright errors; for example, uncertain collar locations, imprecise collar
elevations, error-prone depth measurements, fluctuating pumping influences, unquantified
seasonal effects, climatic variations, and changing usage of groundwater resources. This list is
not exhaustive. We propose that head measurements must be understood as stochastic quantities.
Furthermore, errors and fluctuations in the head measurements will have significant impacts on
the inferred flow directions (e.g. Devlin and McElwee, 2007).
The approach outlined in this paper (project name: Inakeyaa) uses a Bayesian framework
(see, for example, (Vick, 2002)). We start with an uninformed prior distribution that says all
directions are equally likely. We then incorporate the available information (usually error-prone,
noisy information of varied quality) and appropriately update the characterization of the
uncertain direction. If the added information is extensive and internally consistent, a clear flow
direction emerges. On the other hand, if the added information is minimal or internally
inconsistent, the uninformed prior is only slightly modified.
Where is the water coming from – Quo vadis? (Latin: Where are you marching?). Our
response – Inakeyaa! (Ojibwe: In that direction.)
© ASCE
Geo-Congress 2020
THE MODEL
Aquifer Model
We model the aquifer as homogeneous and horizontally isotropic, on a horizontal base. The
aquifer thickness, H[m], horizontal hydraulic conductivity, k [m/d], and vertical hydraulic
conductivity, kv [m/d], characterize such an aquifer for our purposes.
We assume that our model domain is sufficiently far from all specific influential features like
rivers, lakes, and high-capacity pumping wells such that steady uniform flow captures the
essential local behavior. We ignore recharge and leakage. Thus, we are considering only a sub-
domain of any real aquifer. Despite these limiting assumptions, the results are often remarkably
informative — especially when building initial conceptual models.
Coordinates and Notation
We introduce a Cartesian coordinate system (x, y, z), where x [m] and y [m] are the
horizontal coordinates and z [m] is the vertical coordinate; z is measured up from the base of the
aquifer. We denote the piezometric head in the aquifer by ϕ(x, y, z) [m], the saturated thickness
by h(x, y) [m], and the two horizontal components of the specific discharge vector by qx(x, y, z)
[m/d] and qy(x, y, z) [m/d].
Discharge and Discharge Potential
We define the components of the discharge vector, Qx(x, y) [m2/d] and Qy(x, y) [m2/d], as the
vertically integrated specific discharge.
h(x,y) h(x,y)
Qx  x, y    q x (x,y,z)dz Q y  x, y    q y (x,y,z)dz (1)

0 0
Following Bear (1979, p. 77-78) and Strack et al. (2006), we integrate (1) using Leibnitz’s
rule, apply Darcy’s law, substitute in (4), and obtain the following expressions:
 
Q x  x,y   Q y  x,y   (2)
x y
where the discharge potential, Φ(x, y) [m3/d], is given by
1
  x,y   kh kh 2 (3)
2
with the vertically averaged piezometric head, ( x, y) [m], defined by
h  x,y 
1
  x,y     x,y,z  dz
h  x,y  0
(4)
We observe, following Youngs (1966), that (3) is valid even when the hydraulic conductivity
is vertically anisotropic.
Uniform Flow
As stated earlier, we assume that uniform flow is the dominant feature within the model
domain. The resulting discharge potential is given by
  x,y   Ax+By+C (5)
2 2 3
(Strack, 1989, (8.42), p. 57), where A [m /day], B [m /day], and C [m /day] are uncertain
© ASCE
Geo-Congress 2020
parameters.
Flow Direction
The components of the discharge vector are computed from the discharge potential by
applying (2) to (5). The resulting flow direction, α [rad], is computed using the standard four-
quadrant arctangent function as
α  atan2(B,A) (6)
In this paper we define the “flow direction” as the direction from which the flow originates;
e.g. a “north wind” blows from the north. This definition is useful in wellhead protection because
it helps identify the source areas. Our goal in this note is to characterize the probability
distribution of α.
Confined Flow
When h(x, y) = H over the model domain (i.e. confined flow), we can substitute (3) into (5)
and write
1
kH  kH2  Ax  By  C (7)
2
We rearrange (7) to read
1
C+ kH 2
A B 2
 x+ y+ (8)
kH kH kH
Defining three new parameters, {a, b, c}, we can rewrite (8) as
( x, y)  ax  by  c (9)
1
where a  A / (kH), b  B / (kH),and c  (C  kH 2 ) / (kH) . Since k and H are strictly positive
2
numbers, we can rewrite (6) as α = atan2(b, a). Thus, we may carry out our analysis in terms of
the vertically averaged heads and never need to consider the discharge potential. In this simple
confined model, we can infer the flow direction without knowing H, k or kv.
Unconfined Flow
When h(x, y) < H over the model domain (i.e. unconfined flow), we can substitute (3) into
(5) and write
1
kh  kh 2  Ax  By  C (10)
2
The left side of (10) is problematic for two reasons. First, there are two variables: h(x, y) and
( x, y) . Second, the left side is nonlinear.
We embrace the Dupuit-Forchheimer approximation (Dupuit, 1863; Forchheimer, 1886;
Polubarinova-Kochina, 1962; Harr, 1962; Youngs, 1966; Bear, 1972; Haitjema, 1987; Strack et
al., 2006) and write
h ( x, y)  ( x, y) (11)
Substituting (11) into (10) yields
1
k  Ax  By  C (12)
2
© ASCE
Geo-Congress 2020
We linearize (12) to eliminate the problematic nonlinear behavior on the left side. With some
loss of generality, we limit our analysis of unconfined flow to cases where the maximum head
over the model domain, max , and the minimum head over the model domain, min , satisfy the
condition
max  min  min (13)
That is, the range of  over the model domain is much less than the minimum value of 
over the model domain. In practice, this condition is not an additional restriction beyond those
previously identified since the prohibited thinning of the saturated thickness only occurs near
specific influential features like rivers, lakes, and high-capacity pumping wells.
We rewrite the left side of (12) as
 k    min 
1  1 2 1 2
k  kmin   kmin (14)
2 2 2
Substituting (14) into (12) and rearranging yields
1 2
   min  C+ kmin
2
A B 2
  x+ y+ (15)
2min kmin kmin kmin
Defining three new parameters, {a, b, c}, we can rewrite (15) as
     2 
 1    ax  by+c
min
(16)
 2min 
 
1 2
C  kmin
A A 2
where a= ,b , and c  . The assumption expressed in (13) justifies the
kmin kmin kmin
following bounds and subsequent approximation.
       min 
2 2 2
1  
  max min   1
min max
(17)
2min 2min
2
2  min 
and thus,
( x, y)  ax  by  c (18)
Since k and min are strictly positive numbers, we can rewrite (6) as α = atan2(b, a). As with
the confined case, we may carry out our analysis in terms of heads and never consider the
discharge potential. In this simple unconfined model, we can infer the flow direction without
knowing H, k, kv, max , or min .
PRIOR DISTRIBUTION
Concrete Subjectivity
For both the confined and unconfined cases we have three unknown parameters: {a, b, c}.
We organize these three unknown model parameters into the (3 × 1) matrix w = [a, b, c]T.
Our subjective prior distribution for these parameters is built upon three postulates about
what we know before considering any head data from the area. (1) We have no knowledge about
the flow direction — all directions are equally likely. This suggests that unknown parameters a
© ASCE
Geo-Congress 2020
and b are independent and identically distributed. (2) We have some knowledge about the
magnitude of head gradient based on experience in similar flow domains and topographic
information. (3) We have no knowledge about the parameter c. This parameter can be positive or
negative depending on where we place the origin of the local coordinate system.
Building upon these postulates, we select a multivariate normal distribution (e.g. Kotz et al.
(2000, Chapter 45)) as our subjective prior distribution, with
0  σ 2p 0 0 
 
E  w   0  0  Var  w   V  0 σ 2p 0 
(19)
0  0 0  
 
where σ p denotes the common variance of a and b.
2
It would be possible to use some other multivariate distribution to characterize a and b, but
this choice of multivariate normal satisfies our postulates, offers sufficient flexibility, and is
computationally congenial (Calvetti and Somersalo, 2007).
Determining σp from Professional Experience
Since a and b are uncorrelated normal random variables, with means 0 and standard
 a 2 b2 
deviations σp, the random variable  2  2  follows a chi-square distribution with two degrees
σ 
 p σp 
of freedom (e.g., Manoukian (1986, Section 4.25)). The distribution function of the chi-square
gives
Pr  
a 2  b2  xσ p  1  e x
2
/2
(20)
We recognize the term a 2  b2 as the magnitude of the head gradient. We denote the prior,
subjective, assessment for the median of head gradient by i50 and solve (20) for σp.
i50
σp  (21)
2 ln 2
MEASURED HEADS
We have n measured heads, ϕi [m] at known locations (xi, yi) [m] for i = 1, 2, . . . , n. In this
analysis we ignore the potential errors and uncertainties in the well collar locations. This offers a
significant simplification and is consistent with our use of only “located wells”. These head
measurements include measurement errors, εi [m]. The underlying linear regression model is
i  ax i  byi  c+εi (22)
for all i = 1, 2, . . . , n. For notational convenience we organize the data into regression matrices
as follows.
 x1 y1 1   ε1 
 x y 1   ε 
X=  2 2  z   2 e=  2  (23)
     
     
 x n y n 1 n  ε n 
Using this compact notation, we can rewrite the complete underlying linear regression model,
© ASCE
Geo-Congress 2020
(22), as
z  Zw  e (24)
We assume that the measurement errors are independent, zero mean (i.e. the head
measurements are unbiased), Gaussian random noise. We permit the standard deviation of each
measurement error to be individually specified. The associated standard deviations may reflect
knowledge about the data source: carefully surveyed, monitored, and measured wells are given a
lower standard deviation; unsurveyed, unmonitored, and casually measured wells are given a
higher standard deviation. We denote the standard deviation of the i’th measurement error by σi,
and organize the associated variances in a diagonal matrix, G  diag(σ12 , σ 22 , , σ 2n ) .

We assume that these standard deviations are known a priori from alternate studies. The prior
distribution of the measurement errors is multivariate normal: e  (0, G) . By including off-
diagonal terms in G, our model can be extended to include temporally or spatially correlated
measurement errors.
BAYESIAN LINEAR REGRESSION
Bayesian linear regression infers the joint distribution of the model parameters based upon
observed data. Unlike common least squares regression, Bayesian linear regression also
incorporates subjective prior information about the parameters into the inference (see, for
example, Chen and Martin (2009, Section 2)).
The parameter likelihood function is multivariate normal: (Xw, G) . The prior distribution
for the parameters is multivariate normal: (0, V) . Applying Bayes’ Theorem to obtain the
posterior distribution, we conclude that the distribution of parameters, w, conditioned on the
observations, z, is multivariate normal (e.g. Wang (2009, Section 2.2)). The conditional mean
and variance are
E  w | z   H1XTG 1z (25)
Var  w | z   H1 (26)
where
H  V1  XTG1X (27)
Note that, in general, the posterior parameter distributions are correlated since they are based
on the same conditioning information.
FLOW DIRECTION
The flow direction is a function of a and b only, so we need to extract the associated bivariate
marginal distribution from the trivariate posterior distribution. For a multivariate normal
distribution, this is a simple process of extracting the first and second row of the conditional
expected value matrix, (25), and the first and second rows and columns of the conditional
variance matrix, (26) (e.g., Roussas (1973, Section 18.2, Theorem 3)). We use the following
notation to denote these extracted rows and columns.
 μ a|z  σ 2 σ ab|z 
u  S   a|z 2 
(28)
μ
  b|z 
 σ ab|z σ b|z 

The resulting distribution of the flow direction, α, follows Smith’s distribution (Justus (1978,
Equation (4-11)) or Carta et al. (2008, Equation (6))):
© ASCE
Geo-Congress 2020
S 1
1  1  
 D  exp  
1 T 1 
f α  1  
1  2πD  exp  D 2   u S u (29)
2   2 
T
r S r 2π  
cos α 
r T S 1u
D= r  (30)
r T S 1r sin α 
and denotes the standard normal cumulative distribution function.
Smith’s distribution model is relatively uncommon in the literature on directional data
analysis. The most common model is the von Mises distribution, even for Bayesian analysis
(Nuñez-Antonio and Gutiérrez-Peña, 2005). Nonetheless, in our case, Smith’s is appropriate
because we are not analyzing directional measurements. We are inferring the distribution of the
flow direction from a planar fit to head measurements.
APPLICATION
Drillers’ Data
The Bayesian approach outlined above was exercised for the city of Clear Lake, Minnesota.
For the past 50 years the Minnesota Geologic Survey has maintained a database of well-driller
reports. We identified a set of 62 wells from this database with field-verified locations within 2
[km] of the Clear Lake water supply well. We excluded 13 wells that were located within 100
[m] of a high-capacity well and had measurement dates that corresponded with the pumping
period of that well. The heads at the remaining 49 wells were calculated using the static water
depth reported by the well driller in the state database and the elevation from the National
Elevation Dataset (NED) 30-meter grid (Gesch et al., 2010).
We included two sources of uncertainty in the measured heads: elevation error and temporal
variation. We estimated the variance of the elevation error by comparing the survey grade
elevations at 1, 092 points in Minnesota to the associated values extracted from the NED grid.
The resulting variance was 2.07 [m2].
We estimated the variance of the temporal variation using data from similar hydrogeologic
and geographic settings within the state. There are 299 DNR observation wells completed in
similar aquifers, each with more than 30 water level observations. The median of the variances
from these wells is 0.472 [m2]. We estimated the total variance using the direct sum of the
temporal and spatial components: 2.542 [m2].
We estimated the median gradient using the NED elevations for Clear Lake and Jones Lake
as 0.001 [ ] Based on the 49 head data, we evaluated the regional flow direction at Clear Lake
using the Bayesian approach.
Observation Wells
The Minnesota Department of Natural Resources (DNR) maintained three surveyed
observation wells from 1983 to 2000. The locations of the wells form an approximate equilateral
triangle on a circle of radius 1 [km] centered on the Clear Lake water supply well. Water levels
were measured in these wells on the same day, 10 months of the year for nearly 18 years,
yielding 172 sets of synoptic measurements. Using plane geometry we computed the apparent
flow direction (e.g. Pinder et al. [1981]) for each of the 172 sets of three synoptic measurements.
© ASCE
Geo-Congress 2020
Discussion
The Clear Lake example compares two very different data sets and associated analyses: 18
years of synoptic data from the DNR observation wells, and the Bayesian approach applied to the
drillers’ water level data. See Table 1. Both methods identify the same approximate most likely
flow direction. The widths of the 50 percent probability intervals are also similar.
Table 1: Comparison of the results for the Clear Lake area using data from two sources.
Data Set Count Flow Direction [deg]

Median Lower Upper
Quartile Quartile
DNR Synoptic Data 172 126 124 130
Bayesian Approach 49 128 122 134
The advantage of applying the Bayesian approach to the drillers’ water level data is that it
does not require 18 years. The method can be employed immediately to exploit the large quantity
of lower-quality data that has heretofore been largely ignored. The method correctly incorporates
information of varied quality and returns a proper evaluation of the uncertainty. The method is
self-policing. If the available data is extensive and internally consistent, a clear flow direction
with a narrow probability interval emerges. On the other hand, if the available data is minimal or
internally inconsistent, the probability interval is wide.
REFERENCES
Bear, J. (1972). Dynamics of Fluids in Porous Media. Dover Publications, Inc., New York 764
pp.
Bear, J. (1979). Hydraulics of Groundwater. McGraw-Hill International Book Company, New
York 544 pp.
Calvetti, D. and Somersalo, E. (2007). Introduction to Bayesian Scientific Computing: Ten
Lectures on Subjective Computing (Surveys and Tutorials in the Applied Mathematical
Sciences). Springer.
Carta, J. A., Bueno, C., and Ramírez, P. (2008). “Statistical modeling of directional wind speeds
using mixture of von Mises distribution: Case study.” Energy Conversion and Management,
49, 897–907.
Chen, T. and Martin, E. (2009). “Bayesian linear regression and variable selection for
spectroscopic calibration.” Analytica Chimica Acta, 631, 13–21.
Devlin, J. F. and McElwee, C. D. (2007). “Effects of measurement error on horizontal hydraulic
gradient estimates.” Ground Water, 45(1), 62–73.
Dupuit, J. (1863). Etudes Théoretiques et Pratiques le Mouvement des Eaux dans les Canaux
Découverts et á Travers les Terrains Permeables. Dunod, Paris, 2nd edition.
Fetter, C. W. (1994). Applied Hydrogeology. Macmillan College Publishing Company, Inc., New
York, third edition 691 pp.
Forchheimer, P. (1886). “Ueber die ergiebigkeit von brunnen-anlagen und sickerschlitzen.” Z.
Architekt. Ing. Ver. Hannover, 32, 539–563.
Gesch, D., Evans, G., Mauck, J., Hutchinson, J., and Jr., W. C. (2010). “The national map -
elevation: U.S. Geological Survey fact sheet.” Report No. 2009-3053, U.S. Geologic Survey
(April).
Haitjema, H. M. (1987). “Comparing a three-dimensional and a Dupuit–Forchheimer solution for
© ASCE
Geo-Congress 2020
a circular recharge area in a confined aquifer.” Journal of Hydrology, 91, 83–101.

Harr, M. E. (1962). Groundwater and Seepage. McGraw-Hill Book Company, New York.
Justus, C. G. (1978). Winds and Wind System Performance. Solar energy. Franklin Institute
Press, Philadelphia, Pennsylvania 120 pp.
Kelly, W. E. and Bogardi, I. (1989). “Flow directions with a spreadsheet.” Ground Water, 27(2),
245–247.
Kotz, S., Balakrishnan, N., and Johnson, N. L. (2000). Continuous Multivariate Distributions,
Models and Applications, Vol. 1 of Wiley Series in Probability and Statistics. Wiley-
Interscience (May).
Manoukian, E. B. (1986). Modern Concepts and Theorems of Mathematical Statistics. Springer-
Verlag, New York 156 pp.
Nuñez-Antonio, G. and Gutiérrez-Peña, E. (2005). “A Bayesian analysis of directional data using
the von Mises–Fisher distribution.” Communications in Statistics – Simulation and
Computation, 34, 989–999 doi:.
Pinder, G. F., Celia, M., and Gray, W. G. (1981). “Velocity calculation from randomly located
hydraulic heads.” Ground Water, 19(3), 262–264.
Polubarinova-Kochina, P. (1962). Theory of Groundwater Movement. Princeton University
Press, Princeton, NJ.
Roussas, G. G. (1973). A First Course in Mathematical Statistics. Addison-Wesley.
Strack, O. D. L. (1989). Groundwater Mechanics. Prentice-Hall, Inc., Englewood Cliffs, New
Jersey 732 pp.
Strack, O. D. L., Barnes, R. J., and Verruijt, A. (2006). “Vertically integrated flows, discharge
potential, and the Dupuit-Forchheimer approximation.” Ground Water, 44(1), 72–75.
Vick, S. G. (2002). Degrees of Belief: Subjective Probability and Engineering Judgement.
American Society of Civil Engineers 455 pp.
Wang, Z. (2009). “Semi-parametric Bayesian models extending weighted least squares.” Ph.D.
thesis, Ohio State University, Ohio State University.
Youngs, E. G. (1966). “Exact analysis of certain problems of ground-water flow with free
surface conditions.” Journal of Hydrology, 4, 277–281.
© ASCE
Geo-Congress 2020
Supporting a Bridge between Countries Case Study: Construction of Baudette Bridge

Drilled Shafts
Nathan W. Iverson, P.E.1
1
Veit and Companies, Foundation Division, Rogers, MN. E-mail: niverson@veitusa.com
ABSTRACT
Minnesota Department of Transportation (MNDOT) and the Ontario Ministry of

Transportation (MOT), started construction of a new border crossing to replace the existing
historic bridge in Baudette, Minnesota. The new bridge was designed to bear on drilled shafts
instead of driven pile like the adjacent railroad bridge and previously constructed bridge. The
bridge design includes four river piers supported on 8’ diameter drilled shafts up to 100’ below
river elevation. Piers one and two are located on the United States side of the river and piers
three and four are location on the Canadian side. The piers were constructed in the existing river
from barges. The technical challenges and extreme remote location required strong coordination
between multiple agencies, engineers, and two federal governments. This case study details the
technique shaft, bi-directional test, concrete design and production (mass, tremie, and self-
consolidating concrete), thermal integrity profiler results, obstructions, drilling techniques, clean
out procedures, and a dynamic working environment in an extreme northern environment.
INTRODUCTION
The Rainy River Bridge is located on the Northern Border of Minnesota and Canada. The
Baudette – Rainy River International Bridge is a point of entry for approximately 1,350 vehicles
per day serving as a regional link for trade and business. The existing bridge is an historic steel
truss bridge that was constructed in 1959, but the bridge was identified as a Fracture Critical
bridge and Functionally Deficient as it is not permitted to carry overweight loads or over
dimensional loads.
After deliberations and various alternatives, the Department of Transportation and Ministry
of Transportation elected to pursue a bridge replacement. The new bridge was a continuous steel
girder design composed of 5 spans with a total length of 1,350 feet.
Table 1. Soil and Pier Information.

Technique Pier 1 Pier 2 Pier 3 Pier 4
Country USA USA USA Canada Canada
Top of Pier 1065.0 ft 1056.1 ft 1058.3 ft 1058.3 ft 1058.3 ft
Water 1061.8 ft 1061.8 ft 1061.8 ft 1061.8 ft 1061.8 ft
Alluvial Soil 1042.0 ft 1058.0 ft 1042.0 ft 1040.0 ft 1042.0 ft
Dense Till 1021.0 ft 1016.0 ft 1021.0 ft 1015.0 ft 1017.0 ft
Rock NA 972.0 ft NA NA NA
Bottom Cased 1015.5 ft 1015.5 ft 1015.5 ft 1015.5 ft 1015.5 ft
Bottom Pier 935.5 ft 972.0 ft 953.5 ft 953.5 ft 958.5 ft
SOIL CONDITIONS
The soil borings at the drilled pier locations suggested that the soil conditions would be
© ASCE
Geo-Congress 2020
predominantly loose alluvial soils overlying dense glacial till. The water level is approximately
1061.8 feet. The river bottom ranged from 1058.0 feet at Pier 1 to 1040.0 feet at Pier 3. The table
(Table 1) shows the pertinent transition elevations relative to each pier.
The river bottom was composed mostly of very loose granular soils overlaying a fine-grained
material (silts and clays) near the abutment and piers. Towards the center of the river (Pier 3), the
layer of loose granular soils extended deeper almost completely replacing the fine-grained soils
present in other borings.
These surficial soils transitioned to glacial till. The till varied in composition, but was
primarily composed of granular soils with varying amounts of silt and clay. The blow counts in
this formation were extremely dense ranging from 50 blows per 5 inch to impenetrable by a split
spoon. This suggested relatively high percentage of gravels, cobbles and potentially boulders.
This coupled with poor sample recovery led to a drilled shaft specification that called for
obstructions below 22.5 inch in diameter to be considered incidental to construction.
In order to get a better idea of till deposit, the quaternary data for the region was referenced.
The quaternary map for both Minnesota and Ontario (Quaternary Lithostratigraphic Unit of
Minnesota, Quaternary Geology of Fort Frances-Rainy River Area) were examined. These
reports suggest the Blackduck formation from the Minnesota quaternary or Whitshell Till -
Labradorean derived or Whitemouth Lake Till - Keewatin derived from the from the Ontario
quaternary. Both these documents suggested possible glacial fluvial deposits, which could
contain layers of gravel, cobbles and boulders. The cobbles and boulders were thought to be
subangular and of crystalline rock types.
Figure 1. Drilled Shaft Design – Pier 1.

FOUNDATION DESIGN
The existing bridge was bearing on timber pile. The newly constructed Canadian National
Railway (“CN”) bridge located just upstream was also bearing on driven pile. The concern in
utilizing driven pile was potential of scour. MNDOT’s consultant, HZ United, identified the
critical scour elevation was located at an elevation of 1013.0 feet to 1015.0 feet at or near the
elevation of the transition between the dense till and the overlying loose alluvial soils. The
existing bridge had experienced significant flow events historically (1969, 1970, 1974, 1979,
1985, 2001, 2002, 2014). These events created flows, which ranged from 59,000 to 78,600 cubic
feet per second. These events required subsequent grouting and rip rap placement around the
© ASCE
Geo-Congress 2020
piers of the existing bridge caisson to repair excessive scour at multiple of the piers. This posed a
problem for the future bridge bearing on driven pile as the piles would obtain limited embedment
into the underlying till soils and be exposed to similar scour in the future. This also brought to
question how the piers were would perform in a 100-year events
The final foundation utilized driven pile for the abutments and drilled shafts for the river
piers. The drilled shafts were designed as 8.0 feet diameter with permanent 0.75 inch-thick
casing extending to 1015.5 feet or to critical scour elevation. A 7.5 feet diameter socket would be
constructed into the dense glacial till. A pair of these shafts would be used to support the bridge
at each bent location. A precast concrete pile cap would then be placed for each bent.
MATERIALS
The drilled shafts were composed of 3 primary materials; permanent casing 8.0 foot diameter
with a thickness of 0.75 inches, #14 reinforcing steel bars, and ready-mix concrete. The large
diameter casing is typically a lead time concern as it is manufactured specifically for the project.
The reinforcing steel poses some issues as the large diameter of the bar doesn’t allow for lap
splicing, so mechanical couplers would need to be used. The concrete procurement proved to be
the most challenging lead time item for the foundation portion of the project.
Due to the remote location of the project, ready mix in the region is generally limited to basic
concrete mixes. There is not a state certified ready-mix facility within 1 hour of the jobsite. This
was compounded by the fact that most of the providers in the region have little to no knowledge
about mass concrete design or cementitious materials other than fly ash and cement. The ready-
mix facilities in the region typically don’t utilize fly ash in order to avoid an additional silo
onsite. This issue was left it to the contractors to work out a solution for supply that would
reliably fulfill the design requirements. Transporting the drilled shaft concrete the hour or longer
while attempting to consistently adhere to the project specifications for the mass concrete or self-
consolidating concrete would not be practical or even possible. This meant that a solution needed
to be developed for a local supply.
During the estimating process for this project, the local and regional concrete contractors
were contacted for quotes for the project. Most of the ready-mix suppliers refused to quote the
project due to the relatively small quantities ready mix versus the large startup costs of a portable
ready-mix facility. The only price tendered to generals at bid time was almost 8 times the typical
ready-mix price. This escalation would add millions of dollars to the project bid and through
review of the quote still required significant lead time and still posed logistical challenges.
Veit and Lunda Construction (general contractor) reached out to one of the local ready-mix
suppliers and developed a collaborative solution that utilized both company’s experience with
mix designs on other projects to upgrade the capability of the local ready-mix supplier’s local
small batch facility. By providing the local ready-mix supplier the engineering support as well as
providing past experience with similar projects, the local ready-mix supplier upgraded their
facility. By installing two new silos and improving the automated batching system at their local
facility, the ready-mix supplier was able to get the plant certified and ready for concrete
production. The plant would be capable of generating approximately 40 cubic yards per hour,
which would allow for continuous placement of ready mix to the site. The trip time from their
production location to the project site would be less than 10 minutes and posed little to no risk of
delays due to road conditions. Veit and Lunda engaged American Engineering and Testing
(AET) to develop the mix design that would function as a self-consolidating concrete (SCC) as
well as meet the mass concrete requirements. Once the mix design had been completed, tested,
© ASCE
Geo-Congress 2020
and submitted to the MNDOT for approval, Veit started working on the multiple iterations of
trial batching and test placements. Due to the requirements of a mass concrete mix, a full-scale
test placement was performed. The mix design requirements originally called for a 4 inches to 6
inches slump. The trial placement at this slump proved to be less than ideal. Veit requested a
change in the specification to allow for greater fluidity and replacement of the slump test with
the spread test. The slump requirement was ultimately removed in favor of a spread test and a
design spread of 18 inches to 21 inches was adopted. The second placement proved successful
but showed signs of issues with batch consistency. It was decided that a Veit full time quality
control person would test at the plant with the ready-mix provider and be present at the site, to
run tests to provide active feedback to adjust for the temperature change during concrete
placement as the day progressed. Both placements maintained concrete temperatures below the
specified maximum temperature allowed for mass concrete per project specifications.
Figure 2. Test Block 1 (left), Test Block 2 (right)

WORKING ON THE BORDER
While the ready-mix design process was underway and the materials procurement process
started for the lead time components, the work visa process and imported materials
© ASCE
Geo-Congress 2020
documentation was also being completed. This process involved lawyers, customs brokerage
agents, consultants, and multiple meetings with the customs officials from both countries. This
process became more complicated after about three months into the project when the United
States government levied tariffs on Canadian goods and in turn Canada instituted tariffs on
United States goods. Veit’s role in the project limited some exposure to this issue, since the
project was an international border crossing with onsite agents and Veit’s work never required
Veit personnel to set foot on Canadian soil. The process was completed by modifying crews and
working with the border agents to meet the work visa and labor requirements of the border
agents and local governments.
EQUIPMENT
The drill rig used to install these relatively large holes was the CZM EK250. The CZM
EK250 is capable of drilling up to 13 inch in diameter to a max depth of 159 ft with an
interlocking kelly bar. The drill operates at a max torque of 234,500 foot pounds with 73,326
foot pounds of crowd. The dense till and shaft size required large aggressive drill tooling to
remove the very dense till and boulders resulting in a very irregular sidewall. The large drill rig
also created more demand on the barge system to support and resist the forces of the drill rig
during operations.
Figure 3. Test Block 1 (left), Test Block 2 (right)
SHAFT STABILIZATION
In order to install the permanent casing and stabilize the open holed portion of the shaft, Veit
proposed a multiple method approach. A temporary casing would be installed into the alluvial
soils at each pier location. The permanent casing would then be rotated and pushed down to
design elevation. In the scenario where an obstruction was encountered during the permanent
© ASCE
Geo-Congress 2020
casing installation, slurry could be utilized to excavate the obstruction prior to installation of the
permanent casing.
Once the permanent casing was placed and secured to the temporary casing and a temporary
support and alignment jig, slurry would then be introduced to the shaft and the shaft excavation
process would proceed. The large outer casing would help contain the slurry and function as a
secondary containment for the slurry process.
The specification for the project did not allow for bentonite slurry to prevent the potential
development of a filter cake developing on the sidewalls of the soil socket, so a polymer solution
was utilized. Veit selected KB International for the slurry supplier for this project. Working with
KB, Veit helped to tailor a slurry procedure that allowed for a tiered approach for slurry
adjustments. The base polymer slurry would be utilized to start the shaft while fibers and
additional additives would be added as needed to control slurry loss and maintain head.
ACCESS
Pier one was accessed by a coffer cell and causeway as it was located in relatively shallow
water and located close to the United States shoreline. The causeway would also function as the
staging area for subsequent construction and dock headwall for barge support for the remaining
piers.
Figure 4. Barge Spread

Lunda Construction working with input from Veit worked on developing an access plan for
each of the piers. The access plan incorporated a casing support jig at each location. The jig was
© ASCE
Geo-Congress 2020
composed of a welded steel frame and would function as an alignment device and support for the
temporary casing and permanent casing. The jig was supported by a series of H-piles driven 25
feet into the river bottom and then welded to the temporary casing. This allowed for a rigid
platform at each pier bent. The jig was designed to span the distance between both the drilled
shafts at each location.
Piers two, three and four would be accessed by a series of barges. Veit would utilize one
barge spread for slurry containment and one barge spread for the drill rig and crew. Lunda would
then operate two ancillary barge spreads, one for material transport and one for the support
crane.
Concrete would be pumped to the shaft through a series of concrete pumps and barge
transport as required for each placement.
TECHNIQUE SHAFT / BI-DIRECTIONAL TEST

Once the submittals were approved, the construction of the technique shaft could start.
During the construction of the shaft, numerous obstructions were encountered, which
significantly impacted the drilling production. Many of these obstructions were protruding from
the sidewalls of the shaft. These sidewall obstructions appeared to be granitic boulders that were
in excess of 3 foot in diameter and lodged firmly in the side of the shaft, so firmly in fact that
they resisted hundreds of thousands of pounds of force of the tooling that was used to attempt to
dislodge them, which subsequently resulted in significant tooling wear and damage.
Figure 5. Shaft Diameter vs. Depth

In order to verify shaft diameter, a caliper was used to measure diameter versus depth (Figure
© ASCE
Geo-Congress 2020
5). Upon review of the caliper data and discussions with the engineer, Veit was requested to
remove the side wall obstructions from the shaft. The caliper was then used again to check
diameter to confirm removal of the sidewall obstructions. Figure 5 shows the comparison
between the caliper readings. The caliper data shows that at multiple elevations large
obstructions were removed from the sidewalls.
Figure 6. Hoisting and Placement of the 111.5 foot Reinforcing Cage
Figure 7. Rebar Cage with Sensors and Bi-Directional Jacks
© ASCE
Geo-Congress 2020
Upon completion of the drilling process, shaft cleanliness was verified by mini-SiD (Shaft
Inspection Device). Upon verification of bottom clean out, the reinforcing cage was installed.
The technique shaft also required that a bi-directional test be performed to verify the design.
The bi-direction jack was designed to load the shaft in 8,000 kips in both directions.
The shaft was instrumented with 5 levels of strain gages, linear displacement transducers,
telltales, thermal sensors, and thermal integrity sensors. This information was collected by
Applied Foundation Testing and used to analyze the performance of the pier.
Once the sensors were confirmed operational, the high mobility self-consolidating concrete
was placed in the shaft by means of a 6 inch tremie pipe connected to a concrete pump truck. The
shaft was then monitored with the thermal integrity sensors and mass concrete sensors to confirm
max temperature was not exceeded and no thermal anomalies were noted. This information was
then sent to GRL Engineers for processing and review. Figure 8 shows the thermal model for the
shaft.
Figure 8. Thermal Integrity Profile with Shaft Profile Model

The shaft was then allowed to finish curing until design strength had been reach for the
concrete. This process took about 10 days, due to the relatively high percentage of slag and fly
ash making up majority of the cementitious material in the concrete mix design.
Applied Foundation Testing (AFT), then proceeded to load the shaft via the bi-directional
cell. The test proceeded without issue until at a load of approximately 7,382 kips in both
directions. At this point, the hydraulic pumps froze due to the cold ambient air temperature. The
results were then verified with the foundation engineer (Dan Brown and Associates, DBA). It
was determined that enough information had been obtained to end the test early prior to reach the
intended 8,000 kips in both directions. Figure 9, shows the displacement both upward and
downward with resect to the applied load.
© ASCE
Geo-Congress 2020
Figure 9. Bi-axial Test - Load vs Displacement

Based on the bi-directional test results, the technique shaft was considered a success and
production shaft construction began the following week.
OBSTRUCTIONS
The major issue that arose out of the construction of the technique shaft was the amount of
obstruction and the extensive time required to remove the obstructions. The soil borings clearly
identified the presence of cobbles and possible boulders, but the size of the boulders and the
frequency of the oversized boulders far exceeded expectations. During the drilling operations for
the technique shaft encountered hundreds of cobbles and boulders, over 24 of these boulders
exceeded 22.5 inches in diameter. The largest boulder was approximately 43 inches in diameter.
These boulders were present from an elevation 1115.0 feet to bottom of shaft. They were, as
noted above, often lodged into the side wall of the shaft or bundled together in clusters. The
complicating factor were the geologic origin of the boulders. The boulders were predominately
granitic in origin, the compressive strength exceeded that of the steel and the hardness is
comparable to the carbide attempting to penetrate the boulder. The boulders made advancement
through the boulder impossible. In order to adjust to the drilling conditions, Veit mobilized
additional tooling to remove the obstructions. Due to the large diameter of the obstructions, a
significant variety of the sizes were required to accommodate the potential obstructions location
in the shaft and the surrounding soils or boulders present. The sidewall boulders also created
significant issues as the shaft becomes unstable as the rocks are dislodged from the side wall.
© ASCE
Geo-Congress 2020
The process of dislodging the obstructions required significant effort as the large boulders were
embedded into extremely dense soils. The carbide teeth and holders were often ripped from the
tool or boulders would fall onto the tooling and dent or damage the core barrels or drilling
buckets. Figures 10 – 13 show some of the tagged and cataloged boulders that exceeded the
obstruction size specification for the project.
Figure 10-13. Boulders Examples from Technique Shaft

The non-obstructed drilling rate through soil, cobbles, and small boulders was approximately
6 feet per hour. The obstructed drilling rate was often reduced to a standstill or required tooling
to be switched to a significantly smaller diameter to work around or process the soils around the
obstruction in order to allow the boulder to fall into the hole that was created for later removal.
The average production rate in the obstruction soils ranged from 0.5 feet per hour to 1.5 feet
per hour.
Due to the amount of obstructions, the production shafts were delayed until spring and
temporary casings were left in the river until construction could commence in spring after the
fish spawning season (Figure 14).
© ASCE
Geo-Congress 2020
Figure 14. Boulders Temporary Casings and Jig for Winter Shutdown
BRIDGE PIERS
The winter shutdown allowed for ample time for the construction team to meet to discuss and
plan for the spring construction startup. Partnering meetings and multiple series of
correspondence yielded a revised procedure that considered the boulder removal procedures, a
detailed decision matrix for change in conditions reporting, tooling repair and hardening, and a
revised pricing structure to facilitate cost reporting and compensation.
Table 2. Drilling Production and Obstructions

Tech Production Piers
Shaft 1-1 1-2 2-1 2-2 3-1 3-2 4-1 4-2
Duration (days) 22 8 10 10 16 15 10
2019
2019
# of Boulders >
22.5” 18 11 17 14 21 10 7
Drilling - Regular 15 hr 20 hr 15 hr 13 hr 7 hr 59 hr 47 hr
Sept
Drilling - Obstruct 120 hr 69 hr 39 hr 61 hr 99 hr 63 hr 16 hr

Oct
Obstruction Layer 70 ft 57 ft 70 ft 68 ft 72 ft 43 ft 20 ft
© ASCE
Geo-Congress 2020
PRODUCTION CONSTRUCTION
Drilled shaft construction started in May and was completed in early October of 2019, the
obstructions continued to play a significant part in the daily operations. The table below breaks
down the large quantity of obstructions encountered as well as the shaft duration.
When drilling without obstructions the shafts are advanced at about 20 feet per day and the
drilling duration should be about 5 to 6 days between temporary casing installation, permanent
casing installation, drilling and cleanout. The extensive removal of obstructions and removal of
side wall obstructions has required the use of concrete plugs, over 20 different tools of various
diameter and type, and a full-time tooling repair individual onsite to maintain the tools as they
are damaged from the repeated abuse from attempting to extract the numerous boulders present
in each shaft.
CONCLUSIONS
The drilled shaft construction for this project was challenging for an assortment of various
reasons. The project’s remote location created difficulties for material procurement and timely
adjustments to changing conditions. The project taking place in two countries made staffing
complicated as any change required multiple government agencies involvement. The large
diameter and working over the water posed constant logistics issues from concrete deliveries,
equipment repairs, and general access. All those things considered, the obstructions proved to be
the most burdensome aspect of the project. Due to the sheer number of boulders, compressive
strength, hardness, and size made drilling a daily struggle. The average shaft was 75%
obstructed, which resulted in months being added to the construction schedule. The resulting soil
socket performed well with side shear resistance values exceeding 3.6 ksf in the finer granular
material to 12.8 ksf in the dense boulder till. The dense till in end bearing exceeded 133 ksf.
These values were significant enough to potentially allow for a reduction in length, but due to the
potential concern of variability across the site it was elected to installed the drilled shafts at the
original designed lengths.
REFERENCES
A.F. Bajc, (2001) “Quaternary Geology - Fort Frances – Rainy River Area”, Ontario Geological
Survey
Harvey Thorleifson, (2016) “Quaternary Lithostratigraphic Units of Minnesota, Minnesota
Geological Survey, Report of Investigations 68, ISSN 0076-9177.
© ASCE
Geo-Congress 2020
Computational Tools for the Analysis of Stability of Embankments in Frictional-Cohesive

Soils
Carlos Carranza-Torres, Ph.D., P.E.1; and David Saftner, Ph.D., A.M.ASCE 2
1
Dept. of Civil Engineering, Univ. of Minnesota, Duluth Campus, Duluth, MN. E-mail:
carranza@d.umn.edu
2
Dept. of Civil Engineering, Univ. of Minnesota, Duluth Campus, Duluth, MN. E-mail:
dsaftner@d.umn.edu
ABSTRACT
This paper presents a series of computational tools for quick estimation of factors of safety
and critical circular failure surface location for embankments. The embankments are assumed to
be made of homogeneous frictional-cohesive dry soils, obeying a linear Mohr-Coulomb shear
strength criterion. Cases of a firm horizon that the failure surface cannot cut through are
considered to exist at the base of the embankment or at great depth. The latter case corresponds
to the case of a slope in homogeneous soil. Dimensionless functions defining the factor of safety
and position of the critical failure surface are constructed using the Bishop method of slices in
the commercial limit equilibrium software SLIDE. Results are implemented in a computer
spreadsheet that is made available for free downloading. Formulas for quick evaluation of factors
of safety are also presented. The effect of the cohesion and friction angle of the soil and the
position of the firm horizon on the resulting factors of safety and depth of the failure surface are
discussed. An application example solved with the proposed tools and with the shear strength
reduction technique implemented in the commercial software FLAC is presented. Although the
tools provided in this paper have been developed for teaching basic slope stability analysis in an
undergraduate soil mechanics course, the tools could be useful when performing preliminary
estimates of factors of safety of embankments in actual geotechnical design practice.
INTRODUCTION
Computation of factor of safety against failure and location of the critical failure surface are
the basis of the current practice of slope stability analysis in geotechnical engineering. These
stability analyses are an essential component in the design of embankments and cuts for
highways, levees, dams and dikes (Terzaghi et al. 1996; Coduto et al. 2011; Das & Sobhan
2018).
Various methods exist for computing the factor of safety and location of the critical failure
surface for slopes. Among them, limit equilibrium methods, limit analysis methods and full
numerical methods (e.g., Potts & Zdravkovic 1999; Davis & Selvadurai 2005). The so-called
shear strength reduction technique, normally implemented in finite element or finite difference
software (both tools being examples of implementation of the full numerical methods mentioned
above), is another method for computing the stability of a slope that has gained popularity in the
last two decades (e.g., Dawnson et al. 1999; Griffiths & Lane 1999; Hammah et al. 2007). All
existing analysis methods, with a few exemptions (e.g., analytical models that consider a planar
failure surface), require use of computer software for their implementation, as they are
computational-intensive. In this sense, it is remarkable that in contrast with other problems in
geotechnical engineering, there are no rigorous analytical (closed-form) solutions available for
computing the factor of safety and position of the critical failure surface for embankments or
© ASCE
Geo-Congress 2020
slopes. Therefore, with the lack of a true exact solution for the problem to be taken as a
benchmark, the methods used in stability analyses of embankments or slopes (i.e., limit
equilibrium methods, limit analysis methods and full numerical methods), are interpreted to give
approximate solutions to the problem only.
Despite of its introduction in the middle of the last century, the method of slices, which is a
particular formulation of the limit equilibrium method, is still the most popular method to
estimate the factor of safety and location of the critical failure surface for slopes (e.g., Abramson
et al. 2002; Duncan et al. 2014). Within the method of slices, there exist various formulations,
including the Bishop (1955), Janbu (1954a, 1954b), Spencer (1967) and Morgenstern-Price
(1965) methods. The Bishop method is popular since it gives similar results (i.e., factors of
safety and location of the failure surface) as those obtained with more elaborated limit
equilibrium methods (e.g., the Spencer and the Morgenstern-Price methods), with less computing
effort (Abramson et al. 2002). Implementation of the Bishop method of slices to compute factor
of safety and location of the critical circular failure surface for a slope requires use of specialized
computer software. There exist various commercial packages of this type, e.g., GEO05 (Fine Inc.
2016), SLIDE (Rocscience Inc. 2015), SLOPE/W (Geo-Slope Inc. 2012), XSTABL (Interactive
Software Designs Inc. 2007), among others.
Considering that computational-intensive methods are required for the determination of
factor of safety and location of critical failure surface, authors have proposed dimensionless
charts as a means of summarizing and conveying the results of slope stability analyses (e.g.,
Taylor 1948; Bishop & Morgenstern 1960; Spencer 1967; O’Connor & Mitchell 1977;
Michalowski 2002; Baker 2003).
Among the existing stability charts obtained from limit equilibrium models, the ones
presented in the book by Hoek & Bray (1974, 1977, 1981) (also, Wyllie & Mah 2004 and Wyllie
2018) deserve particular attention. These stability charts, which apply to slopes in rock masses
assumed to obey the Mohr-Coulomb shear failure criterion, use a particular form of
dimensionless scaling originally proposed by Bell (1966) that allows a compact representation of
factor of safety for slopes. Since then, other authors have used the scaling rule proposed by Bell
(1966) to produce dimensionless charts (e.g., Cousins 1978; Michalowski 2002; Steward et al.
2010).
This paper extends the dimensionless representations presented in Hoek & Bray (1981),
focusing in the problem of soil embankment stability, with particular regard to the existence of a
firm horizon beneath the embankment that the failure surface cannot cut through. Two limiting
cases of embankments are considered in this paper, one in which the firm horizon (or firm
foundation) is located at the base of the embankment, and the other in which the firm foundation
is located at great depth, and therefore the case becomes that of a slope built or cut in a soil mass.
The developments presented in this paper build on and extend the developments in a previous
paper (Carranza-Torres & Hormazabal 2018) which re-evaluated and extended the dimensionless
representations originally proposed by Hoek & Bray (1981) for the general case of slopes in rock
masses, without consideration of the existence of a tension crack. Based on Carranza-Torres &
Hormazabal (2018), this paper provides algebraic equations obtained from regression analysis of
factors of safety for embankments obtained with limit equilibrium models. It also provides a
computer EXCEL workbook (Microsoft 2016) that can be conveniently used to determine factors
of safety and position of the critical (assumed circular) failure surfaces for the two limiting cases
of embankments mentioned above.
© ASCE
Geo-Congress 2020
PROBLEM STATEMENT
The problem considered in this paper is shown in Figure 1. An embankment is built of an
assumed homogeneous, isotropic and dry soil of unit weight γ, that obeys the Mohr-Coulomb
shear failure criterion, and that is characterized by a cohesion c, an internal friction angle φ , and
null tensile strength. On the plane of analysis, the embankment is wide enough so that
determining the stability of the embankment involves analyzing the stability of one of the faces
only. According to Figure 1, the face is assumed to have an inclination angle α and height H. A
system of cartesian coordinates (x, y) is assumed to have its origin at the toe of the embankment
(point O in Figure 1).
Two cases are considered with regard to the position of a horizontal firm foundation. The
first case corresponds to the firm foundation located at the embankment base, as represented in
Figure 1. The second case corresponds to the firm foundation located at great depth, with the
case now corresponding to that of a slope built or excavated in homogeneous soil.
When the shear strength of the soil is affected by a scalar factor of safety, the embankment in
Figure 1 is assumed to be in a limit state of equilibrium with a critical circular failure surface of
radius R and center of coordinates xc and yc. Because the firm foundation represented in Figure 1
prevents the failure surface from extending beyond the embankment base, the surface is
comprised of the arcs A−Aʹ and Cʹ −Bʹ, and the linear segments Aʹ −Cʹ and Bʹ −B (the latter
associated with the development of a tension crack). Therefore, for a firm foundation at the
embankment base, the failure surface follows the path A − Aʹ −Cʹ − Bʹ − B, with the coordinates
of the points being (xA, yA), (xAʹ , yAʹ ), (xCʹ , yCʹ ), etc. This critical failure surface is referred to as
the critical secant failure surface in the legend of Figure 1.
In Figure 1, part of another critical circular failure surface referred to as critical tangent
failure surface is represented with a discontinuous outline. This is the tangent failure surface that
has been traditionally considered in the literature (Taylor 1948; Bishop & Morgenstern 1960;
Morgenstern 1963; Michalowski 2002). Steward et al. (2010) showed that a secant critical
failure surface yields a lower factor of safety compared with a tangent one. Based on Steward et
al. (2010) and on full numerical models presented in the ‘Application Example’ section, the
critical failure surface considered for the case of embankments in this paper is the one that
follows the path A − Aʹ −Cʹ − Bʹ − B. Figure 1 also indicates with discontinuous outline, a
deeper failure surface that corresponds to the case in which the firm foundation is at great depth.
This is the same critical failure surface that has been considered in Carranza-Torres &
Hormazabal (2018), although without the inclusion of a tension crack.
With regard to the shape of failure surface in Figure 1, some authors have used log-spiral
failure surfaces since such surfaces yield lower values of factor of safety compared with circular
ones (Michalowski 2002; Michalowski 2013; Utili 2013; Utili & Abd 2016). The reason of
adopting circular failure surfaces in this paper is that the commercial limit equilibrium software
SLIDE, used to develop the computational tools, does not implement log-spiral failure surfaces.
According to the method of slices, when the embankment or slope is at a limit state of
equilibrium, the soil shear strength is fully mobilized on the failure surface (Coduto et al. 2011;
Verruijt 2012; Das & Sobhan 2018). For the case of the Mohr-Coulomb soil considered in this
paper, the shear strength is written as follows
 s   n tan  c (1)
where τs and σn are the shear and normal stresses, respectively, at the base of an arbitrary slice on
the failure surface, and φ and c are the internal friction angle and cohesion of the soil,
© ASCE
Geo-Congress 2020
respectively.
Figure 1. Problem of determining the factor of safety and position of the critical failure
surface for the face of an embankment or slope in cohesive-frictional soil. A firm foundation
that the failure surface cannot cut through is considered to be either at the base of the
embankment or at great depth. A tension crack is allowed to develop at the crest.
With regard to equation (1), there is abundant bibliography showing that soils are
characterized by a stress dependent shear failure envelope —i.e., the relationship in equation (1)
is non-linear, rather than linear (e.g., Scott 1994; Terzaghi et al. 1996; Mesri & Shahien
2003).Since the computational tools presented in this paper were originally developed for
teaching basic slope stability analysis in an undergraduate soil mechanics course, the authors
considered more convenient to use a simpler Mohr-Coulomb linear soil model.
The factor of safety (FS) is the ratio of the shear strength of the soil on the failure surface and
the shear stress required for equilibrium (e.g., Abramson et al. 2002; Duncan et al. 2014). Of the
various formulations available for computing the factor of safety, the Bishop Method (as
implemented in the software SLIDE) is employed in this paper. The reader is referred to the
widely available literature for more details about the implementation of the standard limit
equilibrium method used in this paper (e.g., Mostyn & Small 1987; Duncan 1996; Abramson et
al. 2002; Duncan et al. 2014; Huang 2014).
DIMENSIONLESS REPRESENTATIONS OF STABILITY RESULTS

In this section and in the sections that follow, the case with the firm foundation located at the
embankment base (Figure 1) will be simply referred to as the embankment case, while the case
with the firm foundation located at great depth will be referred to as the slope case.
For both, embankment and slope cases, and following the analysis in Carranza-Torres &
Hormazabal (2018), all three mechanical variables (γ, φ and c) and the single length-dimension
variable (H) will be grouped into a single dimensionless factor (X ) as follows
© ASCE
Geo-Congress 2020
 H tan 
X (2)
c
According to Bell (1966) (also, Hoek & Bray 1974; Cousins 1978; Michalowski 2002;
Steward et al. 2010), the factor of safety divided by the tangent of the internal friction angle of
the soil (tan φ ) will depend on X and on the angle of the embankment or slope face (α) only, i.e.,
FS
 f FS  X ,   (3)
tan 
In equation (3), fFS represents two distinct functions, one for the embankment case and the
other for the slope case, that can be traced by solving a series of selected embankment and slope
cases for properly chosen values of the variables X and α.
Additionally, and according to Carranza-Torres & Hormazabal (2018), all ratios of variables
that characterize the position of the critical circular failure surface in Figure 1, will depend on X
and α only. For example, the scaled abscissa and ordinate of the center of the critical circular
failure are defined, respectively, as follows
xc
 f xc  X ,   (4)
H
and
yc
 f yc  X ,   (5)
H
The scaled abscissa and ordinate of the starting point of the critical circular failure surface
(point A in Figure 1) are defined, respectively, as follows
xA
 f xA  X ,   (6)
H
and
yA
 f yA  X ,   (7)
H
Equations similar to the equations (4) through (7) can be stated for all other points defining
the critical failure surface in Figure 1 (i.e., points Aʹ, Cʹ, Bʹ, and B).
As with the case of the function fFS in equation (3), the functions f xc , f yc , f xA and f y A in
equations (4) through (7) (and the similar equations for the remaining points of the critical failure
surface) can be reconstructed solving a series of selected embankment and slope cases for
properly chosen values of the variables X and α. Then, once these functions have been defined,
the ratio of the radius of the circular arcs forming the critical failure surface, and the height of the
embankment or slope (i.e., the ratio R/H) can be computed as the distance between the points C
and A in Figure 1, with the coordinates of these points expressed also as ratios of the
embankment or slope height H, i.e.
2 2
R x x  y y 
  c  A   c  A  (8)
H H H  H H 
To reconstruct the functions of the dimensionless variables in equations (3) through (7), the
limit equilibrium software SLIDE (Rocscience Inc. 2015) was employed. A total of 6, 804 cases
of embankments and slopes were evaluated using the Bishop method of slices. The input
variables in the models were chosen so to obtain 81 equally spaced (in logarithm base-10 scale)
embankment and slope cases with the factor X between 10−2 and 100. Face inclination angles
between 20◦ and 80◦, in increments of 10◦, were chosen. The factors of safety and corresponding
© ASCE
Geo-Congress 2020
critical circular failure surfaces were determined for all cases using the ‘Autorefine Search’
option implemented in the software SLIDE. For the cases considering the firm foundation at the
base of the embankment, the option ‘Composite Surface’ in the software was used. The option to
create a tension crack (using default parameters) was also specified. For full details of the
characteristics of the limit equilibrium models, the reader is referred to Carranza-Torres &
Hormazabal (2018).
Figure 2. Dimensionless stability diagram for the estimation of scaled factor of safety for
soil embankments with a firm foundation at the base, as represented in Figure 1.
Figure 2 shows the graphical representation of the function fFS in equation (3), as obtained
with SLIDE, for the embankment case. The diagram defines the relationship between the scaled
factor of safety FS/ tan φ (vertical axis) and the ratio X (horizontal axis), for different face angles
α. The small dots on the curves represent the computed limit equilibrium models. In Figure 2, as
X decreases, the soil is predominantly cohesive, while as X increases, it is predominantly
frictional. The figure also shows the positive influence of the cohesion of the soil on the factor of
safety: for a fixed value of friction angle, the factor of safety increases with the increase in
cohesion. It also shows the effect of the embankment face angle on the factor of safety: for a
fixed value of the factor X , the factor of safety decreases with increase of the face angle.
Figure 3 shows the relationship between scaled factor of safety FS/ tan φ and the
dimensionless variable X , for different angles α, for the slope case. The interpretation of Figure
3 is similar to that of Figure 2. A visual comparison of both figures indicates that the curves in
Figure 3 are shifted downwards with respect to the curves in Figure 2, in particular, for values of
face angles smaller than ∼ 50◦. This implies that the factor of safety of an embankment with a
firm foundation at the base will always have a factor of safety that is greater than or equal to the
factor of safety of the same embankment when the firm foundation is at great depth. In other
© ASCE
Geo-Congress 2020
words, the stability of the embankment decreases (or remains the same, for embankments with
face angles above ∼ 40◦) with the increase of the depth of the firm foundation.
To quantify the difference in factors of safety for embankments and slopes, the ratio rSE of
the respective factors of safety is defined as follows,
FS slope
rSE  (9)
FSembk
Figure 4 shows the graphical representation of the ratio rSE and the dimensionless variable
X, for the different values of face angle α considered in Figures 2 and 3. The figure shows that
the decrease in stability conditions when transitioning from embankments to slopes depends on X
and α only. The lower the values of X and α, the larger the drop in factor of safety. Also, for
values of X larger than ∼ 1 and α larger than 50◦, the factors of safety for embankments and
slopes are the same. Referring to Figure 4, the discontinuity in the slope of the curve α = 40◦, in
the vicinity of the abscissa X = 1, is related to the development of the tension crack in the SLIDE
models from which the curves were constructed.
Figure 3. Dimensionless stability diagram for the estimation of scaled factor of safety for
slopes in soils (i.e., the case in which the firm foundation in Figure 1 is at great depth).
To quantify the influence of the position of the firm foundation on factors of safety, the point
P represented in Figures 2, 3 and 4, corresponds to embankment and slope cases for which X is 1
and α is 20◦. From Figure 2, the scaled factor of safety for the embankment case is 13.15,
whereas from Figure 3, the scaled factor of safety for the slope case is 11.43. From Figure 4, the
decrease in the scaled factor of safety is ≈ 0.87.
For space reasons, no diagrams for the functions defining the different points that form the
critical circular failure surface in Figure 1 (equations 4 through 7) are explicitly included in this
© ASCE
Geo-Congress 2020
paper. The diagrams can be constructed using the coordinates of the critical failure surface points
that are included in the EXCEL workbook to be discussed in the section ‘Computer Spreadsheet
for Analysis of Stability’ (the EXCEL workbook also includes coordinates of all the points
represented in Figures 2 and 3).
Figure 4. Ratio of factors of safety obtained with equation (9), for cases of slopes (firm
foundation at great depth) and embankments (firm foundation at the embankment base).
MECHANICAL SIMILARITY OF EMBANKMENT AND SLOPE PROBLEMS

The concept of mechanical similarity of slope cases, with particular reference to mechanical
similarity of scaled factors of safety and scaled position of the critical failure surface, has been
introduced in Carranza-Torres & Hormazabal (2018). This section summarizes the concept,
focusing on the expected decrease in stability conditions when transitioning from embankment
cases (firm foundation at the base of the embankment) to slope cases (firm foundation at great
depth).
Table 1. Geometric characteristics and physical/mechanical soil properties for

embankments or slopes that yield the same dimensionless factor X .
Case α [◦] H [m] γ [kN/m3] φ [◦] c [kPa] X = γH tan φ /c [-]
a 10.0 17.0 10.0 30.0
b 19.1 15.7 30.9 179.5
c 20 20.6 16.5 5.0 29.7 1.00
d 33.2 17.4 32.5 368.0
e 42.4 18.1 21.1 296.1
© ASCE
Geo-Congress 2020
Five different cases of embankments or slopes, as listed in Table 1, are considered. At first
sight, and disregarding the fact that all cases have the face angle (α = 20◦), these correspond to
quite different embankment or slope heights, and soil properties. Nevertheless, according to the
last column in Table 1, all cases are characterized by the same dimensionless variable X (equal to
1.00). Therefore, according to equations (3) through (8), the cases are expected to have the very
same values of scaled factor of safety (FS/ tan φ ) and scaled coordinates of the different points
that form the critical failure surface (i.e., the scaled coordinates of points A, Aʹ, Cʹ, etc. in Figure
1). To illustrate this, Table 2 lists the values of scaled factor of safety and scaled coordinates of
the center, starting point and radius of the critical failure surface, for all cases, as obtained with
equations (4) through (8) and with the EXCEL workbook to be discussed in the next section. The
values listed in Table 2 result the same for all five considered cases.
Table 2. Scaled factor of safety and scaled coordinates that define the position of the
assumed critical failure surface, as obtained with the computational tools in this paper.
Cases Type FS/ tan φ [-] xC/H [-] yC/H [-] xA/H [-] yA/H [-] R/H [-]
a, b, c, d, e Embk. 13.15 1.17 2.60 0.25 0.09 2.67
a, b, c, d, e Slope 11.43 1.13 2.15 -0.38 0.00 2.63
Note: For space reasons, the scaled coordinates of the remaining points defining the critical failure surface shown in
Figure 1 are not listed in the table; those values result also the be the same in all five cases.
The concept of mechanical similarity illustrated by the different cases in Tables 1 and 2,
allows explanation of an observation that dates back to the early developments of computational
methods for slope stability analysis in soil mechanics: embankments and slopes in predominantly
cohesive soils tend to show deeper critical failure surfaces than those in predominantly frictional
soil. Indeed, it has also been known that for slopes in purely frictional soil, the thickness of the
critical failure surface becomes null, as proved by the analytical solution of an infinite slope in
cohesionless soil (Scott 1994; Abramson et al. 2002; Coduto et al. 2011; Das & Sobhan 2018). It
will be shown next that it is the dimensionless parameter X (equation 2) that controls how deep
or shallow the critical failure surface is.
The diagrams in Figures 5a and 5b are equivalent to the diagrams already introduced in
Figures 2 and 3. For clarity, only three curves corresponding to embankment and slope face
angles equal to 20◦, 40◦ and 60◦ are included in the diagrams. The different sketches above the
curves in Figures 5a and 5b correspond to embankment and slope cases in predominantly
cohesive soil (characterized by an arbitrarily chosen value X = 0.05), cases in a regular cohesive-
frictional soil (arbitrarily chosen value X = 1), and cases in a predominantly frictional soil
(arbitrarily chosen value X = 25). The sketches show the shape of the critical circular failure
surface as obtained with the EXCEL workbook described in the next section. It is seen that when
the angle of the embankment or slope face is fixed, the average thickness of the critical failure
surface depends on the dimensionless parameter X only. Finally, it should be noted that all five
cases listed in Tables 1 and 2 plot as points B20 in Figures 5a and 5b. Therefore, the sketches
that correspond to case B20 in Figures 5a and 5b are representative of the shape of the critical
circular failure surface for all five cases listed in Tables 1 and 2.
COMPUTER SPREADSHEET FOR ANALYSIS OF STABILITY

Although dimensionless charts have been the traditional way of summarizing and presenting
results of stability analyses, in modern geotechnical design practice, engineers typically prefer
© ASCE
Geo-Congress 2020
the use of computer spreadsheets over the use of these dimensionless charts (Hoek 2000).
Figure 5. Dimensionless stability diagrams showing similarity of scaled factor of safety and
critical failure surface location. The diagrams correspond to cases of a) embankments, with
a firm foundation at the base; and b) slopes, with a firm foundation at great depth.
© ASCE
Geo-Congress 2020
Figure 6. EXCEL workbook for implementation of stability computations. a) Main

worksheet. b) Graphical representation of Case 2 (i.e., slope in soil).
With this idea in mind, the results of the limit equilibrium models that allowed the functions
© ASCE
Geo-Congress 2020
in equations (3) through (7) to be reconstructed by discrete points, have been summarized in an
EXCEL workbook (Microsoft 2016) called ‘Slope Stability Calculator for Embankments in
Frictional-Cohesive Soils’. Figure 6a shows a view of the main spreadsheet where input data is
entered and results are displayed, while Figure 6b shows a view of the spreadsheet that provides
a graphical representation of the problem (in the case shown, the slope problem), together with a
summary of input data and results.
The EXCEL file corresponding to the workbook introduced above can be freely downloaded
from the first author’s web site, at www.d.umn.edu/∼carranza/EMBK20. At this web site,
together with the EXCEL workbook file, the reader will find a ‘PDF’ copy of a manuscript
describing the EXCEL workbook in detail.
FACTOR OF SAFETY EQUATIONS FOR EMBANKMENTS AND SLOPES

Following previous developments in Carranza-Torres & Hormazabal (2018), a multiple
regression analysis was carried out using the results of scaled factors of safety represented in
Figures 2 and 3, for embankments and slopes, respectively. The objective of the analysis was to
come up with algebraic equations to predict values of scaled factor of safety (FS/ tan φ ) as a
function of the dimensionless factor X and the angle α, which could be used as additional tools to
quickly estimate values of factors of safety.
The adjusting function was chosen to be a power inverse function of the variable X of the
following form,
FS 1 g   g 2   
  1  g   (10)
tan  tan  X X 3
The reason for choosing an inverse function was to be able to recover the solution of an
infinite slope for the case of purely frictional soil. Indeed, when X tends to infinity, all different
curves in Figures 2 and 3 will become asymptotic towards the scaled factor of safety predicted
by the analytical solution of an infinite slope in purely frictional soil, i.e., FS/ tan φ = 1/ tan α
(e.g., Scott 1994; Abramson et al. 2002; Coduto et al. 2011; Das & Sobhan 2018).
In equation (10), the functions g1(α), g2(α) and g3(α) are cubic polynomials that are assumed
to have a discontinuity of the first and higher derivatives at α = 50◦, the approximate value of
slope angle above which the critical circular failure surface always start at the toe of the slope
(i.e., at point 0 in Figure 1). Using standard methods of minimization of sum of square of the
estimate residuals (Chapra & Canale 2015), the functions g1(α), g2(α) and g3(α) were found to
be as follows:
For the case of embankments,
If α ≤ 50°
g1 ( )  5.430  9.180 102 (  50)  1.904 103 (  50) 2  9.963 10 5 (  50)3
g 2 ( )  1.333  7.563 103 (  50)  4.674 104 (  50)2  5.87110 6 (  50)3 (11)
g3 ( )  3.682 10  3.068 10 (  50)  3.84110 (  50)  9.460 10 (  50)
1 4 5 2 7 3
If α ≥ 50°
g1 ( )  5.430  6.067 10 2 (  50)  2.088 10 3 (  50) 2  7.738 10 5 (  50)3
g 2 ( )  1.333  1.72110 2 (  50)  4.454 103 (  50) 2  1.217 10  4 (  50)3 (12)
g3 ( )  3.682 101  7.698 103 (  50)  1.419 103 (  50)2  3.664 105 (  50)3
For the case of slopes,
© ASCE
Geo-Congress 2020
If α ≤ 50°
g 1 ( )  5.346  8.603 103 (  50)  3.221104 (  50) 2  1.335 10 7 (  50)3
g 2 ( )  1.414  3.435 102 (  50)  1.656 10 3 (  50) 2  2.175 10 5 (  50)3 (13)
g 3 ( )  3.841101  6.840 103 (  50)  4.102 105 (  50) 2  3.573 10 6 (  50)3
If α ≥ 50°
g 1 ( )  5.346  4.524 102 (  50)  1.247 10 3 (  50) 2  6.336 10 5 (  50)3
g 2 ( )  1.414  2.321103 (  50)  3.642 103 (  50) 2  1.082 10 4 (  50)3 (14)
g 3 ( )  3.841101  4.771103 (  50)  1.259 10 3 (  50)2  3.398 105 (  50)3
It is important to note that the regression analysis considers that the embankment or slope
face angle (α) in the equations above is expressed in degrees and not in radians.
At the same web site where the Excel workbook introduced in the previous section is posted
(www.d.umn.edu/∼carranza/EMBK20), the reader will find a ‘PDF’ copy of a manuscript
including a detailed analysis of factor of safety errors associated with the application of the
equations above.
APPLICATION EXAMPLE
This section presents an application example of stability analysis of an embankment that
illustrates the use of the computational tools discussed in previous sections.
The problem involves computing the factor of safety and location of the critical circular
failure surface for the face of the embankment in Figure 1 with geometry and soil properties as
follows:
- Height, H = 10 m
- Face inclination angle, α = 20◦
- Unit weight of the soil, γ = 17 kN/m3
- Cohesion of the soil, c = 30 kPa
- Internal friction angle of the soil, φ = 10◦
Cases in which the firm foundation is located at the base of the embankment, and in which
the firm foundation is located at great depth will be considered separately. Following the
notation of the EXCEL workbook introduced in an earlier section, the cases will be referred to as
Case 1 (or embankment case) and Case 2 (or slope case), respectively.
According to the ‘Dimensionless Representations of Stability Results’ section, the factor of
safety and location of critical failure surface depends on the dimensionless factor X , and on the
face angle α (in this case, 20◦). For the input values listed above, the dimensionless factor X
results to be
 H tan  17 kN / m3 10 mtan10
X   1.0 (15)
c 30 kPa
The embankment problem introduced above corresponds to the Case a in the Tables 1 and 2
and to the values listed in the EXCEL workbook (Figure 6). Therefore, from Table 2 and Figure
6a, the computational tools introduced in previous sections give the following values of factor of
safety and coordinates of the points that define the critical failure surface:
For Case 1 (or embankment case):
- Scaled factor of safety, FS/ tan φ = 13.15 (FS = 2.32 for φ = 10°)
© ASCE
Geo-Congress 2020
- Scaled abscissa of the center of the critical failure surface, xc/H = 1.17 (xc = 11.7 m for H
= 10 m)
- Scaled ordinate of the center of the critical failure surface, yc/H = 2.60 (yc = 26.0 m for H
= 10 m)
- Scaled abscissa of the starting point of the critical failure surface, xA/H = 0.25 (xA = 2.5
m for H = 10 m)
- Scaled ordinate of the starting point of the critical failure surface, yA/H = 0.09 (yA = 0.9
m for H = 10 m)
- Scaled radius of the critical failure surface, R/H = 2.67 (R = 26.7 m for H = 10 m) For
Case 2 (or slope case):
- Scaled factor of safety, FS/ tan φ = 11.43 (FS = 2.02 for φ = 10°)
- Scaled abscissa of the center of the critical failure surface, xc/H = 1.13 (xc = 11.3 m for H
= 10 m)
- Scaled ordinate of the center of the critical failure surface, yc/H = 2.15 (yc = 21.5 m for H
= 10 m)
- Scaled abscissa of the starting point of the critical failure surface, xA/H = −0.38 (xA =
−3.8 m for H = 10 m)
- Scaled ordinate of the starting point of the critical failure surface, yA/H = 0.0 (yA = 0.0 m
for H = 10 m)
- Scaled radius of the critical failure surface, R/H = 2.63 (R = 26.3 m for H = 10 m)
Figure 7. Geometry and mesh of elements for the embankment model solved with the finite
difference software FLAC (Itasca 2016).
From the resulting values of factor of safety, it is seen that Case 1 and Case 2 are stable (i.e.,
they both have a factor of safety larger than one), and according to the ‘Dimensionless
Representations of Stability Results’ section, Case 1 has a larger factor of safety than Case 2.
Cases 1 and 2 have also been solved with the shear strength reduction technique using the
finite different software FLAC (Itasca, Inc. 2016). Figure 7 shows the mesh of elements used for
the models. The mesh shown in Figure 7 corresponds to Case 1, for which the elements below
the indicated ‘firm foundation level’ were considered to be elastic.
Figures 8a and 8b show the resulting contours of normalized shear strain computed with the
shear strength reduction technique for the models of Case 1 and 2, respectively (these models
correspond to models that are at the verge of failure). When applying the shear strength reduction
© ASCE
Geo-Congress 2020
technique, the contour representation of shear strain is normally used to define the critical failure
surface. The words ‘shear strain’ in the legend of Figures 8a and 8b refer to the square root of the
second invariant of the deviatoric strain rate, while the word ‘normalized’ refers to the fact that
the shear strain rates have been divided by the maximum value of shear strain rate in the model
at the shown state.
Figures 8a and 8b also include the outlines of the critical failure surfaces obtained with the
EXCEL workbook —i.e., as obtained with the software SLIDE (Rocscience Inc. 2015). The
similarity of the critical failure surfaces obtained with shear strength reduction technique and
with the limit equilibrium method is remarkable. In particular, for Case 1 (the embankment
case), and according to Steward et al. (2010), the failure surface given by FLAC is secant to the
firm foundation and not tangent as it has been traditionally considered in the literature (Taylor
1948; Bishop & Morgenstern 1960; Morgenstern 1963; Michalowski 2002).
The values of factor of safety obtained with the shear strength reduction technique in the
software FLAC, and with the limit equilibrium method in the EXCEL workbook are indicated on
the left side of Figures 8a and 8b. The shear strength reduction technique yields a factor of safety
that is comparatively larger than the one obtained with limit equilibrium for Case 1 (Figure 8a),
while the opposite is true for the Case 2 (Figure 8b). If the true values of factors of safety for the
embankment and slope cases are assumed to be the mean values of factors of safety obtained
with FLAC and EXCEL (a reasonable assumption considering that there is no exact solution for
the problem of determining factors of safety), then the normalized errors of the factors of safety
for each method result as follows:
For Case 1 (or embankment case):
- FLAC, (2.40 − (2.40 + 2.32)/2) /((2.40 + 2.32)/2) = 1.69%
- EXCEL (SLIDE), (2.32 − (2.40 + 2.32)/2) /((2.40 + 2.32)/2) = −1.69%
For Case 2 (or slope case):
- FLAC, (1.94 − (1.94 + 2.02)/2) /((1.94 + 2.02)/2) = −2.02%
- EXCEL (SLIDE), (2.02 − (1.94 + 2.02)/2) /((1.94 + 2.02)/2) = 2.02%
The values listed above indicate that the normalized errors are within approximately ±2% for
both Cases 1 and 2, a relative small value that confirms that values of factor of safety obtained
with limit equilibrium method and shear strength reduction technique can be expected to be
similar for the case of simple embankments and slope cases (Cheng et al. 2007).
Further to the FLAC models discussed above, the results represented in Figure 8 correspond
to models with a non-associated flow rule with no dilation. For the particular properties in this
application example, the very same factors of safety and contours of shear strain, as represented
in Figure 8, were obtained when considering an associated flow rule with a dilation angle equal
to the friction angle.
FINAL COMMENTS
The computational tools presented in this paper include graphical dimensionless
representations, an EXCEL workbook and regression equations that can be used to quickly
determine factors of safety and location of critical circular failure surfaces for simple
embankments or slopes in dry homogeneous and isotropic soils. Although these tools were
developed for teaching basic slope stability analysis in an undergraduate soil mechanics course,
the tools could be useful in geotechnical design practice, for example in the pre-design stage of
simple embankments or slopes. The tools could also be useful when performing probability and
reliability analyses of stability of embankments and slopes using the Monte-Carlo simulation
© ASCE
Geo-Congress 2020
technique. In such cases, thousands of cases need to be analyzed to estimate factors of safety for
different input parameters (e.g., Vanmarcke 1980; Juang et al. 1998; Wang et al. 2013; Lacasse
et al. 2019) and computational tools that allow quick estimation of factors of safety are desirable.
The computational tools presented in this paper will not be applicable to cases of
embankments or slopes in purely cohesive soil. Indeed, if the internal friction angle of the soil is
equal to zero, the dimensionless parameter X in equation (2) becomes undetermined. For the case
of embankments or slopes in purely cohesive soil, a new set of tools would be required, with
particular consideration that the position of the firm foundation dictates the extent in depth of the
critical failure surface. For example, the case of a firm foundation located at infinite depth will
always yield an infinitely deep failure surface when the soil is purely cohesive (e.g.,
Michalowski 2002; Baker 2003; Steward et al. 2010).
Figure 8. Critical failure surfaces outlined by contours of shear strain rates given by the
software FLAC, and as explicitly defined by the EXCEL workbook. Plots correspond to
Case 1 (embankment with firm foundation at the base) and b) Case 2 (slope, with firm
foundation at great depth).
© ASCE
Geo-Congress 2020
The adoption of the limit equilibrium method and the Bishop formulation used to develop the
tools presented in this paper was due to practical reasons only. Use of commercial limit
equilibrium software is still very much spread in the field of geotechnical engineering and the
authors needed to generate and solve thousands of models and process the results in a reasonable
amount of time. Similar representations could have been developed with limit analysis and
possibly shear strength reduction methods —although in the latter case, an impractical amount
of time would have been required to process all models needed to develop these tools. For simple
embankment and slope cases considered in this paper (i.e., embankments and slopes with a
planar face in homogeneous, isotropic, frictional-cohesive soil) no significant differences in

values of factor of safety or location of the critical failure surface could be expected when
applying other methods, like limit analysis or shear strength reduction methods (e.g., Yu et al.
1998; Dawnson et al. 1999; Davis & Selvadurai 2005; Cheng et al. 2007; Leshchinsky 2013).
Due to the non-existence of a rigorous closed-form solution for the problem of computing the
factor of safety and location of the critical failure surface for embankments and slopes, all of the
methods available to carry out stability analyses will be expected to give an approximate solution
to the problem only, and small differences in results of factor of safety and location of the critical
failure surface should be always expected, as illustrated in the section ‘Application Example’ in
this paper.
ACKNOWLEDGEMENTS
The authors wish to dedicate this paper to their geotechnical engineering professors at the
University of Minnesota, Twin Cities Campus, and at University of Michigan, Ann Arbor,
respectively. The authors acknowledge that they would not have decided to start a university
career to teach geotechnical engineering to young civil engineering students, without the
motivation and inspiration received from all outstanding geotechnical engineering professors at
the mentioned institutions.
REFERENCES
Abramson, L. W., Lee, T. S., Sharma, S., M., G. & Boyce (2002). Slope stability and
stabilization methods (2nd ed.). John Wiley & Sons. New York.
Baker, R. (2003). A second look at Taylor’s stability chart. J. Geotech. Geoenviron. ASCE
129(12), 1102–1108.
Bell, J. M. (1966). Dimensionless parameters for homogeneous earth slopes. Journal of Soil
Mechanics & Foundations Div. ASCE SM 5(GT9), 51–65.
Bishop, A. W. (1955). The use of the slip circle in the stability analysis of slopes. Géotechnique
5(1), 7–17.
Bishop, A. W. & Morgenstern, N. R. (1960). Stability coefficients for earth slopes. Géotechnique
10(1), 129–150.
Carranza-Torres, C. & Hormazabal, E. (2018). Computational tools for the determination of
factor of safety and location of the critical failure surface for slopes in Mohr-Coulomb dry
ground. In Proceedings of the International Symposium Slope Stability 2018. April 11-13,
2018, Sevilla. Spain. (Available for downoading at www.d.umn.edu/∼carranza/SLOPE18).
Chapra, S. C. & Canale, R. P. (2015). Numerical methods for engineers (7th ed.). Mc Graw Hill.
New York.
Cheng, Y. M., Lansivaara, T. & Wei, W. B. (2007). Two-dimensional slope stability analysis by
limit equilibrium and strength reduction methods. Computers and Geotechnics 34(3), 137–
© ASCE
Geo-Congress 2020
150.
Coduto, D. P., Yeung, M. C. & Kitch, W. A. (2011). Geotechnical engineering. Principles and
practices (2nd ed.). Pearson.
Cousins, B. F. (1978). Stability charts for simple earth slopes. Journal of the Geotechnical
Engineering Division 104(2), 267–279.
Das, B. M. & Sobhan, K. (2018). Principles of Geotechnical Engineering (9th ed.). Cengage.
Davis, R. O. & Selvadurai, A. P. (2005). Plasticity and geomechanics. Cambridge University
Press.
Dawnson, E. M., Roth, W. H. & Drescher, A. (1999). Slope stability analysis by strength
reduction. Geotechnique 49(6), 835–840.
Duncan, J. M. (1996). State of the art: limit equilibrium and finite-element analysis of slopes.
Journal of Geotechnical Engineering 122(7), 577–596.
Duncan, J. M., Wright, S. G. & Brandon, T. L. (2014). Soil Strength and slope stability. John
Wiley & Sons.
Fine Inc. (2016). GEO5. Slope stability analysis software based on the limit equilibrium method.
Praha, Czech Republic.
Geo-Slope Inc. (2012). SLOPE/W Version 2012. Slope stability analysis software based on the
limit equilibrium method. Calgary, Canada.
Griffiths, D. V. & Lane, P. A. (1999). Slope stability analysis by finite elements. Geotechnique
49(3), 387–403.
Hammah, R. E., Yacoub, T. E. & Curran, J. H. (2007). Serviceability-based slope factor of safety
using the shear strength reduction (SSR) method. In L. R. e Sousa, C. Olalla, & N.
Grossmann (Eds.), Proceedings of the 11th Congress of the International Society for Rock
Mechanics, Lisbon.
Hoek, E. (2000). Personal communication.
Hoek, E. & Bray, J. (1974, 1977, 1981). Rock slope engineering. Institution of Mining and
Metallurgy. London. (First, second and third editions, respectively).
Huang, Y. H. (2014). Slope stability analysis by the limit equilibrium method. Fundamentals and
methods. American Society of Civil Engineers (ASCE) Press. Reston, Virginia.
Interactive Software Designs Inc. (2007). XSTABL. Slope stability analysis software based on the
limit equilibrium method. Moscow, Idaho.
Itasca, Inc. (2016). FLAC (Fast Lagrangian Analysis of Continua) Version 8.0. Minneapolis,
Minnesota: Itasca Consulting Group, Inc.
Janbu, N. (1954)a. Application of composite slip surfaces for stability analysis. In European
Conference on Stability of Earth Slopes. Stockholm, Sweden, Volume 3, pp. 43–49.
Janbu, N. (1954)b. Stability analysis of slopes with dimensionless parameters. Thesis for the
Doctor of Science in the Field of Civil Engineering. Harvard University, Soil Mechanics
Series, No. 46.
Juang, C. H., Jhi, Y. Y. & Lee, D. H. (1998). Stability analysis of existing slopes considering
uncertainty. Engineering Geology 49(2), 111–122.
Lacasse, S., Nadim, F., Liu, Z. Q., Eidsvig, U. K., Le, T. M. H. & Lin, C. G. (2019). Risk
assessment and dams. Recent developments and applications. In Proceedings of the XVII
European Conference on Soil Mechanics and Geotechnical Engineering (ECSMGE-2019).
Geotechnical Engineering foundation of the future. Reykjavik, Iceland. September 2019.
Leshchinsky, B. (2013). Comparison of limit equilibrium and limit analysis for complex slopes.
In Geo-Congress 2013: Stability and Performance of Slopes and Embankments III, pp. 1280–
© ASCE
Geo-Congress 2020
1289.
Mesri, G. & Shahien, M. (2003). Residual shear strength mobilized in first-time slope failures. J.
Geotech. Geoenviron. ASCE 129(1), 12–31.
Michalowski, R. L. (2002). Stability charts for uniform slopes. J. Geotech. Geoenviron. ASCE
128-4(GT9), 351–355.
Michalowski, R. L. (2013). Stability assessment of slopes with cracks using limit analysis.
Canadian Geotechnical Journal 50, 1011–1021.
Microsoft (2016). Excel Software. Version 2016. Microsoft. Redmond, Washington.
Morgenstern, N. R. (1963). Stability charts for earth slopes during rapid drawdown.
Géotechnique 13(1), 121–131.
Morgenstern, N. R. & Price, V. E. (1965). The analysis of the stability of general slip surfaces.
Géotechnique 15(1), 77–93.
Mostyn, G. R. & Small, J. C. (1987). Methods of stability analysis. In B. F. Walker & R. Fell
(Eds.), Soil slope stability, pp. 71–120. Balkema. Rotterdam.
O’Connor, M. J. & Mitchell, R. J. (1977). An extension of the Bishop and Morgenstern slope
stability charts. Canadian Geotechnical Journal 14(1), 144–151.
Potts, D. M. & Zdravkovic, L. (1999). Finite element analysis in geotechnical engineering.
Theory. Thomas Telford. London: Thomas Telford.
Rocscience Inc. (2015). SLIDE Version 7. Slope stability analysis software based on the limit
equilibrium method. Toronto, Canada.
Scott, C. R. (1994). An Introduction to Soil Mechanics and Foundations (3rd ed.). Springer-
Science & Business Media, B.V.
Spencer, E. (1967). A method of analysis of the stability of embankments assuming parallel
interslice forces. Géotechnique 17, 11–26.
Spencer, E. (1973). Thrust line criterion in embankment stability analysis. Géotechnique 23, 85–
100.
Steward, T., Sivakugan, N., Shukla, S. K. & Das, B. M. (2010). Taylor’s slope stability charts
revisited. International Journal of Geomechanics 11(4), 348–352.
Taylor, D. W. (1948). Fundamentals of Soil Mechanics. Wiley. New York.
Terzaghi, K., Peck, R. B. & Mesri, G. (1996). Soil mechanics in engineering practice. John
Wiley & Sons.
Utili, S. (2013). Investigation by limit analysis on the stability of slopes with cracks.
Geotechnique 6(2), 140–154.
Utili, S. & Abd, A. H. (2016). On the stability of fissured slopes subject to seismic action.
International Journal for Numerical and Analytical Methods in Geomechanics 40, 785–806.
Vanmarcke, E. H. (1980). Probabilistic stability analysis of earth slopes. Engineering Geology
16(1-2), 29–50.
Verruijt, A. (2012). Soil Mechanics. Delft University of Technology, Delft, The Netherlands
(Available for dowloading at http://geo.verruijt.net/).
Wang, L., Hwang, J., Juang, H. & Atamturktur, S. (2013). Reliability-based design of rock
slopes. A new perspective on design robustness. Engineering Geology 154, 56–63.
Wyllie, D. C. (2018). Rock slope engineering. Civil Applications (5th ed.). CRC Press. Taylor &
Francis.
Wyllie, D. C. & Mah, C. (2004). Rock slope engineering. Civil and Mining (4th ed.). CRC Press.
Yu, H. S., Salgado, R., Sloan, S. W. & Kim, J. M. (1998). Limit analysis versus limit equilibrium
for slope stability. J. Geotech. Geoenviron. ASCE 124(1), 1–11.
© ASCE
Geo-Congress 2020
Eisenhower Bridge North Abutment and Approach Settlement: A Case History of Timber
Pile Downdrag and Comparative Downdrag Effect on Steel Piles
Steven J. Olson, P.E., M.ASCE1
1
HDR, Minneapolis, MN. E-mail: steve.olson@hdrinc.com
ABSTRACT
The north abutment and approach of the Eisenhower Bridge has settled significantly since
completion in 1960 and is currently several feet lower as compared to the 1960 construction plan
profile. The north approach originally consisted of a 25-foot-high embankment placed on a soil
profile of loose/soft alluvial sands, silts, and clays (alluvium) underlain by a very dense layer of
sands and gravels; the north abutment was supported by 26 timber piles driven through the
embankment fill and alluvium to practical refusal bearing on very dense sands and gravels.
Given the timber piles were driven to practical refusal in competent materials, the movement in
the abutment appears to be a result of downdrag loads. As the approach embankment adjacent to
the abutment settled from consolidation of the soft alluvium, drag load developed on the timber
piles and induced excessive internal stress and end bearing pressure on the piles. It is believed
the yield strength of the timber piles was exceeded in compression resulting in damaged or
broken piles. The purpose of this study is to demonstrate how the drag loads on the timber pile
foundation system caused excessive stress and movement of the abutment and provide
comparative analyses to model the effect if steel piles had been selected in place of the timber
piles.
INTRODUCTION
The Minnesota Department of Transportation (MnDOT) Bridge No. 9040 (Wisconsin
Department of Transportation Bridge No. B-47-24), referred to as the Eisenhower Bridge, is
currently under replacement as MnDOT Bridge No. 25033. The new bridge will be located just
upstream of the existing Eisenhower Bridge (see Figure 1) on Truck Highway No. 63 (TH 63)
between Red Wing, Minnesota, and Hager City, Wisconsin.
SITE GEOLOGY
At Red Wing, the geologic profile consists of bedrock near the surface (within 5 to 10 feet)
of elevation 720 feet. The surficial soils consist of alluvial sand, gravel and sandstone colluvium.
The bedrock nearest the surface and exposed at the south abutment is the St. Lawrence
Formation, which consists of intermixed siltstone and sandstone with some dolomitic zones.
Underlying the St. Lawrence Formation is the Franconia Formation that is variably glauconitic,
fine to medium grained sandstone with seams of shale, and zones where sandstone is cemented
with dolomite.
The bedrock surface is deeper north of Red Wing within the Mississippi River Valley and
heading into Wisconsin; it ranges from 85 feet to more than 145 feet below ground/water
surface, varying from approximately elevation 588 feet to 537 feet at the north abutment.
The south abutment at Red Wing is supported on a spread footing bearing on the St.
Lawrence Formation. Pier 1 is supported on driven H-piles through sandy alluvium with piles
bearing on the underlying sandstone. Piers 2 through 7, and the north abutment and approach of
the Eisenhower Bridge, are located within the Mississippi River Valley in Pierce County,
© ASCE
Geo-Congress 2020
Wisconsin. The Normal Pool Elevation of the Mississippi River is 666.64 feet at Red Wing. The
river valley is relatively low in elevation (about elevation 670 feet) and therefore required an
approach embankment with a height well over 25 feet (to elevation 695 feet) placed to support
the north abutment. The north abutment and approach act as a causeway over a seasonal
floodplain and have settled since original bridge completion in 1960. Figure 2 shows how the
existing abutment and approach is currently several feet lower as compared to the 1960 plan
profile provided by MnDOT.
Figure 1. The aerial view shows the existing Eisenhower Bridge.
Figure 2. Comparison of existing TH 63 profile to 1960 plan profile
© ASCE
Geo-Congress 2020
SUBSURFACE PROFILE
Prior to placement of the north approach embankment for the Eisenhower Bridge beginning
in April 1959, the subsurface profile (based on boring T-1) from surface elevation 666 feet
consisted of 9 feet of very soft native silts/clays, and then 12 feet of loose sands, underlain by
more than 50 feet of soft, compressible clay, overlying very dense gravels with sands. Boring T-
1 terminated within the gravel/sand layer. Young’s Modulus of the very dense gravel/sand layer
was conservatively estimated at 20 kips per square inch with Standard Penetration Test (N60)
values of 50 blows per 0.5 feet.

Deeper exploration in 2014 (boring T-109) revealed the very dense gravel/sand layer
supporting the timber piles to be about 10 feet thick and is underlain by medium dense coarse
sands that extend to sandstone bedrock of the Franconia Formation at elevation 537 feet. See
Wisconsin side (right hand side) of Figure 3 from MnDOT Memo (2014).
Figure 3. Subsurface profile along the bridge alignment

BRIDGE FOUNDATION CONSTRUCTION
The north bridge abutment and approach embankment was placed during the summer of
1959 to about final elevation 695 feet. Soon after embankment placement, the north abutment
piles were installed in December of 1959. The abutment foundation system consisted of 26
timber piles driven with a Vulcan 1-M steam hammer. The hammer has a ram weight of 5,000
pounds and available stroke of 3 feet for maximum rated strike energy of 15,000 feet-pounds.
The timber piles were generally 105 feet long and driven to an average length of 97 feet
below cutoff elevation at 689.95 feet. The measured pile diameter averaged about 15 inches at
the pile top and 7 inches at the base. The piles were driven to practical refusal at an average set
of 0.08 inches per blow (equivalent to 150 bpf). The timber piles extended about 2-3 feet into the
very dense sands and gravels to an average pile base elevation at 593 feet.
Based on the Engineering News Record (ENR) formula, the piles were driven to an average
bearing resistance of 101.6 tons (203 kips) per pile. The minimum required installation bearing
resistance was 20 tons per pile and calculated dead load plus overturning was 16 tons per pile, as
© ASCE
Geo-Congress 2020
indicated on the plan document notes. Therefore, the piles were installed with an average safety
factor of 5.0 or more in axial bearing support based on resistance at end of drive.
EMBANKMENT SETTLEMENT
The abutment and approach embankment construction required well over 25 feet of fill
bearing on the native alluvial river valley soils at the north abutment location. The weight of this
fill has induced vertical movement of the north approach as a result of long-term consolidation
settlement of the compressible alluvial soils, especially the 50-feet-thick layer of soft silty clay
loam immediately above the very dense loamy sands and gravels on which the timber piles are
bearing. To estimate the embankment settlement, the consolidation test results (Figure 4) from
MnDOT Memo (2014) were utilized.
Figure 4. Consolidation test results

Estimated embankment settlement was calculated to be about 50 inches over duration on the
order of 50 years.
DEFINITIONS
The following definitions are applicable and important in understanding the pile analysis
results:
Base Resistance — the resistance or support from soil or bedrock at the pile base as a result
of pile movement. If there is no pile end movement, there is no base resistance (i.e., friction
pile). If moderate pile movement, pile is supported in both side resistance and base resistance. If
excessive pile movement, base resistance may be very high.
Downdrag — the movement of the pile as a result of soil movement when the neutral plane
is located above or within a zone of compressible soil.
Negative Side Resistance — soil load (drag load) applied from the top portion of the pile in a
downward direction with respect to the pile.
Neutral Plane — the location or zone along the pile at transition or interface between
positive and negative side resistance. It is also the location or zone along the pile where the pile
and adjacent soil move together. In addition, it is the location of maximum stress within the pile.
Positive Side Resistance — soil support from the bottom portion of the pile in an upward
direction with respect to the pile.
Side Resistance — resistance or support of the pile from soil adjacent to the pile perimeter
that is applied to the pile in either direction, up (positive support) or down (drag load).
PILE ANALYSIS RESULTS

Neutral Plane methodology as described in Section 7.2.1.3.4 of FHWA-NHI-16-009 (2016)
was used to show the effect of downdrag on a single pile. Four models were developed, each
© ASCE
Geo-Congress 2020
showing a condition in time after final bridge construction. The first model represents the
condition immediately after final bridge completion.
The second model represents a point in time when full side resistance was engaged. This
model includes the assumption of full side resistance with no pile base resistance; the assumption
is supported by Section 4.1.2 of Tomlinson and Woodward (2008) that shows the effects of
loading a pile. Figure 10.8.2.2.2-1 in AASHTO (2017) suggests that the majority of side
resistance in cohesive soil is mobilized after only 0.6% of the base diameter of the pile, in this
case, about 0.05 inch settlement; Figure 10.8.2.2.2-4 indicates very little (~15% or less) base
resistance would have mobilized at that amount of settlement.

The third model represents the mobilization of full-side resistance and full-base resistance.
The final model represents the condition at complete pile downdrag movement.
Since side resistance is mobilized with less movement than pile base resistance, it is first
mobilized before pile base resistance. Beginning immediately after pile installation, the pile load
consists of the pile’s own weight, which is supported entirely in side resistance. Side resistance is
believed to be the primary means of support at the initial load condition as supported by
Tomlinson and Woodward (2008).
As the dead load of the new bridge is applied during construction, the dead load is supported
primarily in side resistance as shown in Figure 5. Drag load is 19 kips resulting in a maximum
pile load of about 54 kips, which is well below the 187 kips available (based on pile diameter) at
the neutral plane. At this condition, the neutral plane is located about 12 feet below the top of
pile. Because the neutral is above the compressible layer, additional pile movement occurred
during consolidation of the compressible clay layer.
Figure 5. First model – Immediately after bridge construction

With additional consolidation, all available side resistance is mobilized (101 kips) as shown
in Figure 6. The neutral plane location moved downward to 40 feet below the pile top. This
resulted in less structural pile resistance (66 kips) due to the smaller diameter of the tapered
timber pile. At this stage, the maximum pile load is 69 kips, greater than the available
compressive resistance of the pile.
© ASCE
Geo-Congress 2020
Figure 6. Second model – after additional consolidation of compressible layer

The next model includes full available side resistance. Due to continued consolidation, full
pile base resistance is mobilized as well as shown in Figure 7. Now the side resistance is 101
kips and base resistance is 54 kips. The drag load is almost double the dead load at 60 kips. The
increased drag load caused the internal pile load (95 kips) to exceed the available structural
resistance in compression. The pile base resistance is at 54 kips, resulting in base stress very near
the compressive strength of timber (1.25 ksi). The neutral plane was moved downward with
additional consolidation of the compressible clay layer, now at 66 feet below the top of pile. At
66 feet, the neutral plane is about the midpoint of the compressible clay layer.
Figure 7. Third model – after full-side resistance and full-base resistance

The final model was performed with neutral plane located at the bottom of the compressible
layer at completion of pile downdrag movement as shown in Figure 8. At this location,
maximum pile load is 116 kips and pile base load is 95 kips — both are well above the structural
capacity of the timber pile. Pile stress is 2.6 ksi and 2.2 ksi at the neutral plane and pile base,
© ASCE
Geo-Congress 2020
respectively, greatly exceeding 1.25 ksi, the strength of timber pile in compression.
Figure 8. Final model – at completion of pile downdrag

As shown in Figure 7, the timber strength is exceeded around the midpoint of the
compressible clay layer. Therefore, the portion of pile below the midpoint of the clay layer could
no longer be considered to provide axial geotechnical resistance since it is broken structurally.
Consequently, it is anticipated the abutment will move as a result of reduced axial resistance of
the piles.
Given the movement of the system is driven by structural failure of the timber piles, steel
piles which typically have higher yield strengths and a consistent area along the length, were
analyzed. For the initial comparison, a similar sized 7-inch diameter closed-end steel pipe pile
was modeled with the downdrag condition where the neutral plane is located at the bottom of the
compressible layer. The maximum pile load was 98 kips and pile base load was 88 kips. Pipe
pile stress was about 18 ksi and 2 ksi at the neutral plane and pile base, respectively. Since the
yield strength of Grade 2 steel is 35 ksi, the pipe pile strength in compression would have
supported the dead load and additional dragload.
Another comparison of similar sized steel H-pile was modeled at the downdrag condition. An
HP 8x36 was modeled with the neutral plane located at the bottom of the compressible layer.
The maximum pile load was 82 kips and pile base load was 75 kips. H-pile compressive stress
was about 8 ksi and 7 ksi at the neutral plane and pile base, respectively, well below the
compressive strength of steel.
Table 1 shows the comparison of three pile types each modeled with the neutral plane at the
bottom of compressible soil layer. The results indicate the timber pile has been overstressed and
pile top has moved downward from excessive drag load on the pile during consolidation of the
compressible soil layer, resulting in the phenomenon known as pile downdrag.
Due to the higher strength in compression of steel as compared to timber, neither the pipe
pile nor the H-pile would have been overstressed. Compressive stress at the pipe pile base would
have been similar to the timber pile, at approximately 2 ksi. The compressive stress at the base of
the H-pile would have been slightly higher at 7 ksi. The very dense gravels at the pile base have
a Young’s Modulus estimated at about 20 ksi, so movement at the base of either steel pipe or H-
© ASCE
Geo-Congress 2020
interface.
Pile downdrag movement is a service limit state phenomenon dependent primarily upon
settlement characteristics of soils at the site, but secondarily based on pile section properties and
subsurface conditions at the pile base. Pile drag load is the cause of potential pile downdrag
movement, but more importantly is additional load in the pile and therefore is a section property
concern that should be evaluated at the pile strength limit state.
REFERENCES
AASHTO (American Association of State Highway and Transportation Officials) LRFD (2017).
“AASHTO LRFD Bridge Design Specifications”, 8th Edition, September, 2017.
FHWA-NHI-16-009 (2016). “Geotechnical Engineering Circular No. 12 – Design and
Construction of Driven Pile Foundations”, September, 2016.
MnDOT LRFD (2018). “Load and Resistance Factor Design, Bridge Design Manual”, October,
2018.
MnDOT Memo (2014). “SP 2515-21 Bridge 25033, TH 63 over Mississippi River in Red Wing,
Subsurface Investigation”, December 17, 2014.
Tomlinson and Woodward (2008). “Pile Design and Construction Practice”, 5 th Edition, 2008.
© ASCE
Geo-Congress 2020
Washington Park Reservoir Improvements: Accommodating Ancient Landslide Movement

with a Compressible Inclusion
Thomas Westover, P.E., M.ASCE1; Dan Hogan, P.E., G.E., M.ASCE2;
Gerry Heslin, P.E., M.ASCE3; and Andrew Kost, P.E., M.ASCE4
1
Associate Engineer, Cornforth Consultants, Inc., Portland, OR. E-mail:
twestover@cornforthconsultants.com
2
Engineer, Portland Water Bureau, Portland, OR. E-mail: Dan.Hogan@portlandoregon.gov
3
Vice President, Cornforth Consultants, Inc., Portland, OR. E-mail:
gheslin@cornforthconsultants.com
4
Project Engineer, Cornforth Consultants, Inc., Portland, OR. E-mail:
akost@cornforthconsultants.com
ABSTRACT
The Portland Water Bureau (PWB), until recently, had operated two open drinking water
reservoirs at Washington Park, located in Portland, Oregon. The open reservoirs were part of a
gravity-fed water system constructed during the 1890s. During original construction at the site, a
large ancient landslide was re-activated. A project to improve the reliability and protection for
the water supply is currently in progress, and includes a buried 12.4-million gallon (47 M liter)
concrete reservoir. When completed, the project will provide a seismically resilient supply of
drinking water to 360,000 people, including all of downtown Portland. The buried reservoir will
be isolated from continued landslide movements using an innovative compressible inclusion. The
compressible inclusion will deform in response to continued landslide movement, and in turn
will impart relatively small, predictable loads on the buried reservoir walls. During a seismic
event, the compressible inclusion will absorb landslide displacement as well as dampen reservoir
displacements, reducing the demands on drilled shaft foundations and reservoir walls.
INTRODUCTION
The Washington Park Reservoirs are part of the City of Portland, Oregon’s water system,
owned and operated by the Portland Water Bureau (PWB). The reservoirs form the western
terminus of a gravity-fed system that originates in the Bull Run Watershed northwest of Mt.
Hood. The reservoirs are located in Washington Park, one of the oldest parks in Portland. The
original Washington Park Facility consisted of two, open-water reservoirs (Reservoirs 3 and 4)
that stored drinking water for the west side of the Willamette River including downtown Portland
since the late 1890’s, see Figure 1.
Since construction, the west sides of both reservoirs have been damaged from landslide
movement. The construction of the reservoirs at the base of a narrow, steep-sided ravine
reactivated a large “ancient” landslide. The landslide is approximately 1,700 feet (520 meters)
long and 1,100 feet (335 meters) wide with an average thickness of about 80 feet (25 meters), see
Figure 2. A system of drainage tunnels constructed in the early 1900s succeeded in slowing, but
not stopping, the landslide movement. Regular monitoring over the past two decades has shown
that the landslide continues to move at approximately 1/8 inch (3mm) per year, primarily during
the wet season.
© ASCE
Geo-Congress 2020
Figure 1: Site Plan of the Washington Park Landslide
Figure 2: Cross-Section of the Washington Park Landslide

The current project to replace the two open reservoirs with a single buried reservoir is driven
in part by aging infrastructure, seismic susceptibility of Reservoirs 3 and 4, and the need to
comply with new drinking water regulations issued by the Environmental Protection Agency.
The Washington Park Reservoir Improvement project will replace existing Reservoir 3 with an
irregularly shaped, 12.4-million gallon (46M liter) reinforced concrete reservoir. The marginal
stability of the landslide poses a design constraint for the new reservoir. Several options for
mitigating for the landslide were evaluated, including ground anchors, shear piles, an anchored
soldier pile wall, and a compressible inclusion. The compressible inclusion concept was selected
by PWB as the preferred option. The buried reservoir will be located outside the current toe of
the landslide and isolated from the slide by placing a compressible inclusion between the
© ASCE
Geo-Congress 2020
landslide toe and the new reservoir, see Figure 3. Future landslide movements will be
accommodated by the compressible inclusion. As it deforms, the compressible inclusion will
impart a relatively small, predictable load on the reservoir. Reservoir 4 will be backfilled and
converted to lowland habitat to handle stormwater, reservoir overflow, and subsurface drainage.
Fill will also be placed on the west slope above Reservoir 4 to restore the ground surface as near
as practicable to the grade that existed before landslide movements were initiated by original
reservoir construction in the 1890s.
Figure 3: Detailed section near MSE wall, compressible inclusion, and new reservoir.
BACKGROUND
At the time of original construction, the City’s Engineer was heavily involved in the design
and construction of the new reservoirs. The engineer, D.D. Clarke, was involved with the project
from its initiation in the 1890’s up through the maintenance activities that took place in 1917.
Clarke published several papers on the landslide, which at the time was one of the largest ever
studied. His first paper published in 1904 (Clarke, 1904) provides a comprehensive historical
background of the original landslide investigations and eventual partial stabilization by means of
drainage tunnels. This paper also discusses in detail the ground conditions encountered in hand-
dug shafts and the quantities of groundwater encountered during excavation. A 1917 (Clarke,
1917) paper discusses the effectiveness of tunnel dewatering during the period from 1906 to
1917, and summarizes the reservoir repair work following the 1904-1906 tunnel work.
Construction History
Reservoir Nos. 3 and 4 were constructed between October 1893 and September 1894 in the
north branch of Tanner Creek, a narrow, steep-sided ravine. Excavation for the reservoirs cut
into the west side of the ravine up to 40 feet (12 meters) vertically and 70 feet (21 meters)
horizontally. In August 1894, shortly before the excavation was completed, ground cracks were
detected in the west side of Reservoir No. 4. Similar ground cracks were detected the following
© ASCE
Geo-Congress 2020
month in Reservoir No. 3. While the evidence suggests that landslide movements were
accelerated by excavations, Clarke’s 1904 paper mentions that there was some evidence that the
landslide mass was unstable prior to construction of the reservoirs. A timber trestle for a cable
railway that crossed the landslide was difficult to keep “plumb” and militia target range stakes
were consistently becoming “out-of-line.” While the extent and rate of movement of the
landslide prior to reservoir construction is difficult to gauge, there is no question about the
movements following the excavation for the reservoirs.
At first, the cracks observed above the reservoir cuts were thought to be associated with local
sloughing of the cut slope. To stabilize the presumed local problem, small buttresses were
constructed. However, damage to the west side of the reservoir slopes and the parapet walls
continued following these remedial measures. Between January 1895 and October 1897, the
parapet wall on the west side of Reservoir No. 4 deflected eastward 3.24 feet (1 meter), an
average of 15 inches (38 cm) per year. Over this same period, Reservoir No. 3 moved 1.69 feet
(0.5 meter), an average rate of 7 inches (18 cm) per year.
Site reconnaissance of the ground west of the reservoirs in September 1895 revealed that the
reservoir cracks were associated with a large landslide, extending much further upslope than
previously thought. Ground cracks were detected which outlined the limits of the landslide. The
headscarp of the slide was located 1,700 feet (520 meters) uphill from the reservoirs. The
landslide likely formed hundreds or thousands of years ago due to water eroding and
undercutting at the base of the slope. Evidence suggests that the ancient landslide has probably
remained in a marginally stable condition ever since, i.e. moving slightly during the wet winter
seasons, and stabilizing during the drier summer months. Interviews with residents and people
familiar with the area indicated that ground movements were noted as early as 1850.
Shafts, Tunnels and Dewatering

To evaluate the depth of the landslide and determine some means for stabilizing the
movements, an extensive field exploration program was undertaken. From September 1895 to
January 1899, a total of 33 drillholes and 22 hand-dug shafts were placed within the boundaries
of the slide. The results of the exploratory program indicated that the landslide was shearing on a
thin clay seam near the top of basalt bedrock. Overlying the clay seam was a mixture of clayey
silt and silty clay with numerous basalt gravels. The depth of the landslide varied from 50 to 120
feet (15 to 37 meters) with an average depth of 80 feet (25 meters).
During excavation of some of the shafts, large flows of groundwater (up to 115 gallons per
minute (gpm) 435 liters per minute (lpm)) were encountered. To continue excavation, it was
necessary to pump large quantities of water. For example, during the excavation of Shaft No. 1,
approximately 4 million gallons (15.1M liters) of water was pumped over a 52-month period. It
was noted that the movements of the slide decreased dramatically during this pumping period.
Therefore, stabilization of the slide by means of drainage tunnels was chosen as the best method
for protecting the reservoirs. From July 1900 to September 1901, 2,500 linear feet (760 meters)
of tunnel was constructed at the base of the landslide, interconnecting the vertical shafts by
following along the landslide shear zone. Between 1901 and 1904, daily groundwater flow rates
were monitored in the tunnels. Summer flow rates were found to vary from 7 to 10 gpm (27 to 38
lpm), while winter rates varied from 17 to 35 gpm (64 to 132 lpm). During extremely severe
storms, flow rates up to 50 gpm (190 lpm) were observed.
© ASCE
Geo-Congress 2020
Tunnel In-filling
For the first couple of years following the tunnel work, movements of the slide mass
averaged about ¼ inch (6 mm) per year; however, movements increased to about 1⅓ inch (34
mm) in early 1904 due to heavy rainfall, and a decision was taken to immediately install
additional drainage tunnels. Four thousand feet of additional drainage tunnels were installed by
August 1906. A comprehensive survey grid consisting of 17 lines with 260 points was
established to monitor the slide movement. The grid survey points were monitored closely from
1906 to 1916, with overall slide movements averaging ½ inch (13 mm) per year. In 1925 and
1926, tunnels damaged by the landslide were repaired or replaced. In 1928 and 1929, most of the
tunnels were backfilled with gravel to prevent collapse of the tunnels yet still allow water to
drain from the slide.
WASHINGTON PARK LANDSLIDE

Geology
Washington Park is located in the Portland Basin within the Willamette Valley physiographic
province. The Portland Basin is a forearc basin located between the Coast Range to the west and
the Cascade Range to the east. The bedrock geology of the Portland Basin is dominated by
Miocene flood basalt (Columbia River Basalts), which flowed episodically into the basin via an
ancestral Columbia River channel, from approximately 17 to 6 million years ago (Madin, 2009).
The Washington Park reservoirs are located on the east flank of the Tualatin Mountains
(locally referred to as the Portland West Hills). During the late Pleistocene (approximately 700 to
34 thousand years ago), wind-deposited glacial silt (loess) from the Columbia River Gorge
blanketed the Portland West Hills. This material, known as the Portland Hills Silt (PHS) is
locally up to 120 feet (37 meters) thick (Lentz, 1981). The presence of buried soil horizons
(paleosols) within the PHS indicates that the loess was deposited during multiple events. The
flanks of the Portland West Hills, and the entire Portland Basin, were heavily altered by repeated
catastrophic glacial outburst floods. These Pleistocene-aged floods are referred to as the
Missoula Floods and occurred between 23 and 15 thousand years ago (Madin, 2009).
The landslide is located on the northeast flank of the southeast-trending Portland Hills
anticline, which is composed of Columbia River Basalt flows dipping to the east at 7 to 15
degrees. In the slide area, the Columbia River Basalt is covered by surficial deposits that include
landslide material, colluvium, alluvium, silt, and fill. Basalt adjacent to the slide area is mostly
weathered and easily breaks down into soil-like composition. The landslide shear zone dips 7.5
to 14 degrees east, which coincides with the regional dip of the basalt flows.
The ancient landslide at Washington Park is a large, translational block landslide that moves
on a deep-seated, discrete shear zone roughly parallel to the ground surface. The slide is
approximately 1,700 feet (520 meters) long and varies in width from 400 feet (122 meters) at the
head to 1,100 feet (335 meters) at the toe. The depth of the landslide varies from 50 to 120 feet
(15 to 37 meters), with an average depth of 80 feet (25 meters). The headscarp is located on the
west side of Portland’s Japanese Garden at approximately elevation 700 feet (213 meters). The
historical toe of the landslide toe is located approximately at elevation 260 feet (79 meters) in
Reservoir 3 and elevation 215 feet (66 meters) in Reservoir 4.
The landslide topography has been modified by cuts and fills, some pre-dating construction
of the reservoirs. The overall slide morphology is characteristic of ancient landslide terrain,
exhibiting hummocky topography with a pronounced graben, curved fir trees, and signs of local
© ASCE
Geo-Congress 2020
slumps and ground distress. The graben area of this landslide is currently occupied by the
Japanese Garden. Signs of both the northern and southern lateral scarps can be observed near the
headscarp. The pavement shows signs of distress and displacement. Cracks and vertical offset in
a tennis court adjacent to Kingston Avenue clearly identify a portion of the northern edge of the
slide. Near the toe of the slide, cracking and signs of distress were observed in a retaining wall
located on the northwest edge of Reservoir 3 and in Dam 3.
Movement History
Much of the early movement history of the landslide is cataloged in Clarke’s 1904 paper.
PWB has historical survey records which fill in the intervening years until modern slope
inclinometers were installed in 1986. Movement rates have been extremely consistent in that
time, with very little variation. Movement generally is slightly higher in the winter months when
precipitation and groundwater levels are higher (~10 feet (3 meters) higher than their seasonal
lows). A summary of the movement history is provided in Table 1.
Table 1: Movement history of the Washington Park Landslide, 1893-Present.

Annual Rate of
Date Movement Description of Events
inch/year mm/year
1893-1894 Unknown Reservoirs constructed
1895-1896 15 380 Water Bureau assessing cause of movements
1897-1898 Pump dewatering of exploratory shafts reduces movement
1½ 38
rate
1899-1900 Exploratory shafts completed; movement rates increase
4 102
due to stoppage of dewatering pumps; survey grid installed
1901-1904 ¼ 6.4 Drainage tunnels constructed
1904-1906 Movements increase; additional drainage tunnels are
1⅓ 34
installed
1906-1916 ½ 13 Detailed survey monitoring
1920-1970 ½ 13 Continued survey monitoring
1975-1986 ¼ 6.4 Measurements obtained from 2 EDR casings
1987-2016 0.14 3.6 Measurements obtained from 7 inclinometer casings
2016-present Varies Construction of new reservoir facilities
The decrease in the rate of movement of the Washington Park slide over the last century can
be explained by the actively changing geometry of the slide mass and fluctuations in
groundwater level. As the slide moves downhill, the upper portion of the landslide pulls away
from the stable ground above it and drops in elevation, creating a graben. This downward
movement decreases the driving force of the landslide. Given enough time, the landslide
movements will bring the driving forces into equilibrium with the resisting forces and the slide
will stop; however, the time required to reach equilibrium could take several more centuries.
Stability of the landslide mass is also influenced by changes in groundwater level. Accurate
records of groundwater encountered in the initial borehole and hand-dug shaft programs provides
a considerable basis for modeling and back-calculation. Similar records were taken after the
dewatering tunnels, providing an excellent accounting of response of the landslide mass to the
removal of large quantities of groundwater.
© ASCE
Geo-Congress 2020
Slope Stability Studies

A detailed slope stability evaluation for the Washington Park landslide was performed prior
to the development of final design. The account from Clarke’s original investigation, coupled
with modern monitoring methods implemented since 1986, provide an extraordinary level of
understanding into the internal composition of the landslide, groundwater levels, and landslide
shear zone geometry. The stability study of this landslide benefited greatly from well-
documented stability conditions at various points in time. These include: i) pre-construction
ground surface in the 1890’s; ii) reservoir excavation with full groundwater levels; iii) reservoir
excavations with partially lowered groundwater levels after the first phase of tunnel construction;
iv) reservoir excavations with fully lowered groundwater levels after the second phase of tunnel
construction; and v) current conditions with detailed groundwater and landslide movement
records. The variety of calibration conditions provided a robust and thorough set of strength
estimates, which increase the reliability of the stability model.
The analyses of the landslide were performed at two cross sections, one through Reservoir 3
and one through Reservoir 4. Four options were evaluated to mitigate the impact of the landslide
on the proposed new buried reservoir at the location of the existing open Reservoir 3. The
options included: i) designing new structures to accommodate landslide movement; ii) stabilizing
the landslide with ground anchors; iii) stabilizing the landslide with a soldier pile and tieback
wall; and iv) stabilizing the landslide with shear piles. The analyses showed that it would be very
costly to stabilize the landslide using structural elements, and that any structural stabilization
measures would likely be damaged during earthquake-induced landslide displacement.
After reviewing the four mitigation options, PWB selected the option that would allow for
accommodation of the landslide movement. This option involves placing a low-strength
compressible inclusion between the landslide and the new buried reservoir at Reservoir 3 to
accommodate slide movement. At Reservoir 4, the option includes backfilling the existing open
reservoir and placing fill on the slope west of the reservoir to buttress the landslide. The analyses
indicate that the calculated factor of safety for the landslide increases slightly due to the
proposed improvements. At Reservoir 3 the improvement is approximately 3 percent, due in
large part to constructing a large mechanically stabilized earth wall at the toe of the landslide. At
Reservoir 4, the improvement is approximately 7 percent. Based on these modest improvements,
it was decided that proposed structures and improvements would be designed to accommodate
continued movement of the landslide.
Stability modeling of the landslide was performed using SLOPE/W (GeoSlope 2007). The
shear zone location is based on slope inclinometer readings, exploratory borings, and
descriptions of materials encountered in dewatering shafts and tunnels. The shear zone roughly
follows top of bedrock for the majority of the landslide length. Near the landslide toe, the shear
zone diverges from bedrock and daylights in the western slope of Reservoirs 3 and 4. The
location of the landslide toe is based on descriptions in Clarke (1904) and by observations of
distress in the reservoir’s concrete liner.
Back Analysis: Stability analyses were performed to back analyze the shear strength along
the landslide shear zone. The landslide mass can be subdivided into three different blocks with
different base inclinations. There is an upper driving block with a relatively flat base, an
intermediate block with a steeper base angle, and a lower block with a relatively flat base.
Surface deformations and inclinometer data indicate a relatively uniform rate of movement along
the length of the landslide, so the landslide was calibrated using intermediate stability analyses
that considered the residual shear zone strength and unique geometry of each individual block.
© ASCE
Geo-Congress 2020
The lower landslide block was analyzed assuming that the shear surface extended to the ground
surface from the western limit of the lower block at an angle of 60 degrees to horizontal. The
friction angle along the base of the block was varied to get a factor-of-safety (FS) near 1.0, and
remained fixed for subsequent analyses. The basal friction angle for the middle block was
calculated considering the stability of the middle and lower blocks as a unit. The friction angle of
the middle block was varied to achieve a FS near 1.0. The final step was to evaluate the stability
of all three blocks as a unit. The friction angle of the upper block was varied to get the calculated
FS of the whole mass near 1.0. The back-calculated friction angles varied from 7.5 to 16 degrees,
which are similar to residual strength values obtained from ring shear testing of local weathered
basalts.
Mitigated Factor-of-Safety: Key factors that affect the target FS for landslide improvement
include the rate of landslide movement, the level of understanding of the landslide, the size of the
landslide, and the consequences of failure. For very large landslides with a well-defined shear
zone and a long history of movement data, like the Washington Park landslide, it is possible to
work toward a target FS of just above 1.0. By comparison, for slope stability problems (e.g.
roadway cuts or fill slope developments) where the geometry and strength information is poorly
defined, target FS are typically near 1.5.
In the case of the Washington Park landslide, we can calculate the reduction in stabilizing
force that caused the 1893 reactivation. Slope stability analyses indicate that the original
excavation at the toe of the landslide decreased stability of the landslide by 5 to 7 percent. This
excavation caused the landslide to move at a rate of approximately 15 inches (380 mm) per year.
The dewatering tunnels slowed the rate of movement to approximately ½ inch (13 mm) per year,
which suggests that the marginal stability of the pre-1893 landslide was restored. From this, one
can conclude that the dewatering tunnels improved landslide stability approximately 5 to 7
percent.
For the cross-sections analyzed, applying an additional horizontal force of between
approximately 150 and 200 kips per lineal foot (2,200 to 2,900 kN/meter) to the toe of the
landslide raises the FS approximately 5%. These large loads are difficult and costly to achieve
with standard structural mitigations and achieve only a marginal increase in landslide stability.
The analyses indicate that it is important for the drainage tunnels to continue functioning
properly, and monitoring of the groundwater levels will continue for the life of the project.
Seismic Landslide Response

There are no widely-accepted published procedures for estimating the displacements of large
landslides with low static FS during seismic events. The Newmark (1965) analysis technique is a
theoretical solution that treats the landslide mass as a rigid body and assumes movement occurs
whenever the FS drops below 1.0. The calculated FS varies as the acceleration varies during the
earthquake time history. Case histories of translational slides subjected to earthquake motions
indicate that Newmark-type analyses predict larger movements than actually occur. Potential
reasons for the over-prediction include the fact that the rigid body assumption may not be valid
for large landslides and that clay soil generally has a higher shear strength during rapid loading
than during slow, monotonic loading.
Kulhawy and Mayne (1990) studied the influence of loading rate on the shear strength of
clay soils, and concluded that shear strength increases by approximately 10% for each order of
magnitude increase in loading rate. This study indicates that the shear strength of clay soils could
be 70 to 85% higher during seismic shaking than during static loading, based on the velocities
© ASCE
Geo-Congress 2020
obtained during shaking. Other studies show considerable variation in in the magnitude of
dynamic strength gain, and a value of 40% dynamic strength gain was selected based on the
body of literature, expected velocities during seismic shaking, a parametric study, and agreement
between modified Newmark procedures and FLAC analyses discussed below.
Newmark analyses completed using increased shear strengths predict displacements that are
much closer to those measured in documented case histories (Vessely, 1998). To perform the
modified Newmark analyses, the landslide is back-calculated to its current state (in this case FS
= 1.0). Strengths along the shear zone are then increased by 40%. With increased strengths, the
yield acceleration for the landslide is calculated and used in a standard Newmark deformation
analysis.
The landslide displacement estimates for sections with structural mitigations were very
similar to the unmitigated case, since the relatively small stability improvement does not
substantially raise the yield acceleration. As such, the displacement estimates for the mitigated
and unmitigated cases are similar, indicating that structural landslide mitigation would not
significantly decrease potential seismic displacements. Displacement estimates for the 475-year
ground motion are generally less than ½-inch (13 mm) for all mitigation measures. Displacement
estimates for the 2,475-year ground motion are generally less than 3½ inches (89 mm).
Based on inclinometer measurements and descriptions in Clarke (1904), the thickness of the
shear zone at Washington Park varies from less than 6 inches to 2 feet (0.15 to 0.6 meters). Any
structural elements (i.e. ground anchors or shear piles) that span the shear zone would be
subjected to shear displacements over the thickness of the shear zone. Displacements on the
order of 3 inches (0.07 meters) along the shear zone could damage a conventional ground
anchor. If ground anchors are damaged by landslide movements, the landslide could resume
moving at a slow creep rate.
The software program FLAC was used to create a two-dimensional, finite-difference model
of the landslide and proposed structures. Model geometry and material properties were
developed from exploration and laboratory data. The ground motion time histories and scaling
factors were developed for the 475-year and 2,475-year events, including time histories from the
2010 Chile and 2011 Tohoku interface subduction earthquakes. These events included long-
duration periods of strong shaking, which can have a significant effect on the accumulated
displacements.
A series of parametric evaluations were performed with the FLAC model to evaluate the
influence of specific model variables on the calculated displacements. The drilled shafts
foundation of the tank have a small influence on horizontal movements. Removing the shafts
from the model increased horizontal displacements by less than 7% as compared to the base case.
In addition, the amount of dynamic shear strength increase on the landslide shear zone has a
large influence on calculated movements. Horizontal displacements calculated assuming a 20%
dynamic shear strength increase resulted in horizontal movements 20-60% larger than horizontal
movements calculated using a 40% dynamic shear strength increase.
WASHINGTON PARK RESERVOIR IMPROVEMENTS

The reservoir at Washington Park serves as the system’s largest drinking water storage
facility west of the Willamette River, which runs through downtown Portland. Since the system
is gravity fed, the elevation of the facility is critical for its continued function. Relocating the
reservoir out of the immediate landslide area was considered. However, given the high cost of
reconstructing the supply infrastructure and difficulty in locating a site at the required elevation
© ASCE
Geo-Congress 2020
without substantial geologic challenges, the new reservoir is constructed at the existing site.
The former open-air reservoir consisted of a concrete liner sitting on sloped ground.
Landslide movement caused substantial cracking to the liner, so a flexible membrane was
installed over the concrete. The new reinforced concrete reservoir will have vertical walls, and
thus will require a substantial excavation to obtain a similar volume. The footprint of the new
reservoir has been moved to the east, out of the existing landslide toe. This shift provides a
buffer of 10 feet from the historic landslide toe to the edge of the concrete structure. A 50-foot
(15 meter) tall MSE wall is constructed on the west slope, to provide a self-stable structure
adjacent to the compressible inclusion. The MSE wall will also provide a small buttressing load
to the slide toe.
A tiered system of two soldier pile and tieback walls were constructed outside the landslide,
along the north and east slopes of the existing open reservoir. An upper temporary wall supports
the adjacent Sacajawea Boulevard and allows the contractor to excavate a work area, or drill
bench, from which to install soldier piles for the lower permanent wall. The permanent wall
permits excavation for the buried reservoir and prevents lateral earth pressures from acting on the
side of the reservoir during its design life. Outlet and drainage piping has been reconstructed
through the site, utilizing a jack and bore technique for new pipe runs that pass under the dam
which previously retained the reservoir.
The buried reservoir will be supported on drilled shafts socketed into bedrock. Some of these
shafts will be subjected to high drag loads from settlements induced by fills on the south and
west sides of the new reservoir. The fills will support visual features planned on top of the buried
reservoir. On the west side of the reservoir, a compressible inclusion consisting of expanded
polystyrene (EPS) will be placed between the MSE wall and the reservoir to accommodate
landslide movement. Along the east and south walls of the reservoir, EPS will be placed between
the reservoir and adjacent retaining structures or fill to act as a buffer during a seismic event.
Two large reinforced soil slopes will be constructed on the slope west of Reservoir 4 to
buttress the landslide. Together, these two slopes approximate the pre-reservoir ground surface.
The tiered wall system will support the Murray Street switchback, providing access from the
bottom of the hill to the new reservoir.
ACCOMODATING THE LANDSLIDE

Compressible inclusions are an emerging technology that can be used to reduce earth
pressure against structures as well as provide a uniform and controllable response to
deformation. Expanded Polystyrene (EPS) blocks are a common material for this application, as
the properties of the material can be manipulated by varying the density of the block.
Compressible inclusions are designed as yielding elements placed between an earth material and
a rigid wall to prevent the buildup of passive earth pressures. In the majority of compressible
inclusion designs, the thickness of the compressible inclusion and the expected deformations are
balanced such that the material remains within the elastic range (< 1% strain, see below). The
application of these concepts to the slide movements at the Washington Park Reservoirs requires
an extension of previously defined design methodologies to account for the displacements of the
landslide expected over the design life of the structure. These large permanent deformations of
the landslide contribute to a significant increase in the complexity of the design.
Compressible Inclusion Concept

The purpose of the compressible inclusion (CI) is to reduce the magnitude of earth pressure
© ASCE
Geo-Congress 2020
applied to the sides of the rigid concrete tank. Under non-landslide conditions, with fill adjacent
to the tank walls, the tank would feel at-rest earth pressures. With landslide displacement,
passive pressures would rapidly develop, damaging the tank. By including the CI as a buffer
between the moving ground and the rigid tank, the lateral force imparted by the landslide
movement on the tank walls will develop according to the stress-strain properties of the EPS
rather than earth materials, resulting in significantly lower loads.
The stiffness and strength properties of the EPS vary as a function of density and
manufacturing processes. In the interest of reducing this lateral loading as much as possible, EPS
of the lightest weight will have generally the softest response to the lateral force of the landslide.
For the purpose of this application, conventional block-molded EPS exhibits four primary stress-
strain regimes: i) a fairly stiff elastic response (0-1% strain); ii) a gradually softening yielding
response (1-5% strain); iii) slow strain hardening (5-50% strain); and iv) rapid strain hardening
(>50% strain). This application utilizes the material behavior in the first three regimes, as the
rapid strain hardening regime would quickly overload the structure.
Figure 4: Stress-Strain Properties of Expanded Polystyrene (EPS), Grade 15.

The CI must be able to support the lateral at-rest earth pressure of the backfill material. EPS
that is too soft to counteract the lateral forces of the earth pressure as the fill is placed will strain
excessively and reduce the available design life and seismic capacity of the CI. The adjacent fill
is constructed as a vertically faced MSE wall or an unreinforced concrete block, resulting in
© ASCE
Geo-Congress 2020
zero-earth pressure initially acting on the EPS.
Mechanics of Expanded Polystyrene (EPS)

EPS is a visco-elasto-plastic material with significant strain-hardening characteristics at very
high strains. The material behavior is complex, but the behavior can be broken down into four
primary stress-strain regimes. These regions are defined by strain, and will be discussed
individually to provide insight into the response (see Figure 4). In addition to these zones of
behavior, EPS also exhibits a rate-dependent loading effect (stiffer response under faster
loading).
Elastic Behavior – Strains of <1%: This strain range is the primary regime for nearly all
EPS applications, and represents the majority of the research focus in the literature. The design
of lightweight fills and typical CIs remain almost entirely within this range. Creep behavior
below 0.5% strain is small to non-existent. Between 0.5% and 1% strain, creep models exist to
evaluate the time-dependent (visco) response of the material, but the response is generally minor.
The stress at 1% strain is typically referred to as the elastic limit stress. The initial stiffness of the
material is high compared to the behavior at higher strains.
Yielding Behavior – Strains of 1%-5%: In this region, permanent deformation of the EPS
with increased loading is expected as the material transitions from elastic behavior into a plastic
yielding regime. During this transition period, the stiffness of the material decreases
dramatically. Within this strain range, the material begins to develop significant creep behavior.
Slow Strain Hardening – Strains of 5%-50%: Following the initial yielding, the material
exhibits strain hardening behavior. Although the material does harden, the stress increases only
slightly over a very large strain range. Permanent deformation will accumulate at these strains,
and only about 1% of the strain would be recovered elastically if unloaded. Creep behavior in
this regime has not been rigorously investigated, nor is it well understood. It is generally
acknowledged that the magnitude of creep increases with additional stress. These materials can
also exhibit a behavior referred to as tertiary creep. Tertiary creep can result in a sudden and
catastrophic collapse, whereby the material creeps experiences rapid strain accumulation under
constant stress.
Rapid Strain Hardening – Strains of 50%-100%: The last regime is characterized by an
increasingly rapid hardening of the material as it is crushed back into its constitutive
components. Stresses within the material increase very rapidly with increasing strain. Creep
behavior at these large strains has not been investigated.
Load Transfer Mechanism

The primary goal of the design methodology is to accurately characterize how the landslide
displacement will translate into loading on the structure. Without a CI, the reservoir would
experience passive earth pressures as landslide movement occurs and the reservoir would
become cost-prohibitive to construct. Conventional CI design, where strains remain below 1%,
would result in a CI thickness of over 100 feet (30 meters). By taking advantage of the favorable
behavior of the material (i.e. relatively small increases in stress over a large strain range) at
strains of less than 50%, the thickness can be reduced to less than 10 feet (3 meters).
The order of construction has an effect on the load transfer. First, the adjacent MSE wall is
constructed to its full height. Then, the reinforced concrete reservoir is constructed and leak
tested. Finally, the EPS is installed to make intimate contact with the face of MSE wall on one
side and the concrete reservoir wall on the other.
© ASCE
Geo-Congress 2020
Lateral loads from the CI will result solely from the displacement of the landslide and will
vary over the design life of the structure. Immediately after construction, lateral load on the
reservoir walls will be approximately zero. When the slide moves, the MSE wall will push on the
CI, transferring load to the reservoir. The reservoir walls and foundation were designed for the
anticipated applied load, the magnitude of which will be determined by the displacement of the
fill and the stress-strain characteristics of the EPS. As slide movement ceases at the beginning of
the dry season, the CI will creep in response to sustained loading from MSE, and the load on the
reservoir will decrease. During the remainder of the dry season when the slide typically does not
move, the load on the reservoir will remain relatively constant at a load less than that
experienced during the previous wet season. When the slide begins to move again during the
following wet season, MSE will compress the inclusion further, and the load on the reservoir will
rise to a level determined by the stress-strain characteristics of the EPS. This cyclical loading is
expected to continue for the life of the structure, with maximum load increasing slowly over
years as slide movement occurs.
The total load applied to the reservoir by the CI is controlled by the stress-strain properties of
the EPS material, the thickness of the CI, and the amount that the landslide compresses the
inclusion. Based on an average rate of slide movement of 1/8 inch (3.2 mm) per year, the creep
movement of the slide over a 75-year design life is 9.4 inches (240 mm). Estimated seismic
displacements for a design (2,475-year) event are up to 15 inches (381 mm) at the toe of MSE
wall, which combines landslide and MSE wall seismic performance. The combined displacement
at the end of the CI’s design life is the sum of landslide creep movement and the upper limit of
seismic movements. For the CI as installed, the combined movements equate to strains of about
33 percent.
The useful life of the CI is controlled by the actual displacement of the slide rather than the
length of time it is in service. Change in thickness of the CI will be monitored using
extensometers to determine when it reaches its useful life. Three sets of long-range displacement
transducers will be used to measure the actual compression of the inclusion. Displacement data
will be recorded by dataloggers that can be accessed via below-grade vaults. Once the CI reaches
its maximum allowable compression, the EPS material can be removed and replaced with a
similar material to extend the useful life of the facility. The adjacent self-stable MSE wall will
provide support to the earth materials while the CI is “serviced”. The width of the second CI will
not be as large as the first, thus the useful life will be shorter. Alternatively, the MSE wall could
be reconstructed further away from the reservoir, and additional design life could be realized.
Seismic Response
During a seismic event, the reservoir will generate seismic loads due to its mass, along with
additional hydro-dynamic forces from the water sloshing in the internal cells of the reservoir.
These forces will cause the reservoir to move out-of-phase with the foundation soils, the
landslide/MSE wall system, and the tied-back soldier pile walls. As the reservoir displaces into
the EPS, it will develop resisting forces within the EPS that counteracts the seismic
displacement. This force will be widely distributed across the entire wall area, and is
proportional to the magnitude and rate of displacement into the EPS. In this way, the EPS will
act as a damper or “seismic buffer” that will counteract reservoir displacements. While the
resistance provided by the EPS is relatively small, when combined with the large area of the
reservoir walls, this effect is capable of generating large resisting forces. By taking advantage of
this behavior, the lateral loading demand on the drilled shaft foundations is reduced, resulting in
© ASCE
Geo-Congress 2020
a more cost-effective seismic design. The rate-dependent response of the EPS will provide a
stiffer and stronger response to displacements in the relatively short-duration earthquake, and
return to a softer and lower response once the normal slow rate of landslide creep resumes.
While the landslide load on the reservoir could be avoided altogether through the use of an air-
gap, utilizing the large-strain properties of EPS to dampen the seismic response resulted in an
overall savings to the project.
CONCLUSION
The Portland Water Bureau is replacing existing open-water reservoirs with a single buried
reinforced concrete reservoir. The reservoirs lies near the toe of a large, ancient landslide.
Structural mitigation or relocation of the reservoir were not practical, so the redeveloped site is
being designed to accommodate continued creep of the landslide, as well as seismic deformation
from the Cascadia Subduction Zone earthquake with a return period of 2,475 years. The
landslide deformations are being accommodated by using a compressible inclusion consisting of
Expanded Polystyrene (EPS) blocks. This material will compress under landslide displacement,
and transfer predictable loads to the reservoir walls that are much lower in magnitude than earth
pressures. The presence of the compressible inclusion reduces demand on the reservoir’s drilled
shaft foundations, allowing for a more efficient and cost-effective design. The performance of
the landslide will be monitored using conventional slope inclinometers. These instruments will
be supplemented by long-range displacement transducers to measure the thickness of the
compressible inclusion, which can be used to infer the actual load applied to the reservoir wall
over time.
REFERENCES
Clarke, D.D. (1904). “A Phenomenal Landslide”. Trans. American Society of Civil Engineers,
Vol. 53, Paper no. 984, pp 322-397.
Clarke, D.D. (1917). “A Phenomenal Landslide – Supplement”. Proceedings of the American
Society of Civil Engineers, Vol 43, Issue 7, pp 1445-1474.
Kulhawy, F.H., Mayne, P.W. (1990). Manual on Estimating Soil Properties for Foundation
Design. Electric Power Research Institute, Palo Alto, California.
Lentz, R.T., (1981). “The Petrology and Stratigraphy of the Portland Hills Silt – A Pacific
Northwest Loess.” Oregon Geology. Vol 43, No. 1, pp 3-10.
Madin, I.P., (2009). “Portland, Oregon, Geology by Tram, Train, and Foot”. Oregon Geology.
Vol. 69, No. 1, pp 73-92.
Newmark, N.M (1965). “Effects of Earthquakes on Dams and Embankments”. Geotechnique.
Vol 15, pp 139-160
Vessely, D.A., Cornforth, D.H. (1998). “Estimating Seismic Displacements of Marginally Stable
Landslides Using Newmark Approach.” Specialty Conference on Geotechnical Earthquake
Engineering and Soil Dynamics III. Geotechnical Special Publication #75, Vol 1. pp 800-
811.
© ASCE
Geo-Congress 2020
A Retrospective on the Evolution of Geotechnical Sensing and Instrumentation for

Monitoring at MnDOT
Joel N. Swenson, P.E.1; and Derrick D. Dasenbrock, P.E., D.GE, F.ASCE 2
1
Barr Engineering Co., Minneapolis, MN. E-mail: jswenson@barr.com
2
Geotechnical Engineering Section, Minnesota DOT Office of Materials and Road Research,
Maplewood, MN. E-mail: derrick.dasenbrock@state.mn.us
ABSTRACT
The Minnesota Department of Transportation (MnDOT), other transportation asset owners,
and geotechnical consultants are embracing sensing and monitoring to add project value and
reduce risk. Historically, measurements to assist with field observations of performance have
occurred since the 1930s and 1940s, and in those early years, engineers in the field relied on
hydraulic and mechanical systems to measure field behavior or performance of geotechnical
designs during and after construction. In the decades since, there have been significant
advancements made in communications, and computing technologies which have enhanced
monitoring capabilities in terms of accuracy, precision, reliability, size, cost, networking, and
functionality. With these advances and benefits have also come unexpected challenges and
opportunities. A retrospective on geotechnical monitoring at MnDOT, a discussion on
technologies for applications today, and thoughts on the future of the geotechnical monitoring
practice in the geotechnical transportation discipline will be presented.
INTRODUCTION
Geotechnical instrumentation in highway construction is not new; it aids design by
evaluating the behavior of various types of construction such as soil and rock slopes,
embankments, shallow and deep foundations, earth retaining structures, and ground
modification. Geotechnical engineers have been using a variety of measurement methods and
tools for decades. Usually, a geotechnical engineer identifies a site with an unusual or
problematic nature to justify the use of field measurements as part of an overall site
characterization program. However, field instrumentation which remains in-place following an
exploration program is far less common than short-term programs to characterize the site.
Conditions of unusual groundwater, settlement behavior, instability, or character that changes
with time often leads to additional, follow-up site investigations, with sensors or equipment left
in-place for measurements to be collected over time.
There are two general categories of geotechnical measuring instruments (Dunnicliff, 1998).
The first category is used for subsurface investigations to determine soil or rock properties (i.e.
strength, compressibility, and permeability) normally during the design phase of a project. The
second category is used for monitoring performance, normally during the predesign phase of a
project where there is risk of poor performance due to groundwater, settlement, or slope
instability, or during the construction phase, or when the facility is in-service. This paper
describes MnDOT’s application of sensors and equipment related to the second category.
Peck (1988, 1993) indicates that every geotechnical design is, to some extent, hypothetical,
and every construction job involving earth or rock runs the risk of encountering unexpected
conditions. These circumstances are a direct consequence of working below ground with natural
materials, created and placed by processes seldom providing uniform conditions. The inability of
© ASCE
Geo-Congress 2020
traditional exploratory procedures, which usually assess an incredibly small volume of material
as compared to that impacted by transportation projects, to detect in advance all the possibly
significant properties and conditions of natural materials requires designers to make assumptions
that may differ than what will be encountered in the field.
Field observations, including quantitative measurements obtained by geotechnical field
instrumentation, provides information such that, in spite of these limitations, a project can be
designed to be safe, efficient, and without excessive conservatism. Field instrumentation is
perhaps unusually useful and relevant to the practice of geotechnics, in contrast to many other
branches of engineering where there is greater material control and uniformity.

Dunnicliff (1998) explains that the engineering implementation and use of geotechnical
instrumentation involves a combination of the capabilities of measuring instruments and the
capabilities of people. During the past few decades, manufacturers of geotechnical
instrumentation have developed a large assortment of valuable and versatile products for the
monitoring of geotechnical parameters. Those unfamiliar with instrumentation might believe that
obtaining needed information entails nothing more than pulling an instrument from a shelf,
installing it, and collecting readings. Although successful utilization may at first appear simple
and straightforward, considerable engineering and planning are required to obtain the desired end
results.
HISTORICAL PRACTICE
In 1905 the Minnesota legislature created the State Highway Commission to build roads and
bridges; in 1917 that commission was abolished and the legislature created the Minnesota
Department of Highways. In 1959, records indicate that the Geotechnical Section was started and
the first (Unique Boring #1) of what would become many thousands of soil borings was
advanced.
Historic project files show that in addition to soil borings, sampling, and site exploration,
geotechnical field instrumentation was also used in the early days of the section to assess water
levels, settlement, and slope movements. Groundwater monitoring systems typically consisted of
open-hole standpipe piezometers, with water depth measurements collected using a weighted
measuring tape. Settlement was measured with a rod and level and settlement plates; these plates
consisted of metal pipes attached to plywood bases installed at the base of an excavation prior to
constructing an earth embankment. Slope movements were measured by placing stakes in the
ground to track surface movements. The first use of traversing probe inclinometers likely
occurred in the early 1970’s after the servo-accelerometer-type probe was introduced to the
industry in 1969 (Machan and Bennett, 2008). These methods were largely performed by hand
with readings tabulated on specially formatted paper records. In 1976, the Minnesota Department
of Transportation was created, and it would be another decade before electronic systems were
seen in the lab (for data collection of consolidation and shear testing) and field to aid in data
collection starting with pneumatic piezometers and traversing probe inclinometers.
Field instrumentation was used relatively sparingly through the late 1990’s and exclusively
for preconstruction project development, geotechnical site evaluation, landslide investigations,
and settlement of embankments during construction, as opposed to longer-term performance
monitoring of built works. Perhaps engineers of the time were reflecting on what Ralph Peck
(1984) stated, "The legitimate uses of instrumentation are so many, and the questions that
instruments and observation can answer so vital, that we should not risk discrediting their value
by using them improperly or unnecessarily." Alternately, it could have been that most methods
© ASCE
Geo-Congress 2020
of the time were comparatively labor intensive and time consuming. It appears that most sites
were constructed conservatively with comparatively flat slopes and often removing problematic
soils to the greatest extent possible. Figure 1 shows manual field monitoring systems that were
used prior to the 1990’s, including use of measuring tapes, rods, levels, and settlement plates for
the measurement of slope and embankment movements and water level.
Figure 1. Early monitoring systems: (left) assessing a slope failure using a pole and
measuring tape; (left center) measuring pavement deformation with a rod and level; (right
center) using a weighted tape to measure water level in a standpipe piezometer; and (right)
a settlement plate.
Electronic readout systems were introduced to MnDOT when inclinometers and vibrating
wire type sensors (piezometers, strain gauges) were acquired in the 1990’s. Recording systems,
where data was recorded and later downloaded to a computer, were introduced first for
inclinometers and later for piezometers and other sensors. Figure 2 shows MnDOT engineers
using traversing probe inclinometers on landslide sites in the later 1990’s and early 2000’s.
These systems generally had proprietary hardware and software interfaces and consisted of
custom battery powered electronic data logger boxes, which were downloaded to a computer
workstation, usually using a serial data cable, after a series of readings were collected. Figure 3
shows a typical drive-point piezometer and associated field logger system. In the mid-2000’s use
of field instrumentation became more common as more geotechnical projects with uncommon or
unusual designs were being designed. The first larger projects with more sensors and more
sensor types were column supported embankments and soil surcharge projects. Starting in the
late 2000’s, more automated systems that could record data from multiple instruments and
instrument types were used. In the past 5 years, many systems using cellular modems and web-
based data evaluation and presentation have become common. The number and types of sensors
in use has generally increased, and more information is available for project scoping, engineering
design, or performance evaluation. Today, in addition to providing foundation design
recommendations for bridges and culverts, MnDOT’s geotechnical staff also routinely
establishes instrumentation and performance monitoring programs for slopes, embankments, and
structures where they can provide engineering value and reduce project cost or risk.
One of the earliest performance monitoring applications and the earliest use of multiple
electronic sensors of different types with data collection at a centralized location was in 2004 at
the Glen Road Interchange project in Newport, MN. Here consulting engineers for the contractor
installed earth pressure cells, extensometers, and soil strainmeters to evaluate the performance of
a deep mixing method column supported embankment. Figure 4 shows the instrumentation
© ASCE
Geo-Congress 2020
layout and the project site where retaining walls and embankment fills were being placed for a
new bridge.
Figure 2. Traversing probe inclinometers: (left) collecting data manually from a traversing
probe; (right) placing the traversing probe inclinometer in the slotted casing within an
outer protective casing.
Figure 3. Vibrating wire piezometers: (left) a drive point piezomenter and logger; (center)
the logger box in the field; (right) retrieving data from the logger.
Shortly after that project, and building on its success, a geotechnical solution was provided
for a roadway widening project on MN 241 in St. Michael, replacing a proposed bridge to deal
with relatively deep deposits of soft soils and peat. Unexpectedly poor soil conditions were
found when geotechnical investigations were conducted near wetlands adjacent to a two lane
roadway. A combination pile-supported embankment, lightweight fill, and excavation with soil
pre-load and surcharge was designed to eliminate the need for a multi-span bridge. As part of the
project, a monitoring program was included to show that settlement was effectively controlled to
within the established tolerances.
Figure 5 shows one portion of the project (the column supported embankment below the
westbound lanes) was monitored independently by researchers and the University of Minnesota
as part of a research project (Wachman and Labuz, 2008). For the research project, 48 sensors
were installed, including; strain gauges, earth pressure cells, and settlement systems. The center
© ASCE
Geo-Congress 2020
portion of the roadway where EPS geofoam was not instrumented. The eastbound lanes were
constructed over an area where several feet of highly compressible soils were removed and the
remaining soils were treated with a soil pre-load and surcharge. This portion of the project was
monitored by MnDOT and included earth pressure cells, borehole settlement cells and the first
use of a manually-read fiber optic sensing geosynthetic. Several strips of TenCate GeoDetect
fabric were installed to test the potential of this system for future MnDOT use. This was also the
first trial of a cellular data connection to transfer data to a remote server. A Geocomp iSite
remote monitoring system was installed in the data collection box.
Figure 4. Site monitoring during construction: (left) Layout of sensors to monitor the
performance of Deep Mixing Method soil columns supporting a roadway embankment and
spread footings; (right) construction after sensors are installed. Sensors are read manually
at a data collection box.
Figure 5. Automated data collection and uploading: (left) The data collection and cell
modem station for the MN 241 soil surcharge and column supported embankment
monitoring project in 2006; (center) the solar panel on the EB data collection box and
conduit carrying sensor cables to the sensors monitoring the soil surcharge; (right)
placement of fill over piling for the RSS embankment on the WB roadway.
There were a number of lessons learned on this project that would significantly improve
future installations. The need for trained and qualified engineers and technicians to design,
install, calibrate, test, and maintain the systems became evident as each phase of the installation
progressed. While many of the sensors performed well, the settlement monitoring systems were
particularly prone to erratic behavior. The vibrating-wire type sensors were found to be the most
reliable. The cellular modem connection proved to be inconsistent, likely due to hardware,
software, and service provider compatibility issues. The system did not perform as intended for
© ASCE
Geo-Congress 2020
long. Other, more solvable, but unforeseen problems occurred, such as the cabinet becoming
inundated with water during winter snow melt and spring flooding. The importance of tightly
sealing cabinets was further reinforced when wires were found to have been eaten by mice.
As familiarity with field instrumentation and monitoring increased, it was applied to more
construction projects where beneficial information could be gained on performance. As part of a
design-build project, the contractor proposed eliminating a large amount of piling at two bridge
abutments by adopting a spread footing design. The MnDOT Bridge Office was concerned that a
tall abutment on spread footings may rotate or tilt excessively, but accepted the change. The
performance was monitored by the contractor using survey targets and an inclinometer. Figure 6
shows how performance was independently monitored by MnDOT using tiltmeters and earth
pressure cells (Bentler et al. 2009).
Figure 6. Performance monitoring at CR 110 on the US 212 design-build project: (left)

spread footings replaced piling at the bridge abutments and an instrumentation plan was
designed to monitor the performance of the tall abutments; (center) the south abutment
(top right) a tiltmeter on the wall stem; (right bottom) an earth pressure cell positioned
below the abutment toe.
PRESENT PRACTICE
By the mid-2000’s small-scale geotechnical monitoring was common; manual traversing
probe inclinometer installations and standpipe piezometers were the most frequent field
installations. However, a transitional period occurred in conjunction with MnDOT employing
cone penetrometer testing (Dasenbrock, 2019) and installing drive-point piezometers with data
loggers, and advances in communications hardware, software and data visualization. Two
contributions to the momentum behind this transition were advances in the fully-grouted method
with diaphragm piezometers and the in-place inclinometer.
Dunnicliff (2008) compares the fully-grouted method to the Casagrande piezometer or open
standpipe piezometer techniques by stating, “Forget the sand and bentonite seal! This is no
longer the way to go! Use the fully-grouted method!” The fully-grouted method consists of
“installing a piezometer tip in a borehole which is backfilled entirely with cement-bentonite
grout” as described by Vaugh (1969), Dunnicliff (1993), Mikkelson (2003), and Contreras,
Grosser, and Ver Strate (2007).
In a fully-grouted installation, the diaphragm piezometer sensor is set within a cement-
bentonite-grout, and the combination of a properly prepared sensor and cement-bentonite-grout
allows for measuring rapid pore water pressure changes in low permeability soils. Contreras,
et.al (2007) states, “The high void ratio and low permeability are two reasons the fully-grouted
© ASCE
Geo-Congress 2020
method works; it allows transmission of a low volume of water over a short distance yet
maintains overall low permeability in the vertical direction. With grout mixes, the cement has a
greater influence on the void ratio than the bentonite and is considered the controlling factor in
the permeability of the grout.” Other benefits to this technique are that pore water pressure
measurements are possible at specific depths, multiple diaphragm piezometers can be installed in
a singular borehole for gradient determination, also known as a “nested” configuration, and using
a single borehole to install more than one type of instrument, for example, vibrating wire
piezometers and inclinometer casing for pore water pressure increase and slip plane
determination.
One of the geotechnical instruments that coincided with this transition, was the ShapeArray
by Measurand in 2006. The sensor was previously named the Shape AccelArray, from which the
acronym “SAA” was derived; however, the name has been recently shortened to ShapeArray and
the acronym SAA remains. The SAA is an in-place inclinometer consisting of rigid segments
connected by flexible jointing installed within vertical inclinometer casing or conduit, or in
horizontal or circular arrays depending on the application. Each segment contains a MEMs based
accelerometer and measures acceleration used for computation of angles over the array such that
incremental or cumulative displacements over time are computed. In its earlier uses, MnDOT
had elected to install dataloggers onsite and manually download the data for use in the site
investigation.
Figure 7. A MEMS ShapeArray (SAA) installation: (left) the sensor is being installed by
MnDOT personnel at a site on I-35E; (right) a logger with SAA information is being
downloaded with a notebook computer at the site.
In 2006, funding was procured for a demonstration project in Crookston, MN, where three
SAAs were installed in an active landslide area. Immediately, following a failure in 2003, several
inclinometer casings were installed and the area was monitored by a consultant for a time
following the slide. After the close of that contract, MnDOT continued to read the inclinometers;
movement continued and the traversing probe became stuck in one of the casings. This
demonstration project would be the first project where MnDOT successfully acquired data
remotely over a long-term period for site evaluation. The Crookson landslide monitoring project
was also the first use of a web-based interface where data was automatically updated and could
be viewed on a website with login credentials. The immediate success of this landslide
monitoring effort led the NW District of MnDOT to request additional SAA systems at another
© ASCE
Geo-Congress 2020
nearby landslide. At this second slide, the sensors captured the complete failure process of the
landslide including the evolution and progress of multiple failure planes and over 100-inches of
horizontal translation; these landslides are discussed further by Dasenbrock et al. (2012). SAA
sensors (Figure 7) are now commonly used to evaluate landslide sites; since they were
introduced in 2007, more slope stability sites are monitored with automated instrumentation than
manual instrumentation systems.
Also in 2006, the first large use of automated data collection for project scoping, with more
than 10 piezometers was for a groundwater study, was on US 169 in Elk River, MN. At about
this same time, a research project to investigate the effects of draglaod on driven piling began.
As part of this project, strain gauges were welded to two piles on BR 27V74 located in the
“Crosstown Commons” where I-35W and MN 62 share a section of freeway in south
Minneapolis. This research project launched a series of subsequent pile performance monitoring
projects where strain gauges were installed in piles and loggers locally recorded the pile
performance during construction and service.
During this period, the value of geotechnical monitoring was becoming more appreciated by
other functional disciplines such as preliminary design, construction, and maintenance. As
projects grew in scope and number, some monitoring programs were now being designed by
consulting engineers for MnDOT and executed by consulting engineers during the construction
phase of projects. The next significant advance for geotechnical monitoring for MnDOT was
when it became a part of design-build contracts for the Hastings Bridge (US 61) reconstruction
and other bridge, roadway, and slope repair projects. A distinct work-type was created for
Geotechnical Instrumentation as part of MnDOT’s prequalification program for professional
technical consultant services.
Figure 8. AMTS systems: (left) an AMTS monitoring slope movement at a rock cut site;
(center) an AMTS monitoring an embankment fill; (right) an AMTS prism at an
embankment fill site over potentially compressible soils.
MnDOT would gradually add new technologies to the monitoring portfolio and discourage
techniques that resulted in poor quality data. In 2012, MnDOT performed a trial monitoring
project using a Trimble S8 automated motorized total station (AMTS) unit and a TSC 3 data
collector on the I-494/US 169 project. MnDOT staff had a difficult time keeping the system
operating and transferring data due to significant IT infrastructure challenges. After this
experience, while the technique was determined to be promising, it was determined that it would
be more efficient to contract this type of operation, as there was insufficient in-house expertise.
Later, comprehensive monitoring programs with both AMTS systems and varieties of contact
© ASCE
Geo-Congress 2020
sensors were deployed on the Rock County GRS-IBS Bridge project between 2013 to 2016
(Swenson, 2016), I-35W Cayuga Interchange Forcemain project 2014 to 2015, High Bridge
project 2017 to 2018, and US 53 BR 69129 and 69130 (Swenson, 2018) (left and center shown
in Figure 8) project starting 2016.
GNSS receiver systems (Figure 9) have been used successfully to monitor deformation on
projects beginning in 2012, first on the MN 36 (Beach Road Bridge), then the US 14 Florence
Flood Mitigation, and on the I-35W Minnesota River Bridge North Abutment. Due to poor
performance, usually from being impacted by construction equipment, settlement plates are
discouraged in favor of other techniques in critical applications. Similarly, fluid-reservoir

settlement measurement systems are also discouraged due to their generally poor performance
(they tend to show erratic and inconsistent data).
Figure 9. GNSS systems: (left) a GNSS system measuring settlement over a coal slag fill site
during a surcharge; (right) a GNSS system measuring the rate and magnitude of soil
settlement during surcharge and prefabricated vertical drain system.
Figure 10. Sites where active monitoring is more critical: (left) US 169 causeway project;
(left center) US 63 Red Wing river bridge north abutment embankment widening; (center
right) US 2 Crookston landslide stabilization project; (right) I-35 W Minnesota
The first use of monitoring thresholds and alarms came in 2015, and simultaneously on the
© ASCE
Geo-Congress 2020
MN 210 slope stabilization project (Swenson, 2019) and the US 53 bridge replacement project
(Swenson, 2018; Theroux, 2018). These two projects had enough infrastructure that for the first
time significant contractor effort was directed toward fulfilling the monitoring and
instrumentation requirements. The MN 210 project was MnDOT’s first geotechnical
instrumentation ‘mega-project’ where more than $1M was directed toward evaluating the
performance of 74 landslide sites as part of a design-build contract. For the MN210 project, the
contractor’s consultant employed a relatively new platform at the time with multi-sensor
integration, plug and play capabilities, auto-mesh networking systems, cloud based data
warehousing, and online visualization and alarms due to the site geography and aggressive
construction schedule. As MnDOT continues to employ instrumentation, infrastructure similar to
those in Figure 10 can be seen along Minnesota’s highways housing dataloggers, solar power,
and telemetry equipment to collect and transmit data to servers and online data visualization.
Figure 11. A web based application: (top) sensor locations are shown on an aerial photo
cone resistance; (bottom) a project “dashboard” is established showing selected sensor
information from multiple sensor types, plotted over a time period of interest.
One of the more significant benefits, which is perhaps underappreciated is the cost and time
savings associated with having data available via a web-based interface. Users do not need to
have proprietary software programs loaded on their computers, and information can be accessed
© ASCE
Geo-Congress 2020
from any computer connected to the internet via a web browser. Additionally, there is a
considerable savings of time since data does not need to be manually collected, downloaded,
transferred, uploaded, processed, graphed, and made available to users. This was often a time
consuming process and there was generally a delay associated with the processing of the
information. End users would also then need to keep track on an increasingly large number of
spreadsheets and graphs. Today’s tools allow users to select sensors and time periods of interest
and plot data automatically. Figure 11 shows a collection of sensors from a project site with
multiple sensor types.
FUTURE PRACTICE
MnDOT has used a project-based approach to geotechnical instrumentation and monitoring.
Each project was designed, installed, and managed from a project standpoint; this methodology
was consistent with the way the geotechnical exploration portion of the work was organized and
managed. As the size and complexity of projects has increased, it is becoming clear that a more
systematic approach to geotechnical instrumentation and monitoring as a whole would provide
improved organization and improved quality.
MnDOT is in the early stages of establishing a geotechnical asset management program, and
geotechnical instrumentation may become an asset class as part of that effort, leading to
improved tracking of the physical locations of project installations.
Development of geotechnical sensors and systems will continue to improve. Attached sensors
will be supplemented with remote sensing systems. In addition to improved sensing,
communication, and visualization tools, other opportunities are expected to include:
 Creation of universal data records viewable across visualization platforms in the same
manner as electronic data for the DIGGS initiative (Dasenbrock, 2019)
 Automatically updating geotechnical computational models using field data such as
piezometric pressures and deformation.
 Proactive monitoring-predictive analytics (neural networks); predictive modeling to
adjust reading frequency and auto-establish alarms and notifications based on unusual
data or changes in trends.
 Operations and management (performance degradation curves) to help in capital planning
 Centralized data management and information sharing similar to applications today for
traffic information.
 Incorporation of near real-time geotechnical instrumentation data on local servers or
cloud-based visualization programs into augmented reality platforms, in the near real-
time.
CHALLENGES
Where previously the challenge was not enough good field data, current challenges include
too much data to efficiently manage across multiple projects. Cell modem data connections are
improving and while technology is making them easier to set up and integrate into monitoring
systems, paying the bills and maintaining the systems is an ongoing challenge, particularly when
staff changes and new engineers are unfamiliar with this aspect of the work.
As systems evolve, it is likely that MnDOT will be challenged to keep pace with industry
advancements (sensing technologies, visualization improvements) as staffing levels remain
constant and demand for services increases with time. In addition, as IT costs continue to
© ASCE
Geo-Congress 2020
increase, it is likely that the value of having and maintaining these systems will need to be
systematically quantified and defended in an expected era of declining revenue and increased
expectations to deliver transportation projects inexpensively and quickly. The NCHRP Synthesis
484, Influence of Geotechnical Investigation and Subsurface Conditions on Claims, Change
Orders, and Overruns, identifies that geotechnical problems can significantly impact project cost
and schedule (Boeckmann and Loehr 2016). While there are often drivers to accelerate project
timelines with limited information, assuming that additional information is of comparatively
lesser value than the project timeline, it will continue to be a challenge to ensure that there is
enough time and motivation to gain the necessary insight from field instrumentation and
monitoring where these techniques will reduce project risk, reduce delays, and reduce cost.
New technologies have created challenges in planning, budgeting, and interacting with
professionals in different disciplines (such as IT professionals). Data management is becoming
increasingly important in terms of organization, access, security, volume and storage
requirements, retention, and requirements for redundancy, back-ups, and online security.
From the engineering side, calibration, maintenance and installation of instruments,
collection of instrumentation data, processing and presentation of collected data, interpretation of
processed data and reporting of the results are ongoing challenges, made increasingly important
by the easier access to the information from all members of project teams, on-demand. Dunncliff
(2018) encourages you to pay more attention to human factors in the future than you have in the
past. As Ralph Peck said to us, “We need to carry out a vast amount of observational work, but
what we do should be done for a purpose, and be done well.”
CONCLUSIONS
Field instrumentation and monitoring, following a well-developed and executed site
investigation program, is often useful and important, if not critical, at sites with conditions that
can change with time. These time dependent effects can be related to settlement, groundwater
fluctuations, slope instability or similar conditions. Site monitoring is part of an overall
systematic approach to reducing project risk- and geotechnical risks in transportation
infrastructure construction, and without, risks to the project can manifest in schedule delays and
cost overruns. As with more traditional investigations, the scope of work can vary significantly
and should be tailored to the particular project needs.
Technology is moving at a rapid pace and readily available electronic data and information is
changing geotechnical engineering among other disciplines. While there have been significant
advances in sensors and systems to monitor stress, pressure, load, displacement, tilt, position,
ground motion, and other qualities and behaviors, the most significant developments have been
in data storage, accessibility, exchange, integration, presentation and visualization. While these
advances have been significant for geotechnical engineers themselves, they are perhaps more
significant to project peers, partners, and the wider engineering and project community.
Improved visualization has helped improve discussions of probability, spatial variation, sample
size, risk, reliability, and cost and schedule implications. Indeed, the most critical advances in the
profession are likely to be those that have impacted how we communicate our information
outside the profession.
ACKNOWLEDGMENTS
The authors appreciate the support provided by MnDOT and the consulting engineering and
construction firms who have designed, installed, and operated successful geotechnical
© ASCE
Geo-Congress 2020
monitoring and instrumentation programs for use in predicting or evaluating performance of

slopes, embankments, and structures. Particular appreciation is given to organizational support
from entities including but not limited to FHWA, TRB, ASCE, DFI for their efforts in providing
technology transfer opportunities through publications, conferences, seminars, webinars, and
face to face meetings, for the distribution and sharing of information.
REFERENCES
Bentler, J., Dasenbrock, D. D., and Hoppe, M. (2009). Analysis and Performance Monitoring of
a Spread Footing Bridge Foundation, 2009 International Foundation Congress and
Equipment Expo (IFCEE’09), March 15-19, 2009, pgs 473-480.
Contreras, I.A., Grosser, A.T., and VerStrate, R.H. (2007). “The Use of the Fully-grouted
Method for Piezometer Installation.” Proceedings of the Seventh International Symposium
on Field Measurements in Geomechanics. Boston, MA. ASCE Geotechnical Special
Publication 175.
D. D. Dasenbrock. (2010) “Automated Landslide Instrumentation Programs on US Route 2 in
Crookston, MN”. In: Proceedings of the Annual Conference of the Minnesota Geotechnical
Society. pp. 165–185.
D. D. Dasenbrock (2019). “The Past, Present and Future of Cone Penetration Testing for
MnDOT Highway Projects.” Roads and Bridges. 3030 W. Salt Creek Lane #201, Arlington
Heights, IL 60005. March 25.
Dunnicliff, J. (1988, 1993). Geotechnical Instrumentation for Monitoring Field Performance.
John Wiley & Sons, Inc., New York, 577 p.
Dunnicliff, J. (1998). Geotechnical Instrumentation Reference Manual, FHWA Report No.
FHWA-HI-98-034, National Highway Institute, U. S. Department of Transportation, Federal
Highway Administration, Washington, D.C., 238 pages.
Dunnicliff, J. (2018). Geotechnical Instrumentation News. Some remarks on the importance of
human factors in geotechnical and structural monitoring programs. June 2018, p. 25.
Machan, G. and Bennett, V.G., Use of Inclinometers for Geotechnical Instrumentation on
Transpiration Projects, State of the Practice, Transportation Research Board Circular,
Number E-C129, October 2008
Mikkelsen, P.E. and Green, E.G. (2003). “Piezometers in Fully Grouted Boreholes.”
International Symposium on Geomechanics. Oslo, Norway. September 2003.
Peck, R. B. (1984). Observation and Instrumentation, Some Elementary Considerations, 1983
postscript. Judgment in Geotechnical Engineering: The Professional Legacy of Ralph B.
Peck, Dunnicliff, J. and Deere, D. U., eds., John Wiley & Sons, Inc., New York, 128-130.
Peck, R. B. (1988, 1993). Foreword: Geotechnical Instrumentation for Monitoring Field
Performance. by Dunnicliff, J., John Wiley & Sons, Inc., New York, vii-ix.
Swenson, J.N., Dasenbrock, D.D., Bryant, E.K., Grosser, A.T., and Budge, A.. (2016).
“Performance Study of Minnesota’s 1st GRS-IBS Structure in Rock County.” Proceedings of
the 65th Annual Minnesota Geotechnical Society Annual Conference, University of
Minnesota.
Swenson, J.N., Dasenbrock, D.D., Iverson, N.W. and Axtell, P.J..(2018). “Geotechnical
Performance Monitoring of Foundations and Rock Slopes on the US 53 Realignment in
Virginia, Minnesota.” Proceedings of the 66th Annual Minnesota Geotechnical Society
Annual Conference, University of Minnesota.
Swenson, J.N., Dasenbrock, D.D., and Provost, D.A., (2019). “Geotechnical Monitoring: A Key
© ASCE
Geo-Congress 2020
Element of the Creative and Effective Landslide Remediation Solutions for Minnesota
Highway 210 in J. Cooke State Park.” Proceedings of the 67 th Annual Minnesota
Geotechnical Society Annual Conference, University of Minnesota.
Theroux, B. and Flores, H., (2018). “Shallow Bridge Foundations on Large Rock Fill.”
Proceedings: 66th Annual Geotechnical Engineering Conference, Minnesota Geotechnical
Society, February 23, 2018.
Vaughan, P. R. (1969), “A Note on Sealing Piezometers in Boreholes”, Geotechnique, Vol. 19,
No. 3, pp. 405-413.
Wachman, G. and Labuz, J.L., TH 241 Column-Supported Embankment, Center for

Transportation Studies Final Report CTS 08-11, CTS project # 2006054, June 2008
© ASCE
Geo-Congress 2020
Experimental Study of Forces Induced in Mechanical Excavation of Rock

John Pultorak1; Dmitry Drozdov2; Jia-Liang Le, P.E., M.ASCE3;
and Emmanuel Detournay, A.M.ASCE4
1
mail: pulto001@umn.edu
2
mail: drozd018@umn.edu
3
mail: jle@umn.edu
4
mail: detou001@umn.edu
ABSTRACT
This paper presents preliminary results of an experimental campaign aimed at mapping the
dependence of the cutting force on the depth of cut in scratch tests performed with a sharp cutter.
Tests conducted in Berea sandstone and Indiana limestone confirm that the scaling of the force
with the depth of cut depends on the cutting regime. They also show a dependence of the nature
of the frequency distribution of the cutting force on the modes of failure.
INTRODUCTION
The drive to mechanize the excavation of hard rocks has brought to the forefront the need to
quantify the mechanics of tool-rock interaction. Indeed, machine design relies on our ability to
predict the average cutting force as well as the expected force fluctuations corresponding to
given operating parameters and rock properties (Nishimatsu 1993, Fowell 1993, Nelson 1993).
In this regard, the following fundamental question arises: what is the dependence of the cutting
force (i) on the depth of cut and the cutter geometry, and (ii) on the strength and fracture
properties of the rock. Rephrased in terms of the energy spent per volume of rock removed, how
does the specific energy (Teale 1965) scale with depth of cut and rock properties. Incidentally, a
possible dependence of the specific energy on the depth of cut is usually ignored in the literature
on mechanical excavation, where attempts to correlate the specific energy to rock properties are
presented (Tiryaki et al 2009).
A related issue is the interpretation of the scratch test, which involves tracing a groove on the
surface of a specimen with a cutting tool. The scratch test is typically conducted under kinematic
control: the depth of the cut d (or depth of the groove) and the cutter velocity v (tangential to the
sample surface) are imposed and maintained constant along the entire cut, while the magnitude
and orientation of the force acting on the cutter are measured. This experimental technique has
garnered interest lately, because it appears to offer a simple means to measure strength properties
of quasi-brittle materials (Richard et al 1998, Akono et al 2011, Richard et al 2012, Zhou and
Lin 2014). However, there is strong disagreement between researchers on how to interpret the
results (Lin and Zhou 2013, Akono et al 2014, Lin and Zhou 2015, Le and Detournay 2016). At
stake is which parameters can be determined in these experiments and how to interpret them
from the test data.
To address these fundamental questions, a research program has been initiated that is
exclusively concerned by the force on the cutting face of the tool, noting that in a controlled
© ASCE
Geo-Congress 2020
scratch test, a sufficiently sharp tool can always be used so that the frictional contact force is
negligible compared to the cutting force, or that the frictional force can be assessed from the
force measurements under certain conditions (Detournay and Defourny 1992, Richard et al
2012). Finally, a possible dependence of the cutting force on the velocity will not be
investigated, as cutting experiments on dry rocks do not show any significant rate effects over
the range of cutter velocity typically achieved in laboratory testing (Nishimatsu 1993, Fowell
1993). It is also convenient to introduce the specific energy , defined as the energy expended
per unit volume of fragmented rock; corresponds to the ratio of the average cutting force over
the cross-sectional area of the groove created by the motion of the cutter. Reformulated in terms
of , the fundamental question is the scaling of the specific energy with respect to depth of
cut d and the strength properties of the rock, namely unconfined compressive strength q and
toughness K Ic .
The paper is organized as follows. First we review the dependence of the rock failure mode
on the depth of cut. We then provide a short description of the scratch apparatus and of the
testing procedure. Finally, we give preliminary results of the transition from the ductile to the
fragmentation regime for tests conducted in Berea sandstone and Indiana limestone and discuss
change in the nature of the frequency distribution of the cutting force between the two regimes.
Figure 1: Different regimes of cutter/rock interaction with increasing depth of cut

CUTTING REGIMES
Rock cutting experiments suggest the existence of three distinct modes of failure for depth of
cut larger than the average grain size (Chaput 1992, Richard et al 1998, Richard et al 2012).
Each mode is characterized by a different dependence of the cutting force on the depth of cut and
on the rock strength parameters.
 Ductile regime: shallow depth of cut (typically less than 1 mm for a medium strength
sandstone), the rock is intensively sheared ahead of the cutter; this cutting regime is
mainly characterized by a de-cohesion of the constitutive matrix and grains, with grains
and powder accumulating progressively in front of the cutter, see Fig. 1(a). A
comprehensive laboratory experimental program has shown convincingly that the
specific energy ; does not depend on the depth of cut d at shallow depths of cut (or in
other words that the cutting force is proportional to d ) and that is well correlated to
© ASCE
Geo-Congress 2020
the UCS unconfined compressive strength (Richard et al 2012).

 Fragmentation regime: With increasing depth of cut, the rock breaks into fragments that
are distributed according to a power law over a significant range of sizes. The photograph
in Fig. 1(b) shows the particle size distribution curve for the fragments collected from a
scratch test on a slab of Tuffeau limestone with d  1.3 mm (Pena 2010). The fractal
dimension D f of the fragment-size distribution curve is about 2.5, a value broadly
consistent with D f determined for a variety of fragmentation processes (Turcotte 1986).
A limited set of experiments and numerical simulations suggests that a progressive

transition between the ductile and the fragmentation regime takes place around a depth of
cut d  1   K Ic / q  (Richard et al 1998, Huang and Detournay 2008, Zhou and Lin
2
2014). This transition length scale is of order O (1 mm) for medium strength sedimentary
rocks, and could presumably be of order O (0.1 mm) or less for hard rocks.
 Brittle regime: At larger depth of cut, brittle failure occurs with macroscopic cracks
initiating from the tool tip and propagating unstably ahead of the cutter, see Fig. 1(c).
Large chips are created by the propagation of sub-horizontal tensile cracks (i.e., sub-
parallel to the free surface). An asymptotic fracture mechanics analysis based on treating
the chips as beam-like structures indicates that the specific energy scales according to
 K Ic2 / Ed (Le and Detournay 2016). (The energy scales by d 1 under these
assumptions and not by d 1/2 as claimed by some researchers (Akono and Ulm 2011,
Akono et al 2011). The erroneous scaling stems from not recognizing the cyclic nature of
the chipping process in rock cutting.) Presumably the transition depth of cut from the
fragmentation to the brittle regime, 2 , is proportional to K Ic2 / E 2 , but the (large)
magnitude of the proportionality factor is currently unknown. There are hints, however,
that transition depth of cut 2 , is of order O  1  10  mm in hard rocks.
SCRATCH APPARATUS
The scratch apparatus used for the experiments is a commercial version of the Rock Strength
Device that was developed at the University of Minnesota in the late 1990's (Detournay et al
1997), see Fig. 2. The RSD scratches the surface of rock samples under precise kinematic
control, while enabling the accurate measurements of the force acting on the cutter. The main
components of the frame are: a traverse with a sample holder, a moving cart housing the vertical
positioning system, the load cell, and the cutting element. The horizontal movement of the cart is
operated by a computer controlled stepper-motor driving a horizontal ball screw via a gearbox.
The depth of cut is adjusted manually with the vertical positioning system and a micrometer. A
locking system secures the vertical traveling mechanism against the frame, in order to maintain a
constant depth of cut while cutting.
A load sensor measures the components ( Fs , Fn ) of the cutting force F , which are
respectively parallel and normal to the cutter velocity. The complete system (sensor, data,
acquisition) achieves about 1 N of precision and resolution over the entire measurement range
[0-4000 N]. The scanning rate is typically set at 25 samples per millimeter travelled by the cutter.
The cutter velocity is set to v  20 mm/s. The force components are measured at a sampling
rate of 200 Hz, meaning 10 measurements of force per mm of cut. The sharp cutters used for the
tests reported in this paper are characterized by a width w  10 mm and a back rake angle
© ASCE
Geo-Congress 2020
weighted sum of random strengths of material elements along the failure surface. As a
consequence of the Central Limit Theorem, the resulting failure load would follow a Gaussian
distribution. Meanwhile, it should be noted that the Gaussian distribution cannot be applied to
the tail distribution since the failure load is non-negative. It has recently been suggested that the
far-left tail would exhibit a power-law behavior (Bazant and Le 2017, Le and Xu 2019).
When the cutting depth increases, the probability distribution of the cutting force clearly
shows a non-Gaussian nature (Fig. 4). It has been speculated that, at this range of cutting depths,
the dominant failure mode is material fragmentation. A simple analysis of energy balance
indicates that the cutting force could directly be related to the total area of fragments.
Meanwhile, the individual fragment size is expected to follow some probability distribution. For
example, the well-known model by Kolmogorov (1941) predicts a lognormal distribution of
fragment size. One essential point is that the random sizes of individual fragments are not
statistically independent, and thus one cannot directly apply the Central Limit Theorem to
determine the distribution of the total area of fragments. This is probably why we did not observe
a Gaussian distribution. To properly capture the statistical dependence of the fragment sizes, one
would need a stochastic model for the fragmentation process (e.g. Fowler and Scheu 2016).
CONCLUSION
Scratch tests conducted on different sedimentary rocks confirm the existence of at least two
asymptotic regimes of cutting: ductile for d  1 and fragmentation d  1 . The available
equipment does not permit to assess whether a brittle regime indeed exists for d  2 . A
preliminary analysis of the force statistics indicates that the functional form of the probability
distribution of cutting force varies with the cutting depth. This dependence is fundamentally
related to the prevalent failure mode. In view of the richness of the data measured in the scratch
test, it will be desirable to develop a probabilistic model, which would naturally yield the effect
of cutting depth on the mean specific energy.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge support from the National Science Foundation under
Grant No. 1742823. Any opinions, findings, and conclusions or recommendations expressed in
this material are those of the authors and do not necessarily reflect the views of the NSF.
REFERENCES
Akono, A.-T., P. M. Reis, and F.-J. Ulm (2011). Scratching as a fracture process: From butter to
steel. Physical Review Letters 106(20), 204302.
Akono, A.-T. and F.-J. Ulm (2011). Scratch test model for the determination of fracture tough-
ness. Engineering Fracture Mechanics 78(2), 334–342.
Akono, A.-T., F.-J. Ulm, and Z. Bažant (2014). Discussion: Strength-to-fracture scaling in
scratching. Engineering Fracture Mechanics 119, 21–28.
Bazant, Z. P. and J.-L. Le (2017) Probabilistic Mechanics of Quasibrittle Fracture: Strength,
Lifetime, and Size Effect, Cambridge University Press. Cambridge, U.K.
Chaput, E. J. (1992). Observations and analysis of hard rocks cutting failure mechanisms using
PDC cutters. Technical report, Imperial College, London, United Kingdom.
Detournay, E. and P. Defourny (1992). A phenomenological model of the drilling action of drag
bits. Int. J. Rock Mech. Min. Sci. 29(1), 13–23.
© ASCE
Geo-Congress 2020
Detournay, E., A. Drescher, and D. Hultman (1997), Portable Rock Strength Evaluation Device,
U.S. Patent 5,670,711.
Fowler, A. C. and B. Scheu (2016) “A theoretical explanation of grain size distributions in
explosive rock fragmentation”, Proc. Royal Soc., A, 472, 2015843.
Huang, H. and E. Detournay (2008). Intrinsic length scales in tool-rock interaction. Int. J.
Geomechanics 8(1), 39–44.
Fowell, R. J. (1993). The mechanics of rock cutting. In J. A. Hudson (Ed.), Comprehensive rock
engineering; principles, practice and projects, Chapter 7, pp. 155–176. Pergamon Press.
Kolmogorov, A.N. (1941). The logarithmically normal law of distribution of dimensions of

particles when broken into small parts. Proceedings of Academy of Sciences of URSS, 31(2),
99-101. (NASA Technical Translation, NASA TT F-12, 287, June 1969)
Le, J.-L. and E. Detournay (2016). Discussion on the “fracture mechanics interpretation of the
scratch test” by Akono et al. Engineering Fracture Mechanics 168, 46–50.
Le, J.-L. and Z. Xu (2019). A simplified probabilistic model for nanocrack propagation and its
implications for tail distribution of structural strength, Physical Mesomechanics, Special
Issue in Memory of Professor G. I. Barenblatt, 22(2), 85–95.
Lin, J.-S. and Y. Zhou (2013). Can scratch tests give fracture toughness? Engineering Fracture
Mechanics 109, 161–168.
Lin, J.-S. and Y. Zhou (2015). Rebuttal: Shallow wide groove scratch tests do not give fracture
toughness. Engineering Fracture Mechanics 133, 211–222.
Nelson, P. P. (1993). TBM performance analysis with reference to rock properties. In J. A.
Hudson (Ed.), Comprehensive rock engineering; principles, practice and projects, Chapter
10, pp. 261–291. Pergamon Press.
Nishimatsu, Y. (1993). Theories of rock cutting. In J. A. Hudson (Ed.), Comprehensive Rock
Engineering, Oxford, pp. 647–662. Vol. 1. Pergamon Press Ltd.
Peña, C. (2010). An experimental study of the fragmentation process in rock cutting. Master of
Science Thesis, Faculty of the Graduate School of the University of Minnesota, U.S.A.
Richard, T., E. Detournay, A. Drescher, P. Nicodeme, and D. Fourmaintraux (1998, July). The
scratch test as a means to measure strength of sedimentary rocks. In SPE/ISRM Eurock 98,
SPE 47196, Trondheim, Norway, pp. 1–8. Society of Petroleum Engineers.
Richard, T., F. Dagrain, E. Poyol, and E. Detournay (2012). Rock strength determination from
scratch tests. Engineering Geology 147–148(12), 91–100.
Teale, R. (1965). The concept of specific energy in rock drilling, International Journal of Rock
Mechanics and Mining Sciences & Geomechanics Abstracts, 2, 57–73.
Tiryaki, B., I. D. Gipps, and X. S. Li (2009). Rapid estimation of rock cuttability using fracture
toughness and rock strength. Advanced Materials Research 76-78, 591–596.
Turcotte, D. (1986). Fractals and fragmentation. Journal of Geophysical Research 91(B2), 1921–
1926.
Zhou, Y. and J.-S. Lin (2014). Modeling the ductile–brittle failure mode transition in rock
cutting. Engineering Fracture Mechanics 127, 135–147.
© ASCE
Geo-Congress 2020
An Overview of Performance Monitoring for Drilled Full Displacement Type Rigid

Inclusions under Highway Embankments
Liang Chern Chow, P.E., M.ASCE1; Joseph G. Bentler, P.E., M.ASCE2;
Alex Potter-Weight, P.E., M.ASCE3; and Andrew J. Eller, P.E., M.ASCE4
1
American Engineering Testing, Inc., MN. E-mail: lchow@amengtest.com
2
American Engineering Testing, Inc., MN. E-mail: jbentler@amengtest.com
3
Menard Group USA, Chicago, IL. E-mail: apotter-weight@menardgroupusa.com
4
Minnesota Dept. of Transportation, Maplewood, MN. E-mail: andrew.eller@state.mn.us
ABSTRACT
Column-supported embankments with rigid inclusions have been popular for many
transportation projects that include a combination of rapid timeline, environmental issues, and/or
specific subsurface conditions. In traditional embankments built on soft ground, the most
frequent uses of instrumentation are to monitor the progress of consolidation and to determine its
stability. Conversely, in column-supported embankments, different mechanisms and modes of
failure emerge despite minimal consolidation. Thus, the use of instrumentation must be altered to
address the anticipated behavior of column-supported embankments to effectively monitor
performance. Instrumentation programs for rigid inclusions are not standardized and highly
varied. This paper discusses the overall practices in instrumentation for drilled full displacement
type rigid inclusions for both national and international case studies followed by local experience
in Minnesota. After 3.5 years of continuous monitoring, the data were reassessed to evaluate the
actual soil-structural interaction and the maximum loads acting within several columns. These
findings have been used as an instrumentation and monitoring framework for several other
embankments supported on rigid inclusions in the Minneapolis/St. Paul metro area.
INTRODUCTION
Within the last decade, expansion of roadways and interstate highways within the
Minneapolis/St. Paul metro area has led to increasing use of column-supported embankments
(CSE) with the rigid inclusion ground improvement technique. Embankments successfully built
with this technique need to consider conditions related to rapid project timeline, overall project
budget, environmental issues, and specific geotechnical conditions. The fundamental design
concept of CSE is to utilize arrays of stiff vertical columns to transfer embankment loads through
underlying compressible soil to a firm foundation (FHWA 2016). In very soft soil, the columns
are designed to collectively bear as high as 100% of the embankment load, whereas in stiffer
soil, embankment load is partially shed to the surrounding soil. As a result, because of the load
transfer mechanism, design of CSE must account for various potential failure modes, including
column group capacity, column group lateral extent, vertical load shedding, lateral sliding, global
stability, and foundation settlement (BS8006 2010, FHWA 2016).
In this paper, the drilled full displacement type of rigid inclusion columns (hereafter, referred
to as “columns”) are discussed. These columns are installed by a system described as “augered,
pressure-grouted displacement pile,” around which the soil is displaced horizontally and/or
upward as the drilling tool advances into the ground. While the drilling tool is extracted upward,
concrete grout is pumped into the soil, forming the column. When new embankment fills are
placed, most of the load is supported by the columns. However, some load is typically
© ASCE
Geo-Congress 2020
transferred to the soil, causing settlement and lateral deformation. The overall soil-structural
response is a combination of both embankment and deep foundation systems. There are two
primary mechanisms for load transfer into the CSE system. The first occurs above the top of the
column in the compacted aggregate layer known as the load transfer platform (LTP). This is
often described as “arching” based on the arch shape of the principal stress arrows in the LTP
pointing towards the tops of the columns. The second mechanism occurs below the top of the
column and is commonly described in pile designs as “downdrag.” As the soil consolidates or
reconsolidates due to disturbance, negative (downward) skin friction arises within the upper
portion of the columns when the surrounding soil settles more relative to the columns. Such
negative skin friction increases the compressive stress in the column and reduces its available
axial capacity.
To understand concepts of various failure modes as well as short- and long-term load transfer
mechanisms, geotechnical instrumentation and monitoring (GIM) programs which monitor the
load and deformation response of the structure are tremendously beneficial. Such instrumentation
programs, however, are not standardized but rather vary by state agencies or are tailored to
specific needs. Moreover, most of the published literature provide only the monitoring results
from GIM programs to verify design assumptions and numerical simulations, or vice versa.
Actual field measurements and details of the instrumentation such as planning and execution,
pros and cons of different transducer types, and challenges during installations are rarely
discussed. In this paper, a literature review of existing practice in GIM programs is first
provided, followed by discussion of a local case study and potential problems. Finally,
instrumentation experience and suggestions are given based on the authors’ experience.
STATE-OF-THE-PRACTICE: GEOTECHNICAL INSTRUMENTATION
The authors compiled published case studies strictly on drilled full displacement rigid
inclusion columns exclusively for transportation projects in the United States (note: research type
projects like test sites for studying load transfer mechanism are not included.) For most
geotechnical applications, load (pressure or stress) and deformation are the two most viable
quantitative measurements that engineers are keen to measure. Therefore, in Table 1, reported
sensor types and related measurement (load vs. deformation response) are highlighted. Not
surprisingly, there are few publications on CSE with drilled displacement type rigid inclusions.
For this reason, the authors also included several projects from other countries for reference.
As can be seen, depending on the interests of the owners/engineers, various types of sensors
have been used to monitor the performance of the CSE. First, settlement and lateral deformation
are consistently mentioned in almost all of these projects. It appears that deformation
measurements are greatly emphasized compared to load measurements. For instance, at a
minimum, settlement plates were surveyed for embankment settlement. Other settlement sensors
such as hydrostatic profile gage, horizontal ShapeArray (SAA), and extensometers were placed
on top of several columns or the soil between the columns. Inclinometers or vertical SAAs are
also listed. Lateral deformation monitoring, whether during construction or post-construction,
can be a helpful indicator for embankment overall stability and lateral spread of soil relative to
adjacent structures. On the other hand, pressure cells, load cells, and piezometers have all been
used in monitoring load, total stress, and porewater pressures, respectively. In particular, both
load and pressure cells are favorably used on top of column, soils, or both. Six projects show that
vibrating wire piezometers have also been placed in soil to monitor porewater pressure for
assessing consolidation. Considering that load transfer is the primary mechanism of rigid
© ASCE
Geo-Congress 2020
inclusions, measurements of load or pressure on the column or soil are useful in analyzing
vertical load shedding and soil arching. It is interesting to see a number of geosynthetic-
reinforced CSEs were instrumented with strain gages on the reinforcement layer of LTP, whereas
only the projects in New Jersey, New York, and Minnesota specified embedded strain gages
within the rigid inclusion columns. These embedded strain gages were intended to record axial
compressive strain of the column to provide an axial load distribution profile.
In summary, the current practice of GIM programs for CSE with drilled displacement rigid
inclusions appears to focus more on deformation rather than load monitoring whether on
individual columns or specific areas of interest. Monitoring the load response should be
considered since load transfer is one of the more important mechanisms for determining
performance of CSE. In the authors’ opinion, deformation monitoring is greatly favorable at first
glance because deformation performance criteria (e.g. settlement after n-year ≤ 1 inch) is more
conveniently established in the construction or warranty specifications, while load (pressure or
stress) is relatively difficult to physically quantify and less reliable to measure. Heavy reliance
on deformation response, however, can be precarious and become the engineer’s “worst
nightmare” if load response is overlooked. That is, deformation being the product of stress (i.e. ε
= σ/E for a rigid inclusion column, primary consolidation of soil) is an indication of load transfer
efficacy, hence, confirming the design and anticipated behavior of the CSE. If load transfer
performance of actual CSE is undermined or even worse, unbeknownst to the engineers, the
reported deformation does not necessarily reveal correct information about the geotechnical
response (i.e. degree of primary consolidation) of the underlying soil (i.e. consolidation is a
function of vertical effective stress and dissipation of excess porewater pressure), then there may
be inaccurate assumptions propagated into future designs as a result. The dichotomy between
measurement of deformation and stress also appears analogously in landslide and slope failure.
Slope failure (deformation) occurs because mobilized shear stress exceeds available shear
strength, driving down factor of safety to less than one. Although deformation criteria (e.g.
settlement ≤ 1 inch) is a necessary performance parameter, ignoring load response may provide
an incomplete picture of system performance with respect to the stiff column structure itself.
LOCAL EXPERIENCE IN MINNESOTA
Project Background and Geotechnical Specifications
The I-35E Cayuga project reconstructed 1.4 miles of the corridor in 2013 to 2016 in a highly
dense, urban area from University Avenue to Maryland Avenue, just north of downtown St. Paul,
Minnesota. The project added an interchange, improved access and interchange geometrics, as
well as included the new construction, reconstruction, and rehabilitation of 13 bridges. Design
and staging of the project shifted the alignment of I-35E eastward by approximately 300 feet in
order to construct much of it “offline” (i.e. adjacent to the existing, still-operational roadway)
thereby minimizing traffic impacts during construction, and to smooth traffic transitions between
the project area and the HOV MnPASS lanes to the north. The new alignment crossed an area
with up to about 20 feet of buried organic soils that had been compressed due to previous
earthwork in the well-developed area. The overlying fill was typically dense sand and the
organic soils were slightly overconsolidated but still had the potential for excessive long-term
settlement.
The use of geotechnical tools became paramount in achieving the goals of the project,
especially in terms of protecting existing infrastructure and meeting the project’s rapid schedule.
© ASCE
Geo-Congress 2020
The use of CSE is one such geotechnical solution that served both purposes equally. Design
criteria for the CSE targeted deflections and deformations in two main criteria: (i) The
“Contract” settlement/deformation requirement(s) of nearby proposed structures and (ii) the risks
associated with the unknowns from sensitive in-place utilities, buildings and road infrastructure,
and the aggressive construction schedule.
Table 1. Summary of geotechnical instrumentation and sensor types for drilled
displacement rigid inclusions. Case studies selected for transportation projects only.
Lateral
Case Study Load/Press. Settlement PWP Misc.
Deform.
I-29, Council Bluffs, IA PPC (soil) MPBX INC (MEMS) PZ SG (load test
(Gallant et al. 2018, SP only)
IowaDOT 2014)
Garden State Parkway, NJ EPC (column) Horiz. SAA INC - SG (column)
(Ramakrishna et al. 2013, SMS SG (geosynt.
Masse et al. 2013) SP LTP)
SG (MSE wall)
I-690 Bridge over Teall Ave, EPC (soil) Horiz. SAA Vert. SAA PZ SG (column)
Syracuse, NY MPBX TL
I-35E Cayuga, St. Paul, MN LC (column) Horiz. SAA Vert. SAA PZ SG (column)
(Potter-Weight et al. 2016, EPC (soil)
Chow et al. 2020) EPC (column)
TH 169 Nine Mile Creek, EPC (soil) Horiz. SAA Vert. SAA PZ SG (column)
Edina, MN EPC (column) SG (MSE wall)
I-35E Flyover Bridge, EPC (soil) Horiz. SAA Vert. SAA PZ SG (column)
Forest Lake, MN EPC (column) SP
North Dynon Embankment, EPC (soil) Horiz. INC INC PZ SG (geogrid
Melbourne, Australia TL LTP)
(King et al. 2017)
Case Studies A & B LC (column) HPG INC - -
Motorways, Australia LC (soil) SP
(Larisch et al. 2015)
Forth Replacement Crossing, - SP - - -
Scotland (Mathieu et al.
2015)
Breakwater Road Bridge, EPC (soil) DSM - - -
Victoria, Australia
(Fok et al. 2012)
Notations: PWP = Porewater pressure, LC = load cell, PPC = push-in pressure cell, EPC = earth pressure cell, SP =
settlement plates, LTP = load transfer platform, SAA = ShapeArray, INC = inclinometer, TL = tiltmeter, HPG =
hydrostatic profile gage, PZ = piezometer, MPBX = multi-point borehole extensometer, DSM = differential
settlement monitoring gage, SMS = settlement monitoring system, SG = strain gage (arc-weldable, sister bar, foil
bonded, etc.)
The CSE and associated LTP were designed to meet these project criteria, including risks due
to changes in construction rate and staging as a result of changed conditions encountered in the
field. The end result was a “Method Specification” that detailed all aspects of CSE & LTP
© ASCE
Geo-Congress 2020
construction methods, materials, reinforcement, and depths, QA/QC testing and acceptance, load
testing and monitoring, and a robust instrumentation program to monitor the performance of the
system during construction and in the long-term after the project was completed.
Field Instrumentation and Monitoring Results
As part of the project scope, five test areas or “nests” were designated for the GIM program
that included vibrating wire piezometers, load cells, earth pressure cells, sister bar strain gages,
horizontal- and vertical SAAs. For brevity, the results from only one instrumentation “nest” (i.e.
Nest D) are presented. This nest of instrumentation consists of two groups of three columns, with
one group instrumented as described in (i) and the other instrumented per (ii) below:
(i) Load cell and four embedded sister bar strain gages: Four strain gages (SG) (a.k.a.
rebar strainmeters or sister bar gages) were placed at selected intervals, as depicted on
Figure 3, and tied to a rebar with centralizers, then pushed into the center of the column
after drilling and grouting were complete. Following that, a 1-inch (25.4 mm) steel
bearing plate was placed, centered, and leveled on top of the column before load cell
(LC) placement. A second steel plate was then placed above the load cell, sandwiching
the load cell between the two steel plates. To facilitate installation, the lower plate was 15
inches (381 mm) in diameter compared to the nominal 16-inch (406 mm) diameter
column and top plate. With the use of a threaded steel rod, bolts, and nuts, the load cell
unit was then slightly pre-stressed (e.g. 500 lbs) to the linear range of measurement. The
interior between the two plates was filled with spray foam insulation to prevent soil from
potentially migrating into the void space.
(ii) High-range earth pressure cell placed on top of column: As soon as concrete grout was
stiff enough after column installation, a thick-plate high-range (1,100 psi or 7.5 MPa)
pressure cell (HEPC) was placed on top of the selected column. First, the concrete grout
surface was smoothed using a hand float; the pressure cell was then pressed into the grout
and checked for level. This procedure sometimes took several attempts until the cell was
leveled and properly seated. *Note: Readers shall not be confused by using “earth
pressure cell” on concrete grout for measuring earth pressure. In this context, HEPC are
specially manufactured to address factors affecting measurements (refer to Section
10.2.3, Dunnicliff 1988) and intended for installation on concrete grout.
In addition to the above transducers, nine low-range (50 psi or 350 kPa) earth pressure cells
(LEPC) were also placed on soil in between the six columns. A borehole was drilled at the center
of four columns to install a vibrating wire piezometer (PZ) within the compressible soils. These
sensors were installed according to manufacturer’s recommendation. For all sensors, baseline
measurements were collected periodically both before and after installations with cable routing
through protective conduit to avoid straining the cables or damaging to the transducers or the
datalogging system. All sensors were protected against lightning through surge protectors and
electronic components were grounded. Figure 1 shows the plan view of the instrumentation
layout and graphical presentation of the load cell configuration. Figure 1a also shows the relative
layout of the columns #571, 572, 596, 597, 621, and 622.
Figure 2 presents monitoring results from June 2015 until the end of 2018. In the top figure,
the load cell measurements refer to change in load relative to initial fill placement, indicating the
maximum loads were consistently above 30 kips (133 kN) at columns #597 and #622, while only
about one-third of load was measured at #572 (this makes sense given that #572 is below the
front “toe” of the wall’s spread footing, with little overburden directly above.) With fill height of
© ASCE
Geo-Congress 2020
about 19 feet (5.8 m) behind the wall, the overburden pressure was expected to be 2.3 ksf (109
kPa). Yet, the computed pressures on the columns based on the maximum loads are three to ten
times of that, i.e. 7.4 and 22.3 ksf (354 and 1,068 kPa), respectively, indicating the evidence of
arching of stresses to the stiffer columns. The middle figure shows the strain measurements
within column #597. Clearly, compressive strain began to accumulate as soon as new fill was
placed. It is interesting that the third level of strain gage (D-SG-3-597) exhibited highest
magnitude of compression. This was mentioned in Potter-Weight et al. (2016), but further
discussion is provided later.
Figure 1. Plan view of instrumentation layout and load cell configuration

In Figure 2c, the pressures measured on three columns #571, #596 and #621 and in between
the columns are presented. D-HEPC-571, which aligns with column #572, shows an average
measurement of 4 ksf (192 kPa), while the other two HEPCs on columns #596 and #621
measured pressures averaging 22 and 24 ksf (1,053 and 1,149 kPa), respectively. These HEPC
measurements are consistently similar to the adjacent columns for which load was measured with
a load cell. In contrast, all LEPC measurements, on average, were in the range of 0.7 to 2.6 ksf
(34 to 125 kPa), significantly lower than HEPCs. These pressure differences clearly indicate that
overburden load shed more unto the columns than surrounding soils. Lastly, negative porewater
pressure is seen when both LC and HEPC measurements dramatically increased, then dissipated
relatively fast in late December 2015. The negative and fast dissipation rate of porewater
pressure response are a stiff clay behavior upon loading.
Strain Measurements and Negative Skin Friction
Analysis of above monitoring results, including load transfer efficacy and numerical model
verification etc., is not within the scope of this paper. Monitoring data from other nests
containing similar configurations with an addition of vertical and lateral deformation monitoring
are expected to be useful to future studies on the topic. Some discussions about the design and
comparison of partial monitoring result can be found in Potter-Weight et al. (2016). Herein, the
authors would like to emphasize the use of strain and load cell measurements from instrumented
columns #572, #597 and #622 as an example for presenting soil-structural interaction (e.g.
© ASCE
Geo-Congress 2020
negative skin friction) of the drilled full displacement type rigid inclusions. Detailed subsoil
profile, concept, and interpretations can be found in a companion paper by Chow et al. (2020).
As mentioned earlier, load transfer mechanism in CSE consists of arching to the tops of
individual columns and to the surrounding soil itself. As a result, consolidation of soil leads to
the development of negative skin friction along the upper portion of the columns. The negative
(downward) skin friction is in turn counterbalanced by equal and opposite (upward) skin friction.
When both positive and negative skin frictions mobilize, the neutral plane location should be
readily identified by strain measurements and depth profile. A product of negative skin friction is
called “drag load,” which if unaccounted for, can be excessively high and detrimental to the
column structural integrity (i.e. the failure mode of “column group capacity.”)
Figure 2. Nest D Monitoring results: (a) load cell measurements at three instrumented
columns; (b) example strain gage measurements within rigid inclusion of column #597;
and, (c) pressure cell and piezometer measurements.
Shown in Figure 3, strain measurements within the instrumented columns are converted to
loads by using column modulus, creating axial load distribution with depths on different dates. A
simplified subsoil profile, including soil type and strength, are also given. As simplifying
assumptions, the column area was taken to be uniform along its length and a tangent modulus
was derived from a nearby static load test. The actual moduli of individual columns can be
different, which presents some inherent uncertainty to the calculated loads in Figure 3. Each
symbol represents a strain gage level, except with the uppermost symbol being the load cell
measurements on top of the column. The load distribution curves indicate that the maximum
axial load generally varied with time but tended to remain at the same depth. The actual shape of
© ASCE
Geo-Congress 2020
load distribution curves just below the maximum point of the curve is not known unless
additional strain gage would have been installed below. After about late December 2015 (also
see Figure 2), the rate of development of skin friction sharply decreased, converging to the final
“D-shape” curve. The maximum final load (Qn) was around 320 kips (1,423 kN) at #622. This
translates to the maximum axial compressive stress of 1,650 psi (11.4 MPa). That value is 66%
of the minimum design grout 28-day strength of 2,500 psi (17.2 MPa) but is about 48% of the
3,440 psi (23.7 MPa) average grout strength from the load test columns. The drag load after
subtracting measured top load was about 285 kips (1,268 kN).
This example demonstrates one of the useful and important applications in load monitoring
of rigid inclusions by using strain gages. Without strain measurements, these load distribution
profiles and loads in columns would not be possible. Load cells only measure the load applied on
top, while loads in the column are best measured by embedded strain gages. Unlike driven steel
piles and prestressed concrete piles, determination of column modulus remains challenging when
converting strain to load. In this example, the authors assume column modulus is similar to
modulus of a nearby instrumented load-test column. The role of strain measurements within
columns, for this reason, becomes more prominent for drilled displacement type rigid inclusions.
Figure 3. Measured load distribution along individual columns (After Chow et al. 2020).
RECOMMENDATIONS FOR INSTRUMENTATION
All geotechnical instrumentation planning should start by asking questions: (i) general role of
geotechnical instrumentation, (ii) principal geotechnical questions that require instruments to
help answer, and (iii) overview of typical routine and special monitoring. Readers are
encouraged to refer to Dunnicliff (1988) and FHWA (1998) for planning a GIM program. In the
authors’ opinion, questions for CSE with rigid inclusions include: “which failure mode is of
concern?”, “what is the anticipated load transfer efficacy or soil arching?”, “will there be
significant downdrag?” and possibly questions related to the LTP. Following that, the GIM
program should be designed such that measurements are helpful in answering these questions. In
terms of instrumentation applications, the authors suggest the following drawn from the I-35E
Cayuga project CSE experiences:
(i) Deformation and load monitoring apparatus should be used altogether: For sites
with stiff soils immediately below the LTP (like the dense sand at I-35E Cayuga project),
high arching to soil is anticipated. The use of load monitoring transducers gives an
insight of load transfer efficacy, hence, consolidation of soil. At a minimum, simple
© ASCE
Geo-Congress 2020
settlement monitoring devices (e.g. settlement plates) and embedded strain gages (e.g.
rebar strainmeters) shall be utilized. Piezometers are always helpful in assessing excess
pore pressure and subsequent consolidation progress.
(ii) Use of thick-plate earth pressure cells as alternative to load cell: For practical
purposes, the earth pressure cell placed on top of a column must be stiff relative to the
grout so that column deformation is resisted by the pressure cell and its strain is
insensitive to the grout modulus. The success with thick-plate high-range earth pressure
cells on I-35E Cayuga project demonstrates that the pressure cell measurements are quite
agreeable with that of load cells. Load cells are an excellent transducer at measuring
change of load, but are less convenient and cumbersome to install. Each load cell also
typically contains four to six strain gages, all of which must be logged. On the other
hand, earth pressure cells are convenient to handle and cheaper (one transducer per cell).
Nonetheless, careful placement and workmanship during installation of pressure cells
often dictate the quality of measurements. Practicing geotechnical engineers and
instrumentation specialists should be aware of those factors affecting earth pressure cell
measurements (refer to Section 10.2.3, Dunnicliff 1988.)
(iii)Use embedded strain gages if expected downdrag is significant: This can be done by
attaching strain gages to a rebar with appropriate centralizers at various intervals, then
pushing the rebar into the column after drilling is complete. In order to capture the load
distribution profile, strain gages should be placed near the top of column, at top and
bottom of compressible layer(s), and close to the column toe. In the authors’ experience,
care must be exercised when pushing the instrumented rebar into the grouted column. If
not done carefully, the instrumented rebar can be forced out of grouted column and into
the soil, thereby giving erroneous measurements. Plasticizer or water reducing admixture
can be added into grout to increase workability of concrete and facilitate strain gage
installation. In fact, this is often done for non-production, load-test columns to install
embedded strain gages to measure the load distribution profile during static load testing
of columns.
(iv) When needed, monitor lateral deformation: Monitoring of lateral deformation is often
associated with the concerns for overall stability and excessive deflection of structures
such as retaining walls. In addition, lateral deformation can cause elevated bending
moments and shear forces in perimeter columns which could negatively impact the
performance of the CSE. Fortunately, inclinometer and vertical SAA are relatively easy
to install and salvageable for other applications.
CONCLUDING REMARKS
It has been shown that practices in geotechnical instrumentation and monitoring for column-
supported embankments supported on drilled displacement type rigid inclusions is highly varied
across state agencies and even nationally. The interests of instrumentation on these projects, of
course, are different. However, understanding the mechanism of load transfer and potential
failure modes can be helpful for the instrumentation planning process. The case study of the I-
35E Cayuga project is an example in demonstrating the importance of negative skin friction for
drilled full displacement type rigid inclusions. With proper training and experience, the use of
earth pressure cells and embedded strain gages are economical and resourceful tools for
answering questions regarding load transfer efficacy and downdrag.
© ASCE
Geo-Congress 2020
ACKNOWLEDGEMENT
The authors would like to acknowledge Minnesota Department of Transportation, Shafer
Contracting Co., Menard Group USA, and American Engineering Testing, Inc. for providing
resources and assistance in support of this paper. The contents of this article do not necessary
reflect the official views or policies of MnDOT.
REFERENCES
BS8006. (2010). Code of Practice for Strengthened/Reinforced Soils and Other Fills. British
Standards Institution, London, U.K.
Chow, L.C., Han, J., Reuter, G.R. (2020). “Field monitoring of negative skin friction on rigid
inclusion columns under embankments.” Proc. in Geo-Congress 2020, ASCE, GSP,
Minneapolis, MN. (Manuscript under review.)
Dunnicliff, J. (1988). Geotechnical instrumentation for monitoring field performance. John
Wiley & Sons, New York.
Federal Highway Administration. (1988). Geotechnical Instrumentation Reference Manual,
FHWA-HI-98-034.
Federal Highway Administration. (2016). Geotechnical Engineering Circular No. 13 Ground
Improvement Modification Methods – Reference Manual Volume II, FHWA-NHI-16-028.
Fok, N., Qiu, T., Vincent, P., and Kreminsky, M. (2012). “A case study of ground improvement
using semi-rigid inclusions for Breakwater Road Bridge.” Proc. of the International
Conference on Ground Improvement and Ground Control, 629-643.
Gallant, A.P., Shatnawi, E., Farouz, E., and Jones, T. (2018). “A case study of settlement and
load transfer at depth beneath column-supported embankment.” Proc. Intl. Fds. Congress
and Equipment Exposition, ASCE, GSP, Orlando, FL.
Iowa Department of Transportation. (2014). Special Provisions for Instrumentation. SP-
120228a.
King, D.J., Bouazza, A.B., Gniel, J.R., Rowe, K., and Bui, H.H. (2018). “Geosynthetic
reinforced column supported embankments and the role of ground improvement installation
effects.” Can. Geotech. J., 55: 792-809.
King, D.J. (2017). “The behavior and performance of geosynthetic reinforced column supported
embankments.” Ph.D. thesis, Monash University, Melbourne, Australia.
Larisch, M.D., Kelly, R. and Muttuvel, T. (2015). “Improvement of soft soil formations by
drilled displacement columns.” Ground Improvement Case Histories, 573-622.
Masse, F., Potter-Weight, A., Aziz, S., Sankey, J., Rafalko, S., Walker, M., and Nodide, M.
(2013). “Full scale load test program confirms the design of a rigid inclusions solution for the
support of an MSE wall in southern New Jersey.” 38th Annual Conference on Deep
Foundations, DFI, Phoenix, AZ, Sept 25-28, 2013.
Mathieu, F., Racinais, J., Rochault, J., and Adams, D. (2015). “Successful combination of rigid
inclusions (CMC) and Trenchmix (TM) soil improvement solutions.” Proc. of XVI ECSMGE,
Geotechnical Engineering for Infrastructure and Development.
Potter-Weight, A., Darnell, S., and Sorabella Swift, S. (2016). “Column supported embankment
(CSE) for I-35E Cayuga Bridge.” 64th University of Minnesota Geotechnical Engineering
Conference, ASCE, St. Paul, MN.
© ASCE
Geo-Congress 2020
Ramakrishna, A., Trimpin, E., and Mankbadi, R. (2013). “Embankment construction using
column supported embankment.” 7th International Conference on Case Histories in
Geotechnical Engineering, Apr 29 – May 4, 2013, Chicago, IL, 16.
© ASCE
Geo-Congress 2020
Detecting Pile Length of Sign Structures and High Mast Poles

Daniel V. Kennedy1; Bojan B. Guzina2; and Joseph F. Labuz3
1
Ph.D. Student, Dept. of Civil, Environmental, and Geo-Engineering, Univ. of Minnesota,
Minneapolis, MN. E-mail: kenne407@umn.edu
2
Shimizu Professor, Dept. of Civil, Environmental, and Geo-Engineering, Univ. of Minnesota,
Minneapolis, MN. E-mail: guzin001@umn.edu
3
MSES/Kersten Professor and Head, Dept. of Civil, Environmental, and Geo-Engineering, Univ.
of Minnesota, Minneapolis, MN. E-mail: jlabuz@umn.edu
ABSTRACT
Several hundred high mast light towers (HMLTs) throughout the state of Minnesota have
foundation systems, that are typically concrete-filled steel pipe piles or steel H-piles, with no
construction documentation (e.g. pile lengths) and soil stratigraphy information. Reviews of
designs within current standards suggest that many of these foundations may have insufficient
uplift capacity in the event of peak wind loads. Without knowledge of the in situ pile length, an
expensive retrofit or replacement program would need to be conducted. Thus, providing a
screening tool to determine in situ pile length—as compared to a bulk retrofit of all towers with
unknown foundations—would provide significant cost savings. The goal of the project is to
establish a non-destructive field testing technique, including data analysis algorithm, for
determining in-place pile lengths by way of seismic waves. A unique feature of the proposed
work is the use of computational modeling to account for the effects of soil profile and ground
conditions (e.g. moisture) on the sensitivity of the method. The length of each pile supporting an
HMLT will be identified through a systematic sensing approach that includes (i) collection and
classification of the pertinent foundation designs and soil conditions; (ii) three-dimensional (3D)
simulation of dynamic soil-foundation interaction; (iii) parametric studies of the 3D pile
vibration problem; (iv) field testing; and (v) analysis-driven data interpretation.
INTRODUCTION
High activity areas including freeways, intersections, and airports require additional safety
provisions concerning visibility. Typically, these locations also need to use the available land
area in the most efficient manner. To achieve this, HMLTs are capable of illuminating large
surface areas while maintaining a small footprint.
An HMLT is a tall, thin, tapering steel pole with downward facing lights attached at the top
(see Fig. 1). Typically, HMLTs are designed to reach heights from 30.48 m (100 feet) to over
42.67 m (140 feet), and work in unison with other towers or light sources to achieve the
necessary level of illumination.
With reference to Fig. 2, HMLT foundation systems typically consist of three inclined
concrete-filled steel pipe piles or steel H-piles that connect to the light tower via a triangular
concrete pile cap. Because the tower behaves structurally as a tall cantilever, its foundation
system needs to provide sufficient resistance to overturning moments due to peak wind load in
addition to necessary bearing capacity. Overturning moments and forces produced by wind load
are resisted by both the bearing and uplift capacity of the embedded piles, which increase with
the piles’ embedment length.
Table 1 shows that nearly half of the HMLT foundations in Minnesota (MN) lack pile
© ASCE
Geo-Congress 2020
embedment length information. Upon review of the existing HMLT designs through the prism of
new Load and Resistance Factor Design (LRFD) structural design criteria, it was found that
many HMLT foundations in MN may: 1) have insufficient uplift capacity in the event of peak
wind loads, and/or 2) not receive recertification under new design criteria due to missing pile
depth information.
Figure 1: High mast light towers.

ESTIMATE OF BENEFITS TO MINNESOTA TAX PAYERS
According to current estimates, there are over 200 HMLTs in Minnesota with no construction
records indicating the tower’s foundation pile length. This represents close to 50% of the HMLT
towers in the state. Many of these HMLTs also have no on-site soil information as well. Table 1
lists the available HMLT data for the State of Minnesota.
Numerical simulations suggest that during peak wind loading, the HMLT foundation may not
have sufficient uplift capacity if the length of foundation piles is less than 10 m (32.81 ft). The
soil information is also lacking for some HMLT foundations, implying further that in situ soil-
structure interaction may not be adequately described by preliminary numerical simulations.
The cost estimate to retrofit or replace an individual HMLT foundation, with unknown pile
length, exceeds $20,000 (Dasenbrock, 2017). Table 2 indicates that if none of the tested HMLTs
require additional work on their foundation systems, the benefits to MN taxpayers may exceed
$4 million. On the other hand, if 75% of the HMLTs tested have insufficient pile length, and
require additional work, estimated savings to the MN taxpayers could still exceed $1 million.
© ASCE
Geo-Congress 2020
Transportation (MnDOT) show three typical HMLT foundation designs used in Minnesota: T-
100, T-120, and T-140. The inclined (7.125°) pilings are constructed using steel 12” ø C.I.P. or
steel HP10x42 piles. The depth of the piling is site-specific and determined for individual
foundations (based on the local soil and loading conditions). Figure 2 and Table 3 show the key
design differences between the three types of tower bases.
Table 3: Foundation geometry values for typical designs.
Type L B T D Tower Height
T-100 8'-6" 1'-6" 2'-6" 5'-0"± 100'

T-120 11'-6" 2'-0" 3'-6" 6'-0"± 120'
T-140 11'-6" 2'-0" 3'-6" 6'-0"± 140'
HMLT FOUNDATIONS FOR RESEARCH PURPOSES

Figure 3 shows an aerial photograph of the testing site with existing HMLTs that has been
identified for the research project. The site is located in Eagan, MN, at the intersection of MN 77
(Cedar Ave.) and MN 13 (Sibley Memorial Hwy.).
The available geotechnical report for the site recommends steel HP10x42 piles for the new
construction of tower foundations. No information is provided for the type or depth of in situ
foundation piling. The three towers identified for the project (based on tower type, location,
accessibility, and safety) are circled and are labeled as: T2E T10, T2C T2, and T2C T1.
Figure 3: Tower locations for research project.

All three towers are located sufficiently far enough from the ongoing traffic to provide a
satisfactory level of safety for the study (T2C T1 is located behind a guardrail as well). All three
locations are accessible by the Seismic Cone Penetrometer Test (SCPT) truck.
Towers T2C T1 and T2E T10 are of type T-100, while tower T2C T2 is of type T-120; see
Table 3 for the respective tower foundation geometries.
SOIL PROPERTIES FOR ELASTODYNAMIC MODELING
From SCPT soundings at the testing site, three distinct soil types dominate the area near the
piles: gravel to stiff sand, uniform sand, and clay/silt mix. The three soil types have different
layer distributions in each of the SCPT profiles encountered at the testing site. Representative
elastic properties of each soil type have been identified for the purpose of elastodynamic
modeling. Tables 4 and 5 show the elastic properties chosen (Poisson’s ratio v and Young’s
© ASCE
Geo-Congress 2020
modulus E) along with the constrained and shear moduli, mass density, and the compressional
(P) and shear (S) wave velocities for the three soil profiles.
M E 1   G E
VP  , M : Vs  , G
 1  1  2   2 1  
Table 4: Soil elastic properties.

Elastic
properties v Young’s modulus, E Constrained modulus, M Shear modulus, G

Soil Type [MPa] [psi] [MPa] [psi] [MPa] [psi]
Gravel to
Stiff Sand 0.3 80.0 11603.0 107.7 15619.4 30.8 4462.7
Uniform
Sand 0.3 40.0 5801.5 53.8 7809.7 15.4 2231.4
Clay/Silt
mix 0.3 20.0 2900.8 26.9 3904.9 7.7 1115.7
Table 5: Soil density and wave velocity.

Density and wave speed Density, ρ P wave speed, Vp S wave speed, Vs
3
Soil Type [kg/m ] [pcf] [m/s] [ft/s] [m/s] [ft/s]
Gravel to Stiff Sand 1842.1 115.0 241.8 793.3 129.2 424.0
Uniform Sand 1842.1 115.0 171.0 560.9 91.4 299.8
Clay/Silt mix 1922.2 120.0 118.3 388.3 63.3 207.5
Figure 4: Schematics of the pile sensing approach.

PILE SENSING
With reference to Fig. 4, the proposed Non-Destructive Evaluation (NDE) approach revolves
around mechanical excitation of the pile cap above (or near) the pile stem and monitoring the
near-field ground vibration by a geophone mounted on a seismic SCPT device. During the test,
the pile cap would be continuously excited by a steady-state vibration source, while measuring
the ground vibration at several depths along a vertical line in a close proximity to the pile tested.
© ASCE
Geo-Congress 2020
The recorded seismic data would then be processed and compared to numerical simulations to
determine the pile length. Note that the use of steady-state vibrations, as opposed to deploying an
impact response, is selected for their robust performance in complex/noisy environments (Pak
and Guzina, 1995).
The essence of the proposed approach is the fact that each pile serves, thanks to a high
impedance contrast between the pile and soil, (1) as a waveguide, which facilitates the
transmission of mechanical vibrations all the way to the bottom of the pile, and (2) as a
mechanical antenna, where the wave energy propagating up-and-down the pile is partially
radiated into the surrounding soil. This latter complement of the vibration energy is then picked
up by a geophone mounted near the tip of SCPT and provides the data from which the pile length
is to be estimated.
SHAKERS
The key parameters considered in the selection of an optimal shaker for field testing are: 1)
unidirectional force – force acting in the vertical direction; 2) frequency range of up to 100 Hz or
higher; and 3) peak force amplitude. In addition, a baseplate mounting connector is needed to
efficiently transfer the energy from the shaker to the steel rod that is in contact with the top of the
pile cap. Piston vibratory devices meet the requirement of a unidirectional force by constraining
the moving parts to the axial direction only (as opposed to circulating vibratory devices). Two
piston-type shakers (see Fig. 5) have been selected for the research project. The two devices
(Dayton 1DYR2 and Vibco 4HT77) offer different frequency and force ratings, as described in
Table 6. Clearly, there is a tradeoff between frequency and force rating for the two models.
Table 6: Vibratory device specifications.
Piston Vibrator Frequency Range [Hz] Force Rating [lbf/N] Air Flow [CFM]
Dayton 1DYR2 35 - 50 1323/5885 10.6
Vibco 4HT77 90 - 115 325/1446 9.0
Figure 5: Shakers and baseplate; left = Vibco 4HT77, middle = Dayton 1DYR2, right =
baseplate.
An interchangeable baseplate mount was designed and fabricated for the two shakers, to
allow for quick and convenient shaker interchange in the field. Two screws attach the shaker
device to the top of the baseplate. The baseplate mount also provides positive connection
between the shaker device and the steel rod, to mitigate any anticipated bouncing behavior
© ASCE
Geo-Congress 2020
between the top of the steel rod and the bottom of the vibratory device. The bottom cylinder of
the baseplate is threaded, to screw on to the top of the threaded rod that will be in direct contact
with the pile cap. Figure 5 shows the two shaker devices and the baseplate mounted on the
threaded rod.
ELASTODYNAMIC MODELING
An open source Finite Element Modeling (FEM) software, Code_Aster, was chosen to
simulate and analyze the ground motion involved in detecting the in situ length of pile
foundations. Code_Aster was mainly developed by the R&D group of Électricité de France
(EDF), a large electric utility provider controlled by the French government. The software
provides a platform for numerical simulations over a wide range of applications including
thermal, mechanical, and acoustic problems.
Code_Aster is further bundled into a broader open source software package, Salome_Meca,
that provides a Graphical User Interface (GUI) allowing users to perform necessary pre- and
post-processing through well-defined modules in a central software location. The pre-processing
step includes the Geometry and Mesh modules, and the final post-processing step uses the
Paravis module to visualize the computed fields.
Figure 6: Tower foundation design and installation information provided.
Figure 7: T-100 foundation model and finite element mesh.

During the pre-processing stage, the Geometry and Mesh modules are used in Salome Meca.
As implied by their naming, the Geometry module uses Computer-Aided Design (CAD) to create
geometrical representations of the model, and the Mesh module turns these representations into
1D, 2D, or 3D groups of interconnected nodes, faces, and volumes ready for analysis. Figure 6
© ASCE
Geo-Congress 2020
illustrates the given designs and tower foundation information that were used to create
representative models in the Geometry and Mesh modules.
Figure 7 illustrates the modeled foundation geometry and the finite element mesh along with
observation points in the soil. The small circle shown beneath the y-axis line is the area of
excitation for the modeling.
The featured time-harmonic FEM analysis solves for the steady state visco-elastodynamic
response of the soil-pile system due to sinusoidal force acting at a prescribed frequency. For the
analysis, linear tetrahedral elements were chosen with the minimum number of nodes per
wavelength set at 8. Absorbing boundary elements were used on the model’s outer surfaces
inside the soil, to absorb the propagating wave (minimize wave reflection back into model) and
mimic the infinite soil domain.
Figure 8: Results at 50 Hz; left = overall in-phase component of particle velocity in the z-
direction, right = cross section of results.
Figure 9: Padded annulus on semi-infinite half-space; left = analytical setup, right =

numerical model.
To better understand the SCPT data that will be collected, the results from the numerical
simulation can be selectively chosen for further inspection. Figure 8 displays the overall real
component of the particle velocity field generated by a time-harmonic point force (acting on the
pile cap) at 50 Hz, along with a cross section view including a solid black line next to the pile
foundation that will provide data along the line’s path to compare to a SCPT’s testing data during
in-field tests.
© ASCE
Geo-Congress 2020
Figure 11: Surface motion (   10% ); left = magnitude, right = in-phase, left = out-of-
phase.
Figure 12: 50 Hz ground velocity; left = overall real and imaginary components, right =
magnitude along observation path.
Figure 13: 35 Hz velocity; left = real and imaginary components, right = magnitude along
path.
NUMERICAL RESULTS
To obtain estimates of the ground response due to steady-state vibration supplied by each
shaker, FEM simulations were performed at representative frequencies of f1, f2, … Hz. The
magnitude of the visco-elastodynamic response, vx 2  v y 2  vz 2 , is plotted versus depth (see Fig.
12) along a vertical line that traces the SCPT location (observation path is depicted by solid
black lines in overall vector field) offset from the center of the HMLT by 3.5m. The vertical
© ASCE
Geo-Congress 2020
dashed lines in the plots are placed at the depth where the pile terminates, around 6.75 m deep
for this case.
A large amount of ground movement is seen near the surface, likely due to global vibration
of the large pile cap. Because of this, the response near the end of the pile is not easily
understood. Instead, by focusing on the ground velocity below the bottom of the pile cap (see
Fig. 14), the change in velocity is more noticeable. A notable decrease in velocity is observed in
the soil near the end of the pile, giving confidence that the sensing approach is feasible.
For completeness, the ground motion response is simulated assuming multiple force
amplitude / force frequency combinations relevant to the operation of the two shakers. The
following section (Figs. 13 – 16) show two representative force/frequency combinations for each
shaker.
path.
path.
VARYING PILE LENGTH
Different pile lengths were investigated numerically to better understand the effect of pile
embedment depth on sensory data. Figure 17 shows the difference in ground motion along the
SCPT “scan line”, at two frequencies, for pile embedment depths of 4 m, 6.75 m, and 9 m.
Above the pile end depths the ground motion is similar in form and magnitude for the different
pile lengths. In both figures there is a notable decrease in ground velocity, and a deviation from
© ASCE
Geo-Congress 2020
other pile depth responses, at depths below the end of the pile. The numerically generated ground
velocity at different frequencies for different pile length gives confidence that the multiple force
amplitude / force frequency sensing approach is feasible for detecting the in situ pile end depth.
path.
Figure 17: Ground response at different pile end depths; left = 40 Hz, right = 100 Hz.
CONCLUSIONS
This work describes a preliminary study toward estimating the in situ length of piles
supporting the HMLT structures by seismic waves. The FEM platform Code_Aster has been
selected to simulate the problem and compared to available analytical solution as a means to
validate the numerical model. From numerical simulations, a notable decrease versus depth in
the magnitude of the particle velocity vector is observed in the soil near the end of the pile,
giving confidence that the proposed approach that relies on SCPT measurements is feasible. Two
shaker devices, with unique force amplitude / force frequency combinations, allow for a wide
range of test configurations during field observations.
REFERENCES
B.B. Guzina and S. Nintcheu (2001). “Axial vibration of a padded annulus on a semi-infinite
© ASCE
Geo-Congress 2020
viscoelastic medium,” J. Appl. Mech., ASME, 68, 923-928.

D. Dasenbrock (2017). MnDOT, personal communication.
R.Y.S. Pak and B.B. Guzina (1995). “Dynamic characterization of vertically loaded foundations
on granular soils,” J. Geotech. Eng., ASCE, 121, 274-286.
© ASCE
Geo-Congress 2020
Sky Harbor Airport Runway Realignment

Hector D. Flores1 and Brandon J. Twedt2
1
Professional Engineer, Short Elliott Hendrickson Inc., MN
2
Director of UAS Operations Airport Planning and Design, Short Elliott Hendrickson Inc., MN
ABSTRACT
The Sky Harbor Airport Runway in Duluth, Minnesota, is in the process of being realigned to
increase safety with the runway approach. In 2007, the Duluth Airport Authority identified
several obstructions to the Runway 32 approach surface. As a result, the Authority completed a
multi-year federal Environmental Assessment and a state Environmental Assessment Worksheet
to evaluate alternative solutions with respect to the runway approach. The obstructions consisted
of old growth trees, the majority of which were located within a scientific and natural area
(SNA) protected by the Minnesota Department of Natural Resources. The selected solution
consisted of shortening the runway from 3,050 feet to 2,600 feet and rotating it 5 degrees into
Superior Bay. The geotechnical design of the runway considered slope stability and settlement of
the embankment fill. Staged preloading of the site was completed using surcharge fill with a
geotextile reinforcement base layer to improve stability of the fill. In addition to monitoring
settlement with settlement gauges, drone surveys were completed of the surcharge fill
embankment. Underwater surveys, using a multi-beam echo-sounder (MBES) to provide depth
measurements, detailed surface model, and point cloud, were completed to verify the riprap
embankment protection dimensions placed in Superior Bay. This case study presents the
geotechnical investigation and design of the runway relocation along with results of construction
monitoring of fill placement and settlement.
PROJECT DESCRIPTION
Sky Harbor Airport is owned and operated by the Duluth Airport Authority (DAA) and has
been in operation on Minnesota Point since 1939. The airport consists of a paved 3,050 foot
runway and a ramp and dock for seaplane access to Superior Bay. Over time, a number of red
and white pine trees located off the south end of the runway, within airport property and the
Minnesota Point Pine Forest Scientific and Natural Area (SNA), have grown tall enough to be
considered obstructions for approaching aircraft. A majority of the obstructing trees are part of
the old growth forest protected by the SNA. This project will provide a clear approach surface
for aircraft and allow the airport to be in compliance with state and federal rules for issuance of a
Minnesota Airport License. In order to provide an approach surface that eliminates future
obstructions, the DAA will relocate Runway 14/32 into Superior Bay (See Figure 1). This results
in approximately 7.6 acres and 70,000 cubic yards of fill and serves to protect the valuable
resources within the SNA while increasing aviation safety.
The project at the Sky Harbor Airport included three phases of construction over three years.
Currently, the construction of the project is between Phase 2 and Phase 3. Phase 1 included
placement of engineered fill into Superior Bay within the relocated runway grading limits up to
elevation 605 feet and perimeter riprap shoreline protection. Phase 2 included placement of
surcharge material atop the area constructed in Phase 1 to consolidate the underlying material
(See Figure 2). Phase 3 included removal of the surcharge material and construction of the new
runway and parallel taxiway.
© ASCE
Geo-Congress 2020
FIG. 1. Proposed Runway Relocation.
FIG. 2. Typical Cross Section.

Table 1. Summary of Soil Parameters for Settlement Analysis
Unit Weight Initial Void Ratio Compression Index Recompression Index
Soil Description γsat e0 Cc Cr
(pcf)
Lean Clay 112.6 1.13 0.28 0.04
Organic Silt 105.4 1.34 0.36 0.06
SUBSURFACE CONDITIONS
The subsurface investigation for Sky Harbor Airport was performed by EPC Engineering &
Testing, LLC of Duluth, Minnesota. The drilling and laboratory testing program for the
subsurface investigation were developed by SEH geotechnical staff. The subsurface
investigation, performed over two separate mobilizations, consisted of 34 Standard penetration
test (SPT) soil borings and 4 offset soil borings for field tests and sample collection. Field
testing, in addition to SPT tests, consisted of vane shear tests. Shelby tubes were used to obtain
undisturbed samples. Laboratory testing consisted of moisture content, organic content,
Atterberg limits, mechanical sieve, one-dimensional consolidation, and triaxial compression
© ASCE
Geo-Congress 2020
unconsolidated undrained (UU) as well as consolidated undrained (CU) tests. Soil parameters
were developed from the laboratory and field test results (See Tables 1 and 2) for use in the
settlement and global stability analyses.
Table 2. Summary of Soil Parameters for Global Stability Analysis
Undrained Shear Internal Friction
Unit Weight
Strength Angle
Soil Description
γsat su ϕf
(pcf) (psf) °
Loose Sand 123.2 0 30
Medium Dense Sand 125.0 0 30
Lean Clay 112.6 500 0
Clayey Sand 110.0 210 0
Silty Clay 110.0 150 0
Organic Silt 105.4 160 0
Peat 100.0 100 0
FIG. 3. Typical Stratigraphy.

Twenty-six of the SPT soil borings were drilled off-shore on Superior Bay, during the winter
through ice. Ice/water was measured between 0.5 to 9 feet thick at the off-shore borings. The
general site stratigraphy consists of 0 to 29 feet of very soft and very loose material as measured
from the lake bottom (See Figure 3). Underlying soils consist of dense to hard soil to termination
depths, which ranged from 4 to 60 feet below the existing ground or ice surface. The top very
soft and very loose soil was composed of altering deposits classified as clay, silt, and silty clay.
Organic clay, organic silt, and peat were encountered within these deposits up to a depth of 26
feet below the shoreline surface. The dense to hard soil was primarily comprised of medium
dense sand with some hard silt or silty clay layers.
Groundwater was encountered at two of the on-shore SPT soil borings at a depth of 8 feet
below existing ground. The elevation of groundwater was approximately 601.8 feet, roughly the
same as the water elevation of Superior Bay.
© ASCE
Geo-Congress 2020
Table 3. Finished Grade Settlement Summary

Finished Grade Settlement (in.)
Station Elevation 609 feet
Centerline Crown
97+00 4 6
100+00 15 8
105+00 5 9
116+00 5 11
EVALUATION AND DESIGN

Relocation of Runway 14/32 at the Sky Harbor Airport into Superior Bay entailed up to 11.5
feet of new fill to the finished runway grade. The vertical stress increase due to the height of new
fill over the very soft silts, clays, and organic soil was expected to induce excessive settlement.
Settlement was estimated using a combination of the empirical Hough method (AASHTO 2012)
for granular soils as given in Equation 1 and consolidation settlement theory (Winterkorn and
Fang 1975) for plastic soils as given in Equation 2.
H   '   
Sc  log  v  (1)
C'   'v 
H    'v    
Sc  Cc log   (2)
1  e0    'v  
Soil parameters utilized in the long term primary settlement analysis were determined from
one-dimensional consolidation (ASTM D2435) laboratory test results. It was determined that the
plastic soils were normally consolidated due to the very soft consistency of soils below the lake
bottom. Settlement for granular soils was assumed to be instantaneous. Settlement calculations
were performed with new fill to the proposed runway grade at elevation 609 feet. The results
indicated the runway would settle approximately 4 to 15 inches along the runway centerline and
crown of the new fill (see Table 3).
Time rate of settlement computed from Equation 3 (Liu and Evett 2004), for new fill to the
finished runway grade, estimated up to five years to reach an acceptable settlement tolerance of
half an inch for construction of airport runways. To induce settlement and prevent settlement
greater than half an inch from occurring in the future, a surcharge was recommended to be
placed prior to construction of the runway. The surcharge is to consist of temporary fill placed 5
feet above proposed runway grade. Construction of the runway would proceed once the expected
settlement takes place. Detrimental settlement is not expected to occur during or after the end of
construction.
T
t  v H2 (3)
cv
Settlement calculations were performed for Phase 1 and Phase 2 to design the surcharge
dimensions such that the long term primary settlement takes place prior to the start of runway
construction. The Phase 1 settlement period was estimated to last 9 months between end of
construction to the finished grade elevation of 605 feet and the start of Phase 2 surcharge
construction. Settlement estimates for Phase 1 ranged between 3 to 11 inches (see Table 4). The
Phase 2 settlement period was estimated to last 6 months between end of surcharge construction
to elevation of 614 feet and the start of runway construction during Phase 3. Settlement estimates
© ASCE
Geo-Congress 2020
for Phase 2 ranged between 3 to 14 inches (see Table 4).

Table 4. Phase 1 and Phase 2 Settlement Summary
Phase 1 Settlement (in.) Phase 2 Settlement (in.)
Station Elevation 605 feet Elevation 614 feet
Centerline Crown Centerline Crown
97+00 4 6 3 6
100+00 11 6 14 8
105+00 3 6 5 9
116+00 4 8 3 7
Table 5. Global Stability Analysis Summary

Phase 1 Phase 2 Phase 3
Station Elevation 605 feet Elevation 614 feet Elevation 609 feet
No With No With No With
Geotextile Geotextile Geotextile Geotextile Geotextile Geotextile
100+00 1.6 1.8 1.1 1.2 1.4 1.6
106+00 1.8 2.4 1.1 1.3 1.6 2.0
116+00 1.8 2.4 1.0 1.3 1.5 2.0
Settlement calculations were performed to estimate the long term settlement for Phase 1 and
Phase 2 at each plastic soil layer. Time rate of consolidation theory was applied to estimate the
average degree of consolidation at the end of Phase 1 and Phase 2. Not all layers of plastic soil
were calculated to reach 100 percent average degree of consolidation. Settlement for the plastic
soil layers was estimated for the duration of the given time period using the average degree of
consolidation values for Phase 1 and Phase 2. The sum of the settlement estimated to occur
during the Phase 1 and Phase 2 surcharge time periods was greater than the estimated settlement
for the proposed runway grade. Therefore, it was established that a surcharge to an elevation of
614 feet would theoretically result in 100 percent average degree of consolidation of the plastic
soil layers for a runway constructed to an elevation of 609 feet if constructed as described during
design of Phase 1 and Phase 2.
Global stability of the new fill over existing soils was evaluated at three critical sections
along the runway. The critical sections were modeled using SLOPE/W software for each phase
of the project. Results for global stability of the Phase 2 surcharge were not acceptable because
the factors of safety were less than 1.2 for a temporary condition due to the low strength of the
soils within the bay and 16.5 feet of new surcharge fill (See Table 5). A high strength reinforcing
geotextile was recommended to be installed across the site prior to fill placement. Global
stability analyses required reinforcement with a minimum ultimate tensile capacity of 4,800
pounds per foot in order to achieve a satisfactory factor of safety for the temporary surcharge.
CONSTRUCTION
Phase 1 construction commenced in the fall of 2017 by installing a floating silt curtain
around the perimeter of the site prior to placement of new fill in the bay. A high strength
reinforcing woven geotextile was installed at the existing shoreline approximately at elevation
600 ± 1 feet extending outward along the lake bottom to the edge of the proposed fill section.
© ASCE
Geo-Congress 2020
Geotextile rolls were sewn together side by side to form a larger panel for installation. Each
geotextile panel was three rolls wide, approximately 40 feet, and was pulled into the water by a
boat. Workers helped guide each panel into place. As more panels were installed, steel pipes
were used to join adjacent panels to maintain proper overlap between panels. Steel pipes were
also used on the ends of the panels to keep them in place.
Material for the new fill conforming to Select Granular Borrow (MnDOT 3149.2.B.2) was
barged to a temporary dock facility constructed on site. The new fill material was placed over the
geotextile to approximately one to two feet above the water surface and compacted. Subsequent
lifts of one foot were then placed and compacted to achieve the Phase 1 elevation of 605 feet.
However, survey data for the settlement plates indicates the new fill was placed up to an
elevation of approximately 604.4 feet rather than the proposed elevation of 605 feet. Settlement
plates were installed within the new fill as soon as practical to monitor settlement during Phase 1
and Phase 2 (See Figure 4).
FIG. 4. Settlement Plates Prior to Installation.

Once the Phase 1 fill was in place, construction of the new engineered shoreline commenced.
The new fill was excavated and reworked to the designed dimensions. The contractor
accomplished this by using global positioning systems (GPS) in conjunction with digital terrain
models of the proposed shoreline generated in computer aided design (CAD) software. Riprap
bedding material was furnished on the slope of the new fill. Riprap bedding material conforming
to Granular Filter (MnDOT 3601.2.B.1) was 18 inches thick. Riprap was placed using the same
GPS technology used to construct the new fill slope and placement of the riprap bedding to
establish the proposed shoreline. Riprap material conforming to Class V Riprap (MnDOT
3601.2.A.2) was 30 inches thick. The riprap was finished to the proposed elevation of 605.5 feet
with a slope of 4 horizontal to 1 vertical extending into the bay. This concluded Phase 1
© ASCE
Geo-Congress 2020
construction.
Construction of Phase 2 commenced in the fall of 2018 by barging in approximately 35,000
cubic yards of sand conforming to Select Granular Borrow (MnDOT 3149.2.B.2). The surcharge
was placed in 3 foot lifts and compacted to a standard Proctor dry density of 95 percent to the
proposed height of 614 feet. The surcharge was completed within six weeks and was left in place
until Phase 3 construction, scheduled to start in fall of 2019.
MONITORING
As part of the project design, an instrumentation and monitoring program was developed to
monitor settlement and general construction performance. Settlement plates were used to monitor
settlement associated with the new fill in conjunction with drone survey. GPS was used during
placement of finished riprap both under water and above the water surface. A video camera
system was used to monitor day-to-day progress and project operations.
Settlement across the project was primarily monitored using nine settlement plates
strategically located throughout the project (See Figure 5). The settlement plate locations were
selected to coincide with areas where settlement was estimated and at locations considered to be
critical. Survey of the settlement plates was taken twice per week for two weeks once the
embankment surcharge placement was complete and then once per week until the end of the
surcharge period. Eventually, the survey schedule was revised to once a month. Elevations and
measurements were provided to the nearest 0.01 foot. Survey of the settlement plates was not
started until after the majority of the new fill was placed as part of Phase 1. Due to the project
construction sequencing, the settlement plate survey data did not include the majority of
settlement induced by new fill below an approximate elevation of 604.4 feet. Survey data for
each settlement plate was logged and charted immediately after installation to track settlement
and understand the consolidation behavior of the soil under the load of the new fill.
FIG. 5. Settlement Plate Locations.

An initial drone survey was performed after the surcharge was completed as part of Phase 2.
A subsequent drone survey is scheduled to be completed prior to the start of Phase 3. The two
surveys are to be compared in order to estimate settlement induced by the surcharge across the
site. The results of the drone surveys could be used in the future to validate the results of the
© ASCE
Geo-Congress 2020
settlement plate survey results, the interpretation of the survey data, and compare the different
survey methods.
GPS was used in construction of the Phase 1 fill and the riprap both under and above the
water surface. Different surface models were developed in CAD during design with the proposed
fill, riprap bedding, and riprap. These models were provided to the contractor for use during
construction. The contractor was able to use the model data in conjunction with their equipment
to monitor the elevations of the excavation and placement of material. The use of GPS during
shoreline construction confirmed the material design thicknesses were achieved without the need
to increase material quantities typically associated with uncertainty for underwater placement of
material. This resulted in cost savings for riprap material and confidence in the consistency of the
riprap thickness throughout the shoreline.
A video camera system was installed to monitor project operations in coordination with
EarthCam Webcam Technology Experts. The camera was accessed remotely to monitor day-to-
day progress and project operations. The camera’s pivot and zoom capabilities allowed for
precise monitoring throughout the site. In addition, the camera was set up to automatically take
pictures at intervals of every 5 minutes.
FIG. 6. Phase 1 Settlement.

RESULTS AND CONCLUSIONS
Settlement was evaluated using settlement plate survey data. As previously stated, survey of
the settlement plates was not started until after the majority of the new fill was placed during
Phase 1. However, settlement plates SP2, SP3, and SP4 measured settlement which was
interpreted to be the tail end of a consolidation curve. These settlement plates are located close to
the shoreline where the new fill was higher, due to sloping surface into the bay, with deeper and
© ASCE
Geo-Congress 2020
lower strength soils than at the runway centerline. The results of SP3 and SP4 substantiate that
settlement was still occurring toward the end of Phase 1. Settlement from the survey data varied
between 0 to 3 inches for Phase 1 (See Figure 6). The measured settlement is not comparable to
the estimated settlement during design of Phase 1 given the delay between fill placement and the
beginning of monitoring. It was estimated that approximately 3 to 4 inches of settlement may
have occurred prior to the start of the monitoring period.
Settlement during the Phase 2 surcharge was recorded at all settlement plates, with the
exception of SP6 which was located outside the zone of influence of the surcharge. Settlement
Plates SP2, SP3, SP4, and SP5, located towards the shoreline, recorded settlement despite being
located at the toe of the surcharge. The results indicated the surcharge zone of influence extended
to the edge of the fill as expected. Visual observation by the project geotechnical engineer
reported no horizontal shift or tilting of the settlement plate risers. This suggested no global
stability issues or sliding occurring towards the bay. Thus, the survey data infers the ongoing
settlement is primarily due to the surcharge load. Settlement varied approximately between 1 to 5
inches for Phase 2 at the toe of the surcharge (See Figure 7). The settlement calculated during
design was estimated at the crown of the surcharge, and therefore, the recorded settlement at the
toe of the surcharge is not directly comparable to the calculated settlement.
FIG. 7. Phase 2 Settlement at Toe of Surcharge.

Settlement at the proposed runway centerline during Phase 2 was observed in settlement
plates SP1, SP7, SP8, and SP9. These plates were installed at the end of Phase 1, away from the
proposed shoreline, resulting in minimal observed settlement approximately between 0 to 0.5
inches. Settlement during Phased 2 was recorded between 6 and 11 inches (See Figure 8). The
recorded Phase 2 settlement for settlement plates along the runway centerline is within the
© ASCE
Geo-Congress 2020
estimated settlement of 3 to 14 inches. The consolidation curve for the centerline settlement
plates demonstrated a steep increase in settlement during the construction of the surcharge that
does not resemble a traditional curve typically associated with plastic or fine grained soils. The
consolidation curve during this period resembles more immediate settlement behavior rather than
a typical consolidation curve. This is likely due to the presence of the very loose granular soils
encountered during the subsurface investigation. However, once the more immediate settlement
behavior occurred, a more typical consolidation curve was observed. Given the data available for
all settlement plates to date, it was concluded that all long term primary settlement has occurred
throughout the zone of influence of the surcharge.

The use of a drone in combination with 3D analysis in the point cloud generated from drone
images, provided precise quantity measurements. This allowed for construction quantities to be
measured and used for payment to the contractor. Drone surveys could be used to calculate
settlement extensively throughout a site rather than at limited locations.
The use of GPS provided precise elevations for construction of the shoreline. This
technology allowed the riprap to be constructed without the need to increase riprap quantities,
typically included for underwater placement of material. This resulted in cost savings for less
material and time saved during construction of the shoreline.
The camera system assisted in tracking material delivery to the project site. This was
especially valuable to accurately estimate material quantities for payment. Remote access to the
camera system with internet connectivity was advantageous to view the site and discuss
construction challenges with the contractor and engineer present at the site. In addition to
traditional documentation methods, photos and videos of the construction progress taken by the
camera system provided corroboration to the documentation process.
FIG. 8. Phase 2 Settlement at Proposed Runway Centerline.
© ASCE
Geo-Congress 2020
ACKNOWLEDGMENTS
The authors wish to recognize the assistance and support of Wayne Wambold, Sr.
Geotechnical Engineer at Short Elliott Hendrickson Inc., for providing engineering and peer
review support during design and construction of the project. Special thanks to the Duluth
Airport Authority for the opportunity to work on this project. The general contractor for the
project was Northland Construction LLC.
REFERENCES
AASHTO. (2012). AASHTO LRFD Bridge Design Specifications. Sec. 10.6.2.4.2, Washington,
D.C.
Winterkorn, H.F. and Fang, H.Y. (1975). Foundation Engineering Handbook. Van Nostrand
Reinhold Company, New York, N.Y.
Liu, C. and Evett, J.E. (2004). Soils and Foundations. Pearson Prentice Hall, Upper Saddle
River, N.J.
© ASCE
Geo-Congress 2020
A Failure Mechanism around Axially Loaded Sockets in Weak Rock

Pouyan Asem, Ph.D., A.M.ASCE1; and Joseph F. Labuz, Ph.D., P.E., F.ASCE2
1
Post Doctoral Fellow, Dept. of Civil, Environmental and Geo-Engineering, Univ. of Minnesota,
Minneapolis, MN. E-mail: pasem@umn.edu
2
Professor and Head, Dept. of Civil, Environmental and Geo-Engineering, Univ. of Minnesota,
Minneapolis, MN. E-mail: jlabuz@umn.edu
ABSTRACT
This paper is concerned with the failure mechanism predicted from the cylindrical cavity
expansion model for a rock mass around the sidewalls of axially loaded sockets. The failure
mode, which depends on the initial lateral stresses, load-induced radial stresses, and tensile
strength of the rock, is critical to identify so that the axial capacity and stiffness of the foundation
can be determined. A comprehensive database of in situ load tests, specifically drilled shaft,
anchor, and plug load tests in weak rock provides validation of the model. The initial lateral
stresses in the rock mass are estimated using the coefficient of earth pressure and knowledge of
the average vertical stresses at each load test elevation. The back-calculated shear stress–shear
displacement relationships for socket sidewalls in each load test are used to calculate load-
induced radial stresses developed on the socket sidewalls. The tensile strength is obtained from
tabulated ratios of uniaxial compressive strength to tensile strength and knowledge of the
measured compressive strength. The failure mechanism around the socket sidewalls is analyzed
by applying the Fairhurst (parabolic) failure criterion. Theory shows that the rock mass generally
fails by circumferential tension, which is in agreement with field observations.
1 INTRODUCTION
In the design of foundations for heavily loaded structures founded within a weak rock mass,
the failure mode is critical to identify so that the axial capacity and stiffness of the socket can be
determined. Ladanyi (1967, 1976) suggested that the failure mechanism around a pressurized
cylindrical cavity (e.g., rock socket) in rock initiates at the cavity walls. The failure mode
depends on the magnitude of the far-field geostatic stress (Po) and will take place by
circumferential tension for low values of Po or by confined compression for high values of Po.
Williams (1980) and Baycan (1996) observed that the rock mass failure around shallow sockets
in Melbourne siltstone was by circumferential tension, i.e., by the formation of radial cracks
around the socket sidewalls.
The Lamé solution has been used to investigate the mode of failure of the rock mass around
the sidewalls of axially loaded sockets (Ladanyi 1967, 1976; Jaeger et al. 2007), although the
particular failure criterion (Fairhurst 1964; McClintock and Walsh 1967; Hoek and Brown 1980)
affects the results. Additionally, the analysis of the rock mass failure mode around the socket
sidewalls requires knowledge of Po, which is not often measured and is commonly estimated
using empirical models such as that proposed by Brown and Hoek (1978), which involves
significant amount of uncertainty. Furthermore, the excavation of a cylindrical cavity in the rock
mass and the subsequent pour of concrete alters the initial state of stress in the vicinity of the
cavity walls, which further complicates the estimation of Po. In addition to Po, the uniaxial tensile
strength (T) of rock must be known. Models such as Mohr-Coulomb failure theory over-predict
T. To overcome these limitations, some researchers (Fairhurst 1964) proposed nonlinear models
© ASCE
Geo-Congress 2020
with a “natural” tension cut-off, or others developed correlations for prediction of T in terms of
the uniaxial compressive strength Co (Hoek and Bieniawski 1966; Goodman 1980). The extent of
crack formation around socket sidewalls depends on the far-field geostatic stress (Po) and the
load-induced radial stresses on the socket sidewalls (Pi). Similar to Po, the load-induced radial
stresses are not measured in the in situ load tests and constitute another important unknown in
predicting the failure mode.
This paper presents a simple model and a comprehensive database of in situ axial load tests
to study the failure mechanism of the rock mass around the socket sidewalls when shear stresses
on the socket sidewalls reach the peak side resistance (fsp). In Section 2, the geology of a weak
rock mass is reviewed. In Section 3, a database consisting of drilled shaft, and anchor/plug load
tests is introduced. These load tests are used to calculate the shear stress – shear displacement
(fs–) relationship along the socket sidewalls, which is used to determine the load-induced
changes in the radial stresses acting on the socket sidewalls. Laboratory and the in situ load tests
are used to estimate the rock mass properties (i.e. Co, T, and the rock mass deformation modulus
Em) for each load test. In section 4, we discuss the approach for the prediction of the far-field
geostatic stress (Po) and the load-induced critical stress (Pic) on the socket sidewalls. In Section 5,
theoretical analysis and the load test data are used to determine the dominant mode of failure of
the weak rock mass in axially loaded sockets. A summary of findings is presented in Section 6.
2 WEAK SEDIMENTARY ROCK
The geology of weak sedimentary rock generally consists of three distinct phases (Goodman
1993; Mesri and Shahien 2003). In phase one, parent materials are eroded, carried and
accumulated into rivers, lakes, and oceans, and consolidated or compacted under the weight of
overburden. Phase two consists of the removal of the overburden, over-consolidation of young
deposits and formation of fissures and joints at shallow depths. Phase three consists of an
increase in the water content of material along the exposed walls of the open cracks, and
subsequent softening of the rock blocks that are separated by discontinuity surfaces. Therefore, a
weak rock mass is an assemblage of relatively weathered rock blocks, which are separated by the
structural discontinuity sets (Hoek 1983; Singh and Rao 2005).
Because a weak rock mass consists of relatively weathered rock blocks and structural
discontinuities that are also subjected to the weathering, and both components affect the response
of the rock to external loads, the definition of a weak rock should reflect how weathering affects
the rock blocks and the discontinuity surfaces. Accordingly, the values of uniaxial compressive
strength (Co) that are obtained from the representative samples are used to characterize the effect
of the mineralogy and the weathering state of the rock blocks, and the Geological Strength Index
(GSI) is used to reflect the blockiness of the rock mass and the degree of the alteration of the
discontinuities such as joints and fissures. The review of the technical literature suggests that for
weak rock, the uniaxial compressive strength (Co) ranges from 0.5 – 30 MPa (Deere and Miller
1966; Barton et al. 1978; Rowe and Armitage 1987; Cepeda-Diaz 1987; Kanji 2014), and the
geological strength index (GSI) is commonly less than 70 (Hoek and Brown 1997). This
definition is consistently used in the selection of the load tests in the following section.
3 DESCRIPTION OF THE DATABASE
The database, summarized in Table 1, consists of in situ axial load tests on drilled shafts,
anchors/plugs in weak sedimentary rock. The database is described in detail by Asem (2018) and
© ASCE
Geo-Congress 2020
Asem and Gardoni (2019).

Table 1. Summary of the in situ drilled shaft, anchor and plug load tests (after Asem 2018;
Asem and Gardoni 2019)
Database component Description
Type of soft rock Sandstone and siltstone
Shale, mudstone, claystone
Limestone,
Unconfined compressive strength (Co) 0.5 kPa < Co < 36 MPa

Rock quality designation (RQD) 0 < RQD < 65%
Geological strength index (GSI) 3 < GSI < 70
Rock mass friction angle (m) 12 < m < 42 degrees
Back-analyzed modulus of deformation 10 MPa < Em < 19 GPa
(Em)
Socket geometry 46 < B < 2440 mm
0.25 < DGS < 49 m
0 < DTOR < 33 m
Test shaft concrete 21 < f’c < 55 MPa
22 < Ec < 45 GPa
Slump: 0 to 229 mm
Load test method Osterberg tests, conventional top loaded
tests, anchor load tests and plug load tests
Load test method

The load test methods include (i) the top-down (compression) load test method, where loads
are applied to the drilled shaft butt (Figures 1a, 1b, and 1c); (ii) Osterberg load test method,
where loads are applied bi-directionally using a load cell embedded in the test shaft at a desired
depth (Figure 1d); and (iii) tension load tests on plugs or anchors (Figure 1e). To separate the
side and base resistances, a load cell is located near the base (Figure 1a), strain gauges are
provided along the socket sidewalls (Figure 1b and 1d), a compressible base is provided (Figure
1c) or loads are directly applied in tension to eliminate the base resistance (Figure 1e).
Peak side resistance (fsp) and initial shear stiffness (Ksi)
The values of the peak side resistance (fsp) and the corresponding shear displacement p are
obtained from the fs– relationships (Figure 2 shows a typical fs– response), which are reported
in Asem (2018). The values of fsp are defined as the maximum shear stress achieved on the rock
socket sidewalls in each load test and Ksi is defined as the slope of the tangent line to the initial
portion of fs– response. The estimated values of Ksi represent the rate of increase in side
resistance with axial displacement of the socket at small vertical displacements.
Rock mass deformation modulus (Em)
The values of Em are estimated based on the slope of the tangent line drawn to the initial part
of the corresponding fs– response in each load test, which is represented by the slope called the
initial shear stiffness (Ksi) of the socket sidewalls. Equation (1), based on an elasticity solution, is
© ASCE
Geo-Congress 2020
both sides of Equation (1) by the circumferential area of the socket  A = πBL  , the following
expression is obtained for Em based on the in situ axial load tests:
E m = 2πLK si I (2)
Asem (2018) and Asem and Gardoni (2019) showed that the Em values are in agreement with
Em values obtained (i) from plate load tests on weak rocks in China and Taiwan that are reported
by Chern et al. (2004) and (ii) from drilled shaft base resistance measurements and plate load
tests reported by Asem (2018).
Geological Strength Index (GSI)

Hoek and Diederichs (2006) proposed relationship between Em and GSI:
100,000
Em = (3)
1+e
75 GSI  /11
and the values of Em, back-calculated for rock masses in each load test, are used to estimate GSI
for the rock formations in the database. The rock masses in the database are assumed to be
undisturbed, and thus a disturbance factor, D = 0, is used in Equation (3). The estimated values
of GSI provide an evaluation of the rock mass friction angle (m).
Friction angle of rock mass (m)
The friction angle (m) for each rock mass formation is determined using the method of Hoek
and Brown (1997) based on (i) the estimated values of GSI, and (ii) a material constant (mi),
which reflects the rock type. The corresponding values of mi are selected based on recommended
values reported in the literature (Hoek and Brown 1980; Hoek 1983; Hoek 1990; Hoek and
Brown 1997; Marinos and Hoek 2001). Table 2 shows the typical ranges of mi that are used to
estimate m.
Table 2. Typical ranges of the material constant (mi) for determination of m (after Hoek
and Brown 1980; Hoek 1983; Hoek 1990; Hoek and Brown 1997; Marinos and Hoek 2001)
Rock type Observed range of material
constant, mi
Schist 4-8
Limestone and dolomite 5-7
Siltstone 9
Mudstone and shale 7-10
Sandstone and quartzite 14-15
Andesite and diabase 17
Granite 25-33
4 STATE OF STRESS AROUND THE SOCKET SIDEWALLS

The critical radial stress (Pic) at the onset of peak side resistance (fsp), which results from the
dilation of the socket against the radial stiffness of the adjacent rock mass (Seidel and
Collingwood 2001), affects the extent of the failure zone in the surrounding rock. The value of
the critical radial stress is not measured in the axial load tests but it is estimated as:
© ASCE
Geo-Congress 2020
f sp
Pic = (4)
tan m
The rock mass friction angle (m) is used in Equation (4) because Williams (1980) observed
that the shear surface forms in the rock mass and not entirely at the rock/concrete interface. The
far-field geostatic stress (Po) is also not measured but is estimated as (Brown and Hoek 1978):
Po = Kzγ (5)
where z is the distance from the ground surface to the point of interest, K is ratio of horizontal to
vertical stresses, and  is the unit weight of the rock. In this study, three typical values of the K =
0.5, 1.5 and 2.5 (Brown and Hoek 1978) are used for the evaluation of the Po.
5 FAILURE OF THE SURROUNDING ROCK MASS
Failure mode
The following sequence of events take place as the rock socket is excavated, fluid concrete is
poured, and the axial loads are applied to the socket:
1. Excavation of the rock socket: this results in stress relief around the socket sidewalls. The
normal stresses (Pi) along the rock socket sidewalls become zero and the rock mass
rebounds toward the axis of the socket.
2. Concrete pour: when the concrete is poured, the fluid concrete exerts a radial pressure Pi
< Po on the socket sidewalls. The rock mass response is assumed to remain elastic.
3. Rock failure around the socket: as the axial load increases and slip along the socket
sidewalls initiates, the radial stresses (Pi) on the socket sidewalls increase. Two failure
modes are possible (Ladanyi 1967, 1976): (i) the rock mass fails by circumferential
tension and formation of radial cracks when the far-field geostatic stresses (Po) are small.
Subsequently, when the radial stresses (Pi) on the socket sidewalls attain a value that is
comparable to the uniaxial compressive strength (Co), a zone of crushed rock is generated
inside the radially cracked zone; and (ii) the rock mass failure occurs by confined
compression if the far-field geostatic stresses (Po) are relatively large, as in the case of
very deep sockets or pressurized deep tunnels.
The possibility of formation of a radially cracked zone is evaluated using the criterion
discussed by Ladanyi (1967; 1976). For the axisymmetric problem and ignoring body forces, the
equilibrium equation is:
dσ r σ r  σθ
+ =0 (6)
dr r
Lamé solution for a cylindrical cavity expanding in an elastic material with the boundary
conditions at r =   r = Po and at r = re  r = re yields the following expressions for the
radial (r) and tangential () stresses:
σ r = Po +  σ re  Po  re / r 
2
(7)
and
σθ = Po   σ re  Po  re / r 
2
(8)
where re is the radial distance to the boundary between elastic and cracked zones, and re is the
corresponding radial stress. A relationship between the major principal stress r, the minor
principal stress , and the far-field geostatic pressure Po is obtained by combining Equations (7)
© ASCE
Geo-Congress 2020
and (8):
σ r = 2Po  σ θ (9)
The parabolic failure criterion (Fairhurst 1964) contains a natural tension cut-off. The
condition for tensile failure is simply -3 = T as long as:
σ1  m(m  2)T  0 (10)
Thus, failure at the socket sidewall will take place by circumferential tension when:
m  m  2 1
Po < T (11)
2
where m = Co / T+1 (Fairhurst 1964). The uniaxial tensile strength (T) is often reported in
terms of the ratio Co/T (Hoek and Bieniawski 1966; Goodman 1980; Jaeger et al. 2007). Table 3
provides a summary of typical ratios (Wuerker 1956; Miller 1965; Brandon 1974; Goodman
1980).
Table 3. Typical values for ratios of unconfined compressive strength to tensile strength
(Co/T) for some rocks (after Goodman 1980; Jaeger et al. 2007)
Empirical data
Rock type Description Co/T Average Co/T Reference
Sandstone Berea sandstone 63 55 Miller (1965)
Navajo sandstone 26.3 Miller (1965)
Ohio sandstone 76 Wuerker (1956)
Siltstone Hackensack siltstone 41.5 41.5 Miller (1965)
Limestone Solenhofen limestone 61.3 31 Miller (1965)
and dolomite Bedford limestone 32.3 Miller (1965)
Tavernalle limestone 25 Miller (1965)
Oneota dolomite 19.7 Miller (1965)
Lockport dolomite 29.8 Miller (1965)
Fossiliferous (Ind.) 20 Wuerker (1956)
Shale Micaceous shale 36.3 24 Brandon (1974)
Calcareous shale 11.5 Wuerker (1956)
Granite Granite, pegmatite big 68 37 Wuerker (1956)
Granite (vt.) 32 Wuerker (1956)
Hong Kong Granite 10 Lumb (1983)
The mode of failure around the socket sidewalls is evaluated in Figures 3 and 4. Figure 3
shows the estimated values of T  m  m  2   1 / 2 plotted versus the corresponding values of
far-field geostatic stress (Po) in the rock mass surrounding the socket for K = 0.5, 1.5 and 2.5.
The results indicate that T  m  m  2   1 / 2 is commonly greater that than Po and thus in the
majority of cases considered here, the mode of failure of the rock mass is by circumferential
tension and the subsequent formation of a radially cracked zone (see Figure 5) that is in
agreement with Williams (1980) and Baycan (1996) observations from in situ load tests in
Melbourne weathered siltstone.
Figure 4 compares the values of normal stress (Pic) on the socket sidewalls with the uniaxial
compressive strength of rock (Co). Figure 4 indicates that in the majority of cases, the condition
of Co > Pic holds and thus the formation of a zone of crushed rock is unlikely.
© ASCE
Geo-Congress 2020
Figure 3. Evaluation of the possibility of formation of radial cracks around the rock socket
sidewalls (data from Asem 2018).
Figure 4. Evaluation of the possibility of formation of a zone of broken rock inside the
radially cracked region around the rock socket sidewalls using the method of Ladanyi
(1967; 1976) (data from Asem 2018).
Extent of cracking
Experimental evidence (Williams 1980; Haberfield and Johnston 1986; Briaud 1986) shows
that the dilation and the development of radial and tangential stresses results in the formation of a
zone containing radial cracks around the socket sidewalls and within the adjacent rock mass
(Figure 5). The extent of the zone of influence is determined based on the radius of the cracked
zone re, normalized by the radius of the socket ro. To derive an expression for the ratio re/ro, we
assume that after crack formation, the tangential stresses   0 in the cracked zone and thus the
equilibrium (Equation 6) equation reduces to:
dσ r σ r
+ =0 (12)
dr r
Solving Equation (12) and applying boundary conditions, the result is:
σr = σre  re / r  (13)
© ASCE
Geo-Congress 2020
Additionally, we know that at the ro, the radial stress (r) is equal to Pic when the shear
stresses on the socket sidewall become fsp. Therefore, Equation (14) reduces to:
re / ro =Pic /  2Po +T  (15)
which may be used to estimate re. The results are shown in Figure 6. Data presented in Figure 6
indicate that the re/ro is less than 10 for most load tests reported in the database. Therefore, it is
assumed that the zone of cracking extends to a maximum radius of 10ro.
6 CONCLUSIONS
This paper is concerned with the failure mechanism predicted from the cylindrical cavity
expansion model for a rock mass around the sidewalls of axially loaded sockets. A
comprehensive database of in situ load tests, specifically drilled shaft, anchor and plug load tests
in weak rock provides validation of the model. The following are the findings of this paper:
1. The failure mode depends on the initial lateral stresses, load-induced radial stresses, and
tensile strength of the rock.
2. Theory shows that the rock mass generally fails by circumferential tension, which is in
agreement with field observations.
3. The radial stresses at the onset of development of peak shear strength (fsp) are generally
smaller than the uniaxial compressive strength (Co), and hence a crushed zone is not
expected to form within the radially cracked region.
4. The radial cracks extend to a maximum distance (re) of 10ro, where ro is the radius of the
rock socket.
REFERENCES
Asem, P. 2018. Axial behavior of drilled shafts in soft rock. Ph.D. Thesis, Department of Civil
and Environmental Engineering, University of Illinois at Urbana-Champaign, Urbana,
Illinois, United States.
Asem, P., and Gardoni, P. 2019. Evaluation of the peak side resistance for rock socketed shafts
in weak sedimentary rock from an extensive database of published field load tests: a limit
state approach. Canadian Geotechnical Journal. doi:10.1139/cgj-2018-0590.
Barton, N., Bamford, W. E., Barton, C. M., MacMahon, B., Kanji, M. A., Babcock, K., Boyd, J.
M., Cruden, D., Franklin, J. A. et al. 1978. Suggested methods for the quantitative
description of discontinuities in rock masses. International Journal of Rock Mechanics and
Mining Sciences and Geomechanics Abstracts, 15(6): 319-368.
Baycan, S. 1996. Field performance of expansive anchors and piles in rock. Ph.D. Thesis,
Department of Civil Engineering, Monash University, Melbourne, Australia.
Brandon, T. R. 1974. Rock mechanic properties of typical foundation rocks. US Bureau of
Reclamation.
Briaud, J. L. 1986. Pressuremeter and foundation design. In Proceedings of the Use of In Situ
Tests in Geotechnical Engineering, Blacksburg, Virginia, United States.
Brown, E. T., and Hoek, E. 1978. Trends in relationships between measured in situ stresses and
depth. International Journal of Rock Mechanics and Mining Sciences & Geomechanics
Abstracts, 15(4): 211-215.
Cepeda-Diaz, A. F. 1987. An experimental investigation of the engineering behavior of natural
shales. Ph.D. Thesis, Department of Civil and Environmental Engineering, University of
Illinois at Urbana-Champaign, Urbana, Illinois, United States.
© ASCE
Geo-Congress 2020
Chern, J. C., Chang, Y. L., Lee, K. R., Yu, C. W., Li, T. J., Li, J. Y., and Yuan, C. H. 2004.
Correlation study on the deformation modulus and rating of rock mass. R-GT-97-04, Taipei,
Taiwan.
Deere, D. U. and Miller, R. P. 1966. Engineering classification and index properties for intact
rock. Air Force Weapons Laboratory, University of Illinois at Urbana-Champaign, Urbana,
Illinois, United States.
Fairhurst, C. 1964. On the validity of the Brazilian test for brittle materials. International Journal
of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, 1(4): 535-546.
Goodman, R. E. 1980. Introduction to rock mechanics. 2nd ed., John Wiley and Sons, Inc., New
York, United States.
Goodman, R. E. 1993. Engineering geology: rock in engineering construction. 1st ed., John
Wiley and Sons, Inc., New York, United States.
Haberfield, C. M., and Johnson, I. W. 1986. Concepts for pressuremeter interpretation in soft
rock. In Proceedings of the Specialty Geomechanics Symposium: Interpretation of Field
Testing for Design Parameters, Institute of Engineers, Australia, pp. 65-69.
Hoek, E., and Bieniawski, Z. T. 1966. Fracture propagation mechanism In hard rock. In
Proceedings of the 1st ISRM Congress, Lisbon, Portugal, pp. 243-249.
Hoek, E. and Brown, E. T. 1980. Empirical strength criterion for rock masses. Journal of the
Geotechnical Engineering Division, 106(9): 1013-35.
Hoek, E. 1983. Strength of jointed rock masses. Géotechnique, 33(3): 187-223.
Hoek, E. 1990. Estimating Mohr-Coulomb friction and cohesion values from the Hoek-Brown
failure criterion. International Journal of Rock Mechanics and Mining Sciences &
Geomechanics Abstracts, 27(3): 227-29.
Hoek, E. and Brown, E. T. 1997. Practical estimates of rock mass strength. International Journal
of Rock Mechanics and Mining Sciences, 34(8): 1165-86.
Hoek, E. and Diederichs, M. S. 2006. Empirical estimation of rock mass modulus. International
Journal of Rock Mechanics and Mining Sciences, 43(2): 203-15.
Jaeger, J. C., Cook, N. G., and Zimmerman, R. W. 2007. Fundamentals of rock mechanics,
Blackwell publishing.
Kanji, M. A. 2014. Critical issues in soft rocks. Journal of Rock Mechanics and Geotechnical
Engineering, 6(3): 186-95.
Ladanyi, B. 1967. Expansion of cavities in brittle media. International Journal of Rock
Mechanics and Mining Sciences & Geomechanics Abstracts, 4(3): 301-328.
Ladanyi, B. 1976. Quasi-static expansion of a cylindrical cavity in rock. In Proceedings of the
3rd Symposium on Engineering Applications of Solid Mechanics, pp. 219-240.
Lumb, P. 1983. Engineering properties of fresh and decomposed igneous rocks from Hong
Kong. Engineering Geology, 19(1982/83): 81-94.
Marinos, P. and Hoek, E. 2001. Estimating the geotechnical properties of heterogeneous rock
masses such as Flysch, Bulletin of Engineering Geology and the Environment, 60(2): 85-92.
McClintock, F. A., and Walsh, J. B. 1962. Friction on Griffith cracks in rocks under pressure. In
Proceedings of the Fourth U.S. National Congress of Applied Mechanics, New York.
Mesri, G., and Shahien, M. 2003. Residual Shear Strength Mobilized in First-Time Slope
Failures. Journal of Geotechnical and Geoenvironmental Engineering, 129(1): 12-31.
Miller, R. P. 1965. Engineering classification and index properties for intact rock. Ph.D. Thesis,
Department of Civil and Environmental Engineering, University of Illinois at Urbana-
Champaign, Urbana, Illinois, United States.
© ASCE
Geo-Congress 2020
Pells, P. J. N. and Turner, R. M. 1979. Elastic solutions for the design and analysis of rock-
socketed piles. Canadian Geotechnical Journal, 16(3): 481-87.
Rowe, R. K. and Armitage, H. H. 1987. A design method for drilled piers in soft rock. Canadian
Geotechnical Journal, 24(1): 126-42.
Seidel, J. P. and Collingwood, B. 2001. A new socket roughness factor for prediction of rock
socket shaft resistance. Canadian Geotechnical Journal, 38(1): 138-53.
Singh, M., and Rao, K. S. 2005. Empirical methods to estimate the strength of jointed rock
masses. Engineering Geology, 77(1–2): 127-137.
Williams, A. F. 1980. The design and performance of piles into weak rock. Ph.D. Thesis,
Department of Civil, Resources and Environmental Engineering, Monash University,
Melbourne, Australia.
Wuerker, R. G. 1956. Annotated tables of strength and elastic properties of rocks. Society of
Petroleum Engineers, pp. 23-45.
© ASCE
Geo-Congress 2020
Kennedy Bridge Instrumentation: A Pier Review

James C. Bennett, P.E., M.ASCE1
1
Braun Intertec, Minneapolis, MN. E-mail: jbennett@braunintertec.com
ABSTRACT
If you choose to read this whole story, without interruption, it will not take you 25 years to
do so. You will also not need to read this story while you are sopping wet, covered in snow,
unbearably hot, freezing cold, trapped underwater, perspiring profusely, or pelted by frozen
spheres. You will still not need to read this story with mud oozing around you, wind piercing
through you, machinery driving over you, lightning trying to strike you, stray critters nipping at
you, insects crawling all over you, artists spraying paint on you, thieves trying to steal stuff from
you, or pigeons dropping surprises on you. This story, however, is about that story. Specifically,
it is about what it takes to read (or record) such a story; while enduring the most extreme of
environments, without interruption, for 25 years. And why we would do so. But don’t worry, we
will try to keep it brief; after all, this is a bridge version. So, fasten your fall protection, and
enjoy the ride.
INTRODUCTION
This story is about the historic Kennedy Bridge. Specifically, it is about one of its piers—Pier
6—and the instrumentation sensors and systems adorning it and the surrounding soils. The
instrumentation was designed and installed to endure the most extreme of environments and
recount an uninterrupted, quarter-century long story of infrastructure performance. The following
discussion will explore how this was done, and more importantly, why.
Located 125 kilometers south of Canada, in Grand Forks, North Dakota, Pier 6 became a
dubious supporter of the historic Kennedy Bridge, which straddles the Red River of the North
and the North Dakota-Minnesota border. A major rehabilitation of the bridge was substantially
completed in December 2018, including the reconstruction of problematic Pier 6 and its
foundation system.
Knowing that the original Pier 6 would soon be replaced, why would officials plan to
permanently ground the replacement Pier 6, and still seek to design and install a babysitter for
the next 25 years? It was not just original Pier 6 behaving so badly that had officials concerned;
there was a much more complicated, bigger picture at its root.
Before we explore how the instrumentation was designed and installed to tell a story
continuously for a quarter century, it is important to understand the context of the story and why
it needs be told in the first place. There is a long, complex history behind the ultimate decision to
rehabilitate Kennedy Bridge, as well as the methods by which the rehabilitation was ultimately
executed. The following narrative focuses on some of the key historical information that
influenced how the instrumentation was ultimately designed and installed.
HISTORY
Designed in 1962 and constructed in 1963, the John F. Kennedy Memorial Bridge
(commonly referred to as the “Kennedy Bridge”) conveys US Highway 2 between East Grand
Forks, Minnesota, and Grand Forks, North Dakota. The Kennedy Bridge is known for, and easily
recognized by its steel, consisting of two Parker camelback through truss main spans with riveted
© ASCE
Geo-Congress 2020
and bolted connections. The bridge has six eastern approach spans, two river spans, and five
western approaches—the westerly pier is Pier 6 (Figure 1).
Figure 1: Kennedy Bridge West River Span and Pier 6

The Kennedy Bridge is situated amidst a complex geologic setting. The bridge straddles the
Red River of the North, which has carved its way north through a vast, deep, wet, and soft
lacustrine (lake-deposited) landscape, compliments of bygone Glacial Lake Agassiz; which all
translates to bridge piers and bents supported on piles. The surficial geologic profile at pile-
supported Pier 6, approximated below, consists of 6 distinct deposits, all dominated by clayey
soil: fill, alluvium (river-deposited soils), three lacustrine fat clays, and glacial till (Figure 2).
Soil parameters, approximated from a Braun Intertec Geotechnical Evaluation Report for the
Kennedy Bridge Rehabilitation, were included to present just how deep, wet, and soft most of the
surficial geology is at Pier 6. The remainder of this historical narrative references this same
report (Hubbard 2015).
Figure 2. Approximate Geologic Profile at Pier 6, with selected soil parameters

Complicating matters, Pier 6 is located along the outside bend of the Red River (Figure 3); an
area more prone to bank instability. Zones of instability along Red River banks often include en-
© ASCE
Geo-Congress 2020
echelon failure (sliding along multiple shear surfaces) “systems” (comprised of multiple failure
masses) whose depth increases in proportion to their upslope “reach”. Failure masses with scarps
(cracks in the ground suggesting the upper limit of a slide) in close proximity to the water are
relatively shallow in comparison to those with scarps farther upslope. Based on historical
measurements and investigations, Pier 6 lies among such en-echelon failure systems.
Figure 3. Pier 6 along Exterior Bend of Red River of the North

Although the speed is relatively low (customary for soil creep), the scale of the active failure
systems around Pier 6 is not minor. Based on boring samples with slickensides, the US Army
Corps of Engineers (USACE) originally postulated historic movement along sliding surfaces to
depths as great as 12 to 15 meters. From 1998 to 1999, coinciding with a major flood event
(Figure 4), they used surviving slope inclinometers to measure 5 centimeters of bank translation
at depths of 7.5 to 9 meters below grade. From 2004 to 2009, movement detected in three new
slope inclinometers monitored by the North Dakota Department of Transportation (NDDOT)
confirmed USACE results, and revealed shallow (1.5 to 3 meters below grade) bank translations
of 50 centimeters, and deep (6 to 7.5 meters below grade) translations of 23 centimeters.
Figure 4: Stability Analysis calibrated to 1999 event (also, Pier 6 Geologic Cross-Section)
What did this all mean for Pier 6, stuck in the middle of the action? By 2012, CH2MHill
reported 97 centimeters of total movement at the base of south end of Pier 6, and 63 centimeters
of movement relative to the overlying truss. Although that is surely extreme movement for a
© ASCE
Geo-Congress 2020
structure, this bridge was actually designed to manage it. Like many bridges along the Red River,
the Kennedy Bridge’s original design contained many engineering features that account for
future ground movement, like adjustable connections between the pier and the truss system,
whose capacity for pier offset was 127 centimeters at Pier 6. However, inspections in 2012 by
MnDOT and NDDOT showed that the structure was not simply moving uniformly; Pier 6 was
bowing, twisting, and tilting (5 centimeters horizontally per 1.2 meters vertically), all producing
up to 6-millimeter cracks in the structure. The unfavorable distress of Pier 6 had progressed to
the point where serious repairs were required (Figure 5).
Figure 5. Pier 6 before reconstruction, exhibiting tilt, cracks, and spalling

MnDOT and NDDOT collaborated and chose to reconstruct Pier 6 and its foundation system,
along with a general bridge rehabilitation. When the request for design went out for bidding,
included were prescriptions for designing a set of instrumentation sensors and systems that
would measure infrastructure performance for the following 25 years. Again, if Pier 6 and its
foundation system were going to be fully reconstructed and a host of inspections and
investigative work was already performed, why would an intense instrumentation program be
prescribed?
Following all inspections, measurements, investigations, and engineering, the level to which
Pier 6 movement was linked to ground movement remained unclear. Yes, the foundation system
was designed to move with the ground, however, it was also generally assumed that there was a
degree of plastic flow around and through the pile system, especially in this geologic scenario.
Yes, instrumentation had been installed and monitored over time, however, it was located outside
of the foundation (pile) system. Worse yet, because the foundation system likely penetrates
multiple failure masses, correlations between Pier 6 and the surrounding soil could involve
different loads, distributed differently, over different depths, and over different time intervals.
Therefore, utilizing historic instrumentation data or analyses to project future ground
movements and related structural impacts would prove futile. In addition, the nature of future
ground movement is unpredictable in this geologic setting. Future foundation systems could be
exposed to ground movements of similar, more, or less magnitude than what has already
occurred, and those movements could occur as small steps uniformly over time or as a giant leap
(or set of leaps) in the near or distant future. For example, the bank may lie dormant for years
with limited river water level and piezometric changes, and then it could experience pronounced
movement in response to significant flood events, intensified by potential channel geometry
© ASCE
Geo-Congress 2020
changes, such as levee systems increasing river capacity, scour, or other factors.
Essentially, the ground around a new Pier 6 and its foundation systems, will continue to
move, in dynamic ways we don’t entirely comprehend, and we will not know the extent or
timing until the movement occurs. As Derrick Dasenbrock with MnDOT puts it simply, “in order
to develop the best solution, you must figure out the problem.” Since the problem remains
unsolved and will pervade following reconstruction, we can, and we should, continue to study
the problem while also safeguarding against its effects (mitigating risk). However, this is only
possible if the problem is measured and monitored appropriately.
Instrumentation can tell a story that shows us something, which in effect, may ultimately
change (and improve) the ways we design, build, operate, or maintain infrastructure. For Pier 6
specifically, instrumentation can shed light on the performance of, and dynamic relationships
between a structure, its foundation system, surrounding soils, and water. In turn, this will inform
(or predict) operation and maintenance of this project, inform other facilities (or future projects),
or inform something else we did not anticipate. Thus, the purpose of the instrumentation
prescribed for the Kennedy Bridge project is for monitoring the short-term and long-term
performance of Pier 6, as it is subject to challenging site conditions whose effects we cannot
predict. This is why the story needed to be told in the first place.
Next, we will explore what it takes to tell that story. Although the specifics may be
interesting (e.g. what was done), the process of design and installation (e.g. how it was done) is
more important. The following discussions highlight arguably the most visible (thus, successful)
vehicles for instrumentation design and installation at Pier 6—a proactive approach,
collaboration, integration/continuity, creativity/innovation, and a steady slant on survivability.
These best support the ultimate goal of telling a story, continuously for a quarter century, while
enduring the most extreme of environments.
INSTRUMENTATION DESIGN
SRF Consulting Group, Inc., was selected to coordinate rehabilitation efforts and lead the
roadway and bridge design. Braun Intertec was selected as the geotechnical (and
instrumentation) consultant, whose ultimate task was to design the instrumentation program
subject to the robust minimum requirements specified by MnDOT in the contract documents.
This holistic and proactive design approach to instrumentation, set forth by MnDOT at the onset
of design simply through formulating these requirements, was a bold and unique step forward for
the traditional delivery method employed, and it was crucial to the successful installation of the
system.
Typically, instrumentation components are assigned after most of a project is designed,
which means instrumentation specifications are customarily disparate and general. The
instrumentation drawings (if they even exist) convey instrumentation locations and details on a
separate canvas and in a very general way—leaving much room for interpretation by the installer
and not much insight for the rest of the contracting team. On the contrary for this project, the
Kennedy Bridge design featured a highly collaborative process between the instrumentation
team, structural design team, and both DOTs from the onset. In effect, the fundamental intent
and expectations behind the robust MnDOT prescriptions were well understood from the very
beginning. From that point on, the design and specifications were built upon, and continually
refined as all project players brought important and practical input to the proposed solutions. By
each design submission, the proposed instrumentation program had been increasingly vetted for
structural, geotechnical, operational, and historic considerations, as well as instrumentation
© ASCE
Geo-Congress 2020
constructability and reliability. For example, the existing pile system consisted of 1:6 (horizontal
to vertical) battered 14x73 H-piles, but the replacement pile system evolved to 1:4 battered 40-
centimeter Cast-In-Place (CIP) pipe piles, which required a changing approach for strain gauge
array and ShapeArray design layout, installation details, and survivability.
As the overall design progressed, the highly collaborative journey led to opportunities to
further integrate the instrumentation design and specifications into the bigger picture. Although
Braun Intertec developed the instrumentation-specific drawings and details, Jamison
Beisswenger of SRF Consulting Group, Inc., lead structural designer on the project, spearheaded
fully integrating all instrumentation components (including the sensors, cable runs, conduit runs,
vaults, cabinets, protection, and installation details) into the complete plan set. This dramatically
improved the quality and consistency of the subsequent bidding and construction process, as all
potential project players were well aware of instrumentation types and locations, as well as how
to work with and around these features. This integration left little room for misinterpretation of
critical details—meaning the odds were drastically improved that instrumentation would be
installed as intended, and monitor exactly what it was intended to record.
The innovative instrumentation design featured new, creative, and more complete (holistic)
ways to accommodate the original program requirements. Vibrating wire piezometers were
doubled and concentrated in nested, fully grouted boreholes (5 or 10 in each), extending down
27.5 meters through both suspected piezometric zones, yielding a much more complete
piezometric profile. A radar water-level sensor was designed to be installed beneath the bridge
deck near the central pier in an out-of-the-way location above most floods. Biaxial vibrating wire
tiltmeters were situated at three locations embedded in the Pier, in order to automatically
communicate any structural tilt, twist, or bowing, which inspectors would otherwise need to
measure. The ShapeArray systems (automated inclinometers) were designed to give two
complete profiles of Pier 6, from the top the pier all the way to the bottom of the battered pile, as
well as two profiles of the surrounding soils near the south and north end of the pile systems. In
addition to simply profiling flexure and deformation over time, the duality of the ShapeArray
systems will indicate how well the new, more robust pile system is able to arrest the active
failure masses. Inclinometer systems were designed with future serviceability and instrument
cost in mind; all are accessible through protective vaults for maintenance. Each inclinometer
system was designed with two vertical conduits (27mm and 41mm), for ShapeArray relocating if
greater movement is experienced, as well as one grooved inclinometer casing so there is always a
manual baseline profile to reference if the automated system does not perform long-term. The
instrumented borings directly adjacent to Pier 6 were designed to combine and contain nested
piezometers, inclinometer casing, and ShapeArray casing.
Everything related to instrumentation was designed such that survivability was of paramount
importance, but we also sought balance between necessity and practicality. Factors considered
range from sensor and monitoring system selection, location, and protection, to conduit types and
geometries, installation details, direct-powering tamper-proof monitoring stations, embedding as
many components as possible in the concrete of Pier 6, and planning for inevitable water
intrusion. Referencing the abstract, we designed the instruments, components, and the system as
a whole to operate properly when it is sopping wet, covered in snow, unbearably hot, freezing
cold, inundated, humid, or hailing. Further, the system must operate with active surficial failures,
wind howling, machinery driving over, lightning strikes, stray critters burrowing, insects nesting,
graffiti artists defacing, thieves tampering, or an overabundance of pigeon business.
© ASCE
Geo-Congress 2020
INSTRUMENTATION INSTALLATION
Zenith Tech, Inc., (ZTI) was selected as prime contractor for the Kennedy Bridge
rehabilitation project, while Braun Intertec was selected to provide vibration monitoring services
and install all project instrumentation, which first included the authorship of instrumentation
methods statements for construction and performance instrumentation. Construction, in general,
followed a very proactive and collaborative approach, chiefly administered by David Burt,
project manager for ZTI, who was diligent with advanced project planning, communications,
scheduling, and assuring cooperation at every step of the way.

In order to reconstruct Pier 6, a temporary pier was constructed to handle bridge loads during
Pier 6 reconstruction. Automated tilt sensors were used to monitor the existing Pier 6, the
temporary pier, and new Pier 6 in real-time with alerts, to ensure confidence of structural
conditions during jacking operations and load transfer from old pier to temporary pier to new
pier. (Figure 6).
Figure 6. East-west tilt trends of existing Pier 6 prior to demolition, including load transfer
The Contractor assisted most installations for construction instrumentation, and helped to
formulate better and safer ways to perform the installations needed. The contractor provided
instrumentation installers with personnel lifts and custom tie-off points to allow for safer, more
efficient installations, modifications, and removals. No unsafe structural movement was noted
during the Pier 6 transitioning.
Over a sporadic, yet intensive, series of strategically timed installations between August 2017
and November 2018, the performance instrumentation sensors and system were installed. Since
the majority of instrumentation sensors and systems were embedded in the Pier 6 structure, their
respective installations were closely coordinated with the personnel and schedules of contractors
performing pile driving, formwork, erection, concrete, flatwork, painting, electrical, and so on.
To further promise quality, we collaborated with Moorhead Electric Inc., a local electrical outfit
already on the project, for installing all conduit systems, vaults, and routing instrument cables. In
the end, it was truly a team effort; and since the instrumentation was interwoven throughout the
plan set, the initial heavy lifting of collaboration and managed expectations was already
complete.
Noteworthy installation activities include the variety of collaborative effort with the
Contractor that resulted in all instrumentation components successfully installed into the piles,
before, during, and after the piles were filled with concrete and the foundations were formed.
Furnishing sections of our strain gage array with in-house machined centralizers—cut from large
steel pipe, welded, and hole-drilled for rebar to slide through (Figure 7)—the Contractor assisted
installation down the pipe pile using a crane to suspend the array as welders attached each
lowered section. Similarly, the Contractor assisted the installation of the ShapeArray conduits
utilizing a personnel lift and rigging system to support the bottom of the increasingly heavy
combined section, while attaching and lowering additional sections. Close care was provided to
© ASCE
Geo-Congress 2020
protect the instrumentation in a tight space, and successfully route all cabling and conduit to the
top of Pier 6, where the monitoring cabinets would reside. Time lapse video was employed to
highlight and monitor the construction of the new pier.
Figure 7. Strain gage arrays, with fabricated centralizers, before installation in pile
Figure 8: Instrumented borehole installation, utilizing a well drilling rig

As with instrumentation design, creativity and survivability was integral to installation.
Every single step of installation considered additional care or protection, and often times,
resulted in subtle changes or supplements that promote system survivability. For example, none
of the instrumented boreholes were even drilled until nearly substantial completion, when
grading was basically complete, so that no materials, equipment, or related construction activity,
could negatively impact the instrumented areas. Also, to accommodate large annular borehole
requirements (recall that a set of nested piezometers, two sets of ShapeArray conduit, one tremie
© ASCE
Geo-Congress 2020
pipe, and one grooved inclinometer casing, were required to all fit in one borehole), our drillers
successfully utilized a well drilling rig to perform the instrumented boreholes. This both
dramatically expedited the drilling process, and reduced the potential for damage when pulling
auger or casing up around instruments and associated cabling that is still contained within the
borehole (Figure 8).
By the end, a total of twenty-five vibrating wire piezometers, twelve strain gages, six
ShapeArrays (two adjacent to the pier in the ground, two within Pier 6 piles, and two within both
Pier 6 columns), six grooved inclinometer casings for manual readings, six tiltmeters, and a
radar-sensing river elevation gage were installed. All of these installations included cables routed
through a series of conduit and junction boxes, which ultimately connected to monitoring
cabinets that have a constant dedicated power source and a strong cellular signal for remote data
collection.
INSTRUMENTATION PERFORMANCE
For the rehabilitation of the Kennedy Bridge, which included the full reconstruction of Pier 6,
instrumentation sensors and systems were designed and installed to endure the most extreme of
environments and recount an uninterrupted, quarter-century long story of infrastructure
performance. The discussion above explored why dynamic and challenging site conditions—that
we do not fully comprehend and cannot currently predict—required the short-term and long-term
performance monitoring of Pier 6. The discussion above later explored how this was done
through a proactive approach, collaboration, integration, creativity, and a focus on survivability.
Currently, several months into telling the 25-year long story, we have already encountered a
major flood event, where the river level rose 9 meters at the Kennedy Bridge (Figure 9).
Figure 9. Flood Effects at Kennedy Bridge, comparing Fall 2018 (left) to Spring 2019
(right) (Photo on right compliments of Paul Konickson, MNDOT)
So how did the instrumentation sensors and systems hold up? We are happy to report that the
system operated continuously during the flood, and all sensors continue to report data accurately.
© ASCE
Geo-Congress 2020
We included a quick glimpse of some instrumentation performance data in response to the

flooded conditions (Figure 10).
Figure 10. Kennedy Bridge performance instrumentation data, 4 sets

Over the course of 25 years, we can expect to see the rest of the story unfold as Pier 6 and its
foundation system performs, while other related conditions also continue to inform the situation.
In the meantime, we have many photos, videos, and lessons learned, to consider and reflect upon
for the next successful instrumentation project.
© ASCE
Geo-Congress 2020
REFERENCES
Beisswenger, Jamison. (2016) Kennedy Bridge Rehabilitation Bridge Plans. SRF Consulting
Group, Inc., Minneapolis, MN
Bennett, Jimmy. (2016) Kennedy Bridge Rehabilitation Division SI, Drawings G1 through G8.
Braun Intertec, Minneapolis, MN
Google Earth. (2019) Satellite Imagery
Haugstad, Matt. (2017-2019) Untitled Photographs during Kennedy Bridge Rehabilitation
Project. Braun Intertec, Minneapolis, MN

Hubbard, Chuck. (2015) Geotechnical Evaluation Report: Kennedy Bridge Rehabilitation. Braun
Intertec, Minneapolis, MN
Konickson, Paul. (2019) Untitled Photograph of 2019 Red River Flood in Grand Forks.
MnDOT, District 2
MnDOT (Minnesota Department of Transportation). “Hwy 2 Kennedy Bridge”
<https://www.dot.state.mn.us/d2/projects/kennedybridge/index.html> (2019)
MnDOT (Minnesota Department of Transportation). (2014) Request for Proposal: Detail
Rehabilitation Design – Bridge 9090 (Kennedy Bridge)
© ASCE
Geo-Congress 2020
A Review of LRFD Bridge Foundation Design and Construction in South Dakota

Brett E. Belzer1; Bret N. Lingwall2; and Lance A. Roberts3
1
Project Engineer, RESPEC, Rapid City, SD. E-mail: Brett.Belzer@RESPEC.com
2
Assistant Professor, South Dakota School of Mines and Technology, Rapid City, SD
(corresponding author). E-mail: bret.lingwall@sdsmt.edu
3
Professor and Chair of Mining Engineering, South Dakota School of Mines and Technology,
Rapid City, SD. E-mail: lance.roberts@sdsmt.edu

ABSTRACT
Load and resistance factor design (LRFD) is a reliability-based limit state design
methodology. In this study we investigate the concept, application, and implementation status of
LRFD in shallow and deep foundation designs for the South Dakota Department of
Transportation. Challenging the use of LRFD in South Dakota and neighboring states are the
frequent occurrence of difficult geologic and geotechnical conditions including expansive soil
and rock, deep seated landslides, soft and/or highly weathered shales, intermediary geologic
materials, highly corrosive soils, and lack of available high-quality engineered fill materials. For
deep foundations, pile load testing data are available to evaluate effectiveness of LRFD
methodologies in the geologic materials present within the state. However, whereas deep
foundation design parameters are relatively well established for LRFD methods, local calibration
of shallow foundation design parameters has not yet been performed in most states. This paper
examines the current status of LRFD and discusses its implementation for shallow and deep
foundation design in South Dakota and provides a set of recommendations that engineers can
consider as they pursue implementation of LRFD on construction projects with the ultimate goal
of economical designs incorporating quantitative estimations of failure.
INTRODUCTION
Designing a structural foundation requires an iterative process between structural and
geotechnical engineers. With the Load and Resistance Factor Design (LRFD) approach, the
factored load is determined by selecting appropriate limit states and Load Factors and applying
to the nominal load. The Load Factor varies for each type of load, and different magnitudes of
the factored load can be obtained by choosing from combinations of Load Factors. Load
carrying capacity (Resistance) is similarly modified with Resistance Factors. These Resistance
Factors are generally “typical” or “universal,” taken from design codes. If local conditions or
site-specific testing is available, the typical Resistance Factors may be adjusted to account for
locally known variability in load, materials, construction, model error, acceptable risk to the
public and failure consequences (Fenton et al. 2008). To properly apply LRFD for foundation
design, code defined Resistance Factors are adjusted based on their different probabilities of
occurrence and acceptable level of risk to the public for the specific conditions of a location or
area (Fenton et al. 2008). A major challenge when converting from Allowable Stress Design
(ASD) [i.e. Working Stress Design (WSD)] and its use of broad Factors of Safety (FS), to LRFD,
is that it is essential for designers to be aware that modifying the code defined default Resistance
Factors for local specific demands or conditions is crucial to successful reliability-based design.
Foundations are designed for two limit states: ultimate (ULS) and serviceability (SLS). ULS
are associated with dangerous conditions and can involve serious negative outcomes including
© ASCE
Geo-Congress 2020
structural collapse. ULS are conditions of large and occasionally uncontrolled deflections.
Although SLS are less severe, they are associated with impaired functionality and can involve
adverse outcomes such as excessive settlement in the foundation (Paikowsky et al. 2009). Many
Transportation Agencies, and other local, state and federal regulatory agencies (i.e. FHWA, and
AASHTO) have defined the SLS to be 25 to 50 mm displacement for shallow and deep
foundations. The SLS is more difficult to define than the ULS, as the functionality limitations
from excessive settlement vary by structure. More rigid structures tend to have more stringent
SLS, while more ductile structures tend to have less stringent SLS. For example, a pre-cast
concrete bridge with short span can tolerate less differential movement before cracking than a
long steel bridge. Neither can tolerate the large uncontrolled deflections that occur at an ULS.
Spread footings are common shallow foundation elements used to support buildings and
lightly loaded structures. The use of spread footings to support highway bridges is primarily
limited to 1) bridges with no hazard of scour, 2) bridges founded on shallow rock or sand/gravel
formations, and/or 3) use within systems such as the GRS-IBS systems or Mechanically
Stabilized Earth (MSE) abutment systems. Shallow foundation load carrying capacity is derived
primarily from the bearing capacity of the soil-foundation system as long as a SLS is not
reached. Driven steel piles, meanwhile, are common deep foundation elements used to support
bridges and other heavy structures. In deep foundations, the load carrying capacity is developed
through a combination of side friction and tip resistance. Static capacity prediction methods are
often employed to predict load carrying capacity of foundations and the computed capacity can
vary widely. Improved implementation of LRFD (i.e. refinement of Resistance Factors) is based
on quantifying the uncertainty within a capacity prediction method by comparing the predicted
resistance to the actual resistance acquired from field load test data. Reducing the variability in
predicted load resistance leads to more efficient designs as Resistance Factors are refined.
Two research projects funded by the South Dakota Department of Transportation (SDDOT)
are being conducted to support the implementation of LRFD for the design of shallow
foundations and driven piles. The shallow foundation project consists of the review of historic
and current successful use of shallow foundations for bridge or wall projects, and an examination
of the implementation of LRFD (or LRFD-like concepts in projects that pre-date widespread use
of LRFD such as Load Factor Design [LFD]). The deep foundation project consists of Static
Load Tests (SLT) conducted on driven HP-section piles at several test sites across the state of
South Dakota. In general, the historic project sites and driven pile test sites are described
geologically as either: (1) a shale site or (2) a glacial till site. However, the shale sites are often
overlain by several meters of sediments, typically loose silty sands, which are referred to as
overburden in this study. Additionally, some sites consist of sand and gravel overlying siltstone
formations.
The LRFD approach proportions safety margins by ensuring that the sum of the risk
acceptable probable loads is not larger than the sum of the risk acceptable probable resistances.
This is done by multiplying the nominal design load effects by the Load Factors to make them
higher percentile probability, and multiplying calculated load carrying capacities by Resistance
Factors to them lower percentile probability, as shown in Equation and Figure 1 (Lesny and
Paikowsky 2011; Paikowsky et al. 2010). Figure 1 is an example of a probability density
function for load effect (Q) compared to two different resistances (R) where the relative
uncertainty in loads and resistances are demonstrated. Note in Figure 1 that the nominal values
are not the mean values, and that the Load and Resistance Probability Density Functions overlap
(as they always do).
© ASCE
Geo-Congress 2020
Figure 1. Probability Density Function Example (From Paikowsky et al. 2010). The two
blue lines represent two different uncertainty distributions for a resistance.
ΦRn   in1 ηi λ i Q i Equation 1.

where:
Φ = Resistance Factor that accounts for uncertainty in nominal geotechnical resistance
Rn = nominal geotechnical resistance available to resist given load effects
λi = Load Factor for a given load and load combination that accounts for uncertainty in
nominal design loads
Qi = nominal load of a given type (e.g. dead, live, etc.) that acts on the foundation
ηi = load modifier factor that relates to ductility, redundancy, and operational importance
Qn = nominal load effect
Rn = nominal resistance
mQ = mean load effect
mR = mean resistance
(FS) ̅ = central factor of safety (mR/mQ) as defined by Paikowsky et al. (2010); the ratio of
the mean resistance and the mean load effect distributions which represents the safety margin
between means and is NOT a conventional FS as used in ASD.
The primary lesson of Figure 1 for this paper, is that there are many possible PDFs of
Resistance that can occur at a site. These various PDFs depend on the load carrying capacity
calculation method used, the subsurface investigation program used, the foundations used, and
the actual variability and uncertainty at the site. AASHTO has provided Load and Resistance
Factors for bridge design in the United States that account for the uncertainty and variability
inherent to the calculation methods, foundation types and the subsurface investigation program.
Table 1 presents select LRFD typical AASHTO Resistance Factors for foundation design.
However, AASHTO cannot define universally the uncertainty and variability at a particular site.
Indeed, AASHTO requires the Load Factor for SLS to be defined on a project basis. Thus,
AASHTO allows for refinement of these Resistance Factors if site-specific testing is available.
© ASCE
Geo-Congress 2020
Table 1. Select code defined Resistance Factors for foundations (AASHTO 2017)
Resistance
Method/Soil/Condition
Factor
Theoretical method (Munfakh et al. 2001), in clay 0.50
Theoretical method (Munfakh et al. 2001), in sand, using CPT 0.50
Theoretical method (Munfakh et al. 2001), in sand, using SPT 0.45
Bearing
Resistance
b
Semi-empirical methods (Meyerhof 1957), all soils 0.45
Footings on rock 0.45
Plate Load Test 0.55

Driven Piles Driving criteria established by successful static load test or dynamic
0.75
– Dynamic testing conducted on 100% of production piles
Analysis  dyn Driving criteria established during dynamic testing on at least two piles
0.65
and Static per site condition
Load Tests Wave equation analysis without PDA or load test 0.50
Clay and mixed soil – α-method 0.35
Driven Piles Clay and mixed soil – λ-method 0.40
Static
Analysis
 stat Sand – Nordlund/Thurman method 0.45
Methods Sand – SPT method of Meyerhoff 0.30
End bearing on rock 0.45
Side resistance – clay 0.45
Drilled
Shafts Tip resistance – clay 0.40
Static  stat Side resistance – IGM 0.60
Analysis Tip Resistance – IGM 0.55
Methods
Tip Resistance - Rock 0.50
Moreover, the adoption of LRFD in geotechnical-based designs has not been as universal as
in structural engineering, because: 1) the high geographical variability and natural uncertainties
in soils and rock, and 2) the lack of singular universal calculation methods for standard design
calculations that applies to all possible geologic conditions. Soil is a heterogeneous and
anisotropic material that changes with both time and space, making universal statistical
Resistance Factors difficult to develop. Geotechnical load carrying capacity calculation methods
are generally developed under certain constraints of applicable geologic conditions further
complicating the effectiveness of universal geotechnical Resistance Factors. Thus, more reliable
estimation of the Resistance Factors for geotechnical LRFD calculations is desirable. This can
only be achieved by reviewing previous design parameters that have been implemented across
the wide range of possible calculation methods and comparing to large amounts of full-scale test
data. To take full advantage of the benefits of the LRFD approach in geotechnical design, it is
recommended that the accuracy and reliability of the design methods and input parameters be
reviewed for applicability of typical (i.e. universal code based) LRFD Resistance Factors and
that local refinements to Resistance Factors be developed (Paikowsky et al. 2009, 2010).
SHALLOW FOUNDATIONS
The implementation of LRFD for shallow foundations has been limited compared to that for
deep foundations. Load testing of shallow foundations is much more difficult than for deep
foundations, especially to failure states. Thus, there is the shortage of the available field data
needed to execute competent statistical analyses and obtain reliable correlations at scale for
© ASCE
Geo-Congress 2020
development of LRFD Resistance Factors. As a result, there is more uncertainty in Resistance

Factors for shallow foundations. Indeed, in some cases geotechnical shallow foundation LRFD
Resistance Factors have been estimated rather than rigorously derived due to the lack of
available data to use in robust statistical analyses.
Even with the mandatory use of LRFD for federally funded construction projects, the level of
implementation has differed markedly from state to state. For shallow foundations, reaching the
SLS is much more likely to occur prior to reaching a factored ULS, driving SLS design to be the
primary concern for many agencies when approving shallow foundations. For many designs, the
Resistance Factor value is dependent on the degree of understanding of the site and structure.
Fenton et al. (2008) elucidate the three levels of understanding and demonstrate the
implementation in potential local refinement of Resistance Factors in Table 2, in which universal
code defined Resistance Factors are the “typical” and with either high or low understanding, the
Resistance Factor may be refined as appropriate.
 High Understanding – This level of understanding applies when models are used that
have been proven to achieve a high level of confidence with their predictions; extensive
site investigations.
 Typical Understanding – Here, conventional prediction models are combined with a
standard project-specific site investigation.
 Low Understanding – Conventional prediction models are combined with an
understanding of the site based on previous experience or extrapolation from nearby or
similar sites
Table 2. Shallow foundation Resistance Factors depending on the degree of understanding
of the site (modified after Fenton et al. 2008)
Shallow Foundation Degree of Understanding
Resistance Factors Low Typical High
Bearing 0.45 0.50 0.60
Vertical (settlement) or
0.70 0.80 0.9
lateral movement
Bridges on shallow foundations in South Dakota are most common when the foundations are
bearing on a weak rock or an Intermediary Geologic Material (IGM) such as a shale, very hard
clay, or a weathered siltstone. SDDOT does not permit shallow foundations to bear on expansive
shale formations due to the high risk of uplift and swell in these materials. Note in Table 1 that
AASHTO has only well-defined the Resistance Factors for IGM geologic materials for drilled
shaft deep foundations! As bridge shallow foundations in South Dakota most often bear on a
weak rock or an IGM material and AASHTO has not provided code defined typical Resistance
Factors, LRFD shallow foundation design methods used by SDDOT must be based at least partly
on modifications of Resistance Factors in Table 1 using rationale such as shown in Table 2,
while knowing and understanding past performance of shallow foundations on similar structures.
This “past performance” and “degree of understanding” driven design framework (approach)
is based on the idea that if sufficient high quality and reliable field and laboratory tests are not
available (and rarely are given the geologic materials in South Dakota), then rather than use
inferior test data in a complicated calculation, or assume a bearing capacity presumptively, that
bearing capacities estimated or back-calculated from past performance of other foundations in
the same geologic formation is a superior method. This is reasonable as the bearing capacity
equations and presumptive values used for conventional AASHTO designs assume consistent
sand, clay, or hard rock below the shallow foundations, a situation which is rare in South Dakota.
© ASCE
Geo-Congress 2020
These consistent geologic materials tend to be easy to investigate, drill, sample, test and quantify.
However, the IGM materials that are most common in South Dakota are not easy to investigate,
drill, sample, test or quantify. As an example; in conventional clay soils, a Shelby Tube sample
of undisturbed clay is sampled in the field and transported to a laboratory where the soil is
extruded, trimmed and then tested in Unconfined Compression, 1D consolidation,
Unconsolidated Undrained Triaxial Compression, and other tests. In contrast, the IGM materials
in South Dakota are difficult to sample, experiencing large amounts of disturbance during
sampling, transportation, extrusion, trimming, and testing. This makes the field and laboratory
test results less reliable and raises questions as to the veracity of the parameters derived from
testing these samples for conventional design calculations.
Past performance design methods on the other hand, follow the logic that the best predictor
of future system performance is past system performance. If a thorough investigation is
performed at a bridge site, that characterizes the geologic material sufficient that the variability
and uncertainty can be qualitatively or quantitatively known, then bearing capacities observed in
previous footings in similar geologic materials will provide an upper-bound value for design of
the current system. This upper bound allows local refinement of Resistance Factors. This method
is not optimal, as it depends on sufficient past performance observations in a sufficient number
of geologies. However, as the risks associated with designs using inferior laboratory or field data
are high compared to the past performance approach, past performance approaches may be
favored. Past performance design methods are part of the Observational Approach recommended
by Terzaghi and other pioneers of the profession.
To document the design methods used by SDDOT over the last 60 years, the research team
reviewed the paperwork for 27 geotechnical reports for 22 selected SDDOT construction projects
carried out between 1957 and 2015. Every project design calculation brief and/or Engineer’s
Notes carefully considered potentially expansive soils or rock for the project. Not all project
documents included testing for expansive soils, but all projects considered it in design (not all
test data survives decades of file management). Occasionally the use of published geologic maps
for an area were used to document otherwise known or suspected expansive formations in project
documents. All project documents contain recommendations and construction details for
mitigation of expansive soil or rock if present, or suspected, at the site. Though not a Resistance
Factor design particular in LRFD, expansive or collapsible soils are a present hazard to shallow
foundations world-wide and require expert consideration in design and construction.
While limited field performance data are available for these projects in the form of surface
settlement data during construction, unfortunately for furthering the use of LRFD for the State of
South Dakota, no local full-scale load testing up-to or past failure was identified to be used in
full reliability-based calibration of LRFD factors for local conditions. Insufficient field and
laboratory data were present for nearly all of the projects for statistical analysis of subsurface
conditions. The authors of this report are unaware of any regional rigorous instrumentation or
observational tracking program of presumptive shallow foundation design on IGM [short to
long-term] performance that has been undertaken to prove if the past performance driven
approach is superior to other approaches. Thus, past performance was entirely based on limited
quantitative surface settlement performance and qualitative observations. The past-performance
driven and degree of understanding modification of LRFD Resistance Factors appears to work
well for the projects reviewed. However, the research team evaluated dozens of unique or
innovative approaches to shallow foundation design that are being utilized in other states and
nations that can improve the process.
© ASCE
Geo-Congress 2020
Several field and laboratory tests have been developed in states with geology similar to South
Dakota, wherein conventional field and laboratory tests are insufficient due to the challenging
nature of the materials. The following three field or laboratory tests are suggested to SDDOT to
incorporate into their shallow foundation design to reduce uncertainty.
 Pressuremeter testing has been shown to be a rapid and reliable means to obtain design
parameters for foundation design in hard soils, IGMs, and soft rock (FHWA 2002).
 Geophysical tests have also been demonstrated (FHWA 2002) to be fast and reliable
means to characterize sites with the geologic materials common to South Dakota.
Geophysical tests can reliably profile rock or be used to correlate to design parameters.
 Simple Shear Testing is beginning to be used by several states to obtain friction angles
and cohesion in a more robust manner than the conventional testing via Unconfined
Compression or Direct Shear testing and is often faster than Triaxial Tests.
Part of a past performance driven approach is to know both quantitatively and qualitatively
how close a design is to a failure condition (i.e. the limit states). In past performance monitoring
of previous construction, it is very unlikely that SDDOT (or any DOT) purposefully allowed any
structure or system to reach an ULS or even SLS. Therefore, the database is likely to be heavily
skewed to the conservative side. A database skewed is a database that is difficult to develop
statistical relationships for. Thus, there is a need to couple the field data with full scale load tests
that show where failure states actually are for the geologic conditions in South Dakota. Full scale
load tests are much more common in deep foundations, and AASHTO LRFD (2017) allows for
more aggressive Resistance Factors for sites/regions with full scale load testing. AASHTO
allows these higher Resistance Factors as full-scale load tests prove the failure condition and
reduce the uncertainties that drive the need for the Resistance Factors. If regional transportation
agencies and/or private owners were to perform a small number of full-scale shallow foundation
load tests on well characterized, typical Dakota geologic materials, the uncertainties could be
reducing, and higher Resistance Factors can then be recommended for use in the state and
surrounding states.
DEEP FOUNDATIONS
Data gathered during the SLTs of driven steel pile foundations include head and tip
settlement, and load-in-pile measurements from strain gauges located at various depths along the
length of the piles. The effectiveness of LRFD use in the South Dakota focuses on t-z model
(Misra et al. 2007; Misra and Roberts 2006) parameters for the driven piles from the evaluation
of the SLTs. The t-z method is widely used for prediction of the load-displacement transfer
relationship for piles subjected to axial load due to its simplicity and capability of incorporating
non-linear soil behavior. Load-settlement curve fitting between the measured data and the
theoretical t-z model were conducted to back-calculate the soil-pile interfacial and tip
parameters. Once computed, the t-z model parameters are compared to side and tip resistance
values derived from Case Pile Wave Analysis Program (CAPWAP®) analyses conducted on
restrike pile driving data. The values of the side and tip resistance parameters from both the t-z
method and CAPWAP analysis are compared across the test sites. This in turn is compiled into a
database, analyzed for reliability, and compared to published general LRFD Resistance Factors.
The t-z method models the soil-pile interface resistance as a series of non-linear springs. The
interface properties are given in terms of the shear modulus of soil-pile interface sub-grade
reaction, K, the ultimate soil-pile interface shear strength, τu, the modulus of tip soil sub-grade
reaction, Es, and the unit tip bearing resistance, qt. In the case of a pile in compressive loading,
© ASCE
Geo-Congress 2020
the springs displace depending on the interface properties and begin to yield as load increases.
Spring yielding initiates from the top of the pile and progresses downward. As the soil-pile
interface springs progressively yield, the tip carries additional load until shear failure occurs.
Table 3. Field load test site descriptions
H-Pile
Site Soil Conditions Depth to Top of Shale
Type
Corson 12x74 Loose silt-sand, dense gravel with weathered shale (Shale) 4.5-6.1 m
Lincoln 12x74 Brown clay-silt and sand-clay (Glacial Till) -

Fall River 10x57 Silt-Clay with Sand, Silt Clay (Shale) 7.6-9.1 m
Spink 10x42 Silt-Clay with Sand (Glacial Till) -
Minnehaha 10x42 Silt-Sand, Silt-Clay with Sand (Glacial Till) -
Buffalo 10x42 Silt-Clay, Clay-Sand, Silt-Clay (Shale) 6.1-7.6 m
Brookings 14x73 Clay-Sand, Sand-Clay, Silt-Clay (Glacial Till) -
Meade 12x53 Sand-Gravel, Silt-Clay (Siltstone) -
Figure 2. Dynamic Side and Tip Resistance Values from CAPWAP Restrike.
To characterize the four t-z parameters, back-calculations were performed by curve fitting
between the measured load-settlement and load transfer curves from field load tests and
theoretical curves from the t-z method. The curve fitting process was initiated by dividing the
soil layers based on boring logs at each site and unit side resistance values calculated between
strain gages. The properties τu and qt, were initially estimated based on unit resistance values
© ASCE
Geo-Congress 2020
determined from strain gage values, CAPWAP results, and/or telltale measurements converted to
side and tip resistance values (Fellenius 1980). The properties for K and Es were initially
estimated as a first approximation. Subsequently, the values of all parameters are adjusted using
a trial and error method until both the measured and theoretical curves show the same trend.
Results of the static and dynamic test results were used to draw a comparison between sites
with different geologic characteristics specifically sites characterized as being either a glacial till
site or a shale site. Table 3 provides a summary of the sites at which side and tip resistance
values were measured with SLT, based on classifying each site as either a glacial till site or a
shale site. A direct comparison of the SLT side and tip resistance values compared based on this
type of classification resulted in a minimal difference in the average unit resistance values
suggesting that if limited information was known about a site, the average value determined
across all sites would provide a reasonable starting point to estimate the pile capacity based on
the driven depth.
Figure 3. Static Side and Tip Resistance Values from Static Load Tests.
Evaluating the unit side and tip resistance results based on a soil layer basis changes the
average computed side resistance values drastically. Figures 2 and 3 present the side and tip
resistances for tested piles as a function of geologic material and test type (Dynamic versus static
load tests) at depth. These results show that pile capacity varies considerably in all materials.
Recall from Figure 1 that the wider the Probability Distribution Function, the more the nominal
Resistance needs to be adjusted. The variability in Figures 2 and 3 exceeds 100 MPa in all
formations, indicating a wide range of possible resistances in the materials for the given
foundation type, showing that LRFD Resistance Factors for design without load testing
presented in Table 1 for drilled shafts are likely appropriate to be applied for driven piles in the
local IGM materials (shale, siltstone, and glacial till). Capacity data and in-depth analysis are
© ASCE
Geo-Congress 2020
being published separately on the results of the pile load test program performed across South
Dakota.
Table 4 provides a summary of the Coefficient of Variation (COV) in unit side and tip
resistance values for all layers, all glacial till layers, and all shale layers. Siltstone formation pile
resistances had lower variability than other materials, but insufficient data in siltstone is available
for determination of COV. The results in Table 4 indicate that when considering the materials
along the length of the pile, a significant difference in unit side resistance exists between shale
and glacial till soil and IGM layers, with large variation in capacities in each material,
confirming the Resistance Factors shown in Table 1 for IGM materials tabulated for Drilled
Shafts and static design methods can be applied reliably to driven piles in similar materials when
so designed.
Table 4. Summary of Static and Dynamic Field Test Variability on Driven Steel HP piles
Dynamic CAPWAP Static Load Test Strain Static Load Test t-z
Results Average Gage Results Average Model Results
Classification Statistic Resistance Resistance Average Resistance
Side Tip Side Tip Side Tip
All Layers COV 0.72 0.56 0.59 0.36 0.52 0.32
All Shale Layers COV 0.32 0.40 0.09 0.25 0.24 0.23
All Glacial Till Layers COV 0.69 0.80 0.62 0.41 0.57 0.40
All Overburden Layers COV 0.73 - 0.53 - 0.35 -
CONCLUSIONS
Use of LRFD in South Dakota for highway bridge foundations is challenged by the geologic
materials found within the state. Many of these geologic materials are expansive clays and
shales. Others are difficult to sample, characterize and model IGM materials spanning the gap
between “soil” and “rock” in the form of shale, siltstones and glacial tills. In these challenging to
design conditions, engineers have used past-performance driven LRFD design successfully to
engineer bridge foundations. However, improvements to the past-performance selection of
LRFD Resistance Factors can be enhanced by use of additional investigation tools suitable for
IGM materials such as Pressuremeter testing and geophysical tests. Meanwhile for driven pile
deep foundations, the use of AASHTO (2017) LRFD Resistance Factors for IGM materials
derived for drilled shaft deep foundations appears to be appropriate for sites within South Dakota
and surrounding states with similar geologic materials.
REFERENCES
AASHTO (2017). "AASHTO LRFD Bridge Design Specifications, 8th Edition." American
Association of State Highway and Transportation Officials (AASHTO), Washington, D.C.
Fellenius B. H. (1980). "The analysis of results from routine pile load tests." Ground
Engineering, Foundation Publishing Ltd., Vol. 13, No. 6, pp. 19-31.
Fenton, G. A., Zhang, X., and Griffiths, D. V. (2008). "Load and Resistance Factor design of
strip footings." Proc., GeoCongress 2008, American Society of Civil Engineers, 106-113.
Lesny, K., and Paikowsky, S. G. (2011). "Developing a LRFD Procedure for Shallow
Foundations." Proceedings of the 3rd International Symposium on Geotechnical Safety and
Risk, N. Vogt, B. Schuppener, D. Straub, and G. Bräu, eds.Munich, Germany, 47-65.
© ASCE
Geo-Congress 2020
Misra, A., Roberts, L., and Levorson, S. (2007). "Reliability analysis of drilled shaft behavior
using finite difference method and Monte Carlo simulation." Geotechnical and Geological
Engineering, 25(1), 65-77.
Misra, A., and Roberts, L. A. (2006). "Probabilistic analysis of drilled shaft service limit state
using the "t-z" method." Canadian Geotechnical Journal, 43(12), 1324(9).
Munfakh, G., Arman, A., Collin, J. G., Hung, J. C.-J., and Brouillette, R. P. (2001). "Shallow
Foundations Reference Manual." FHWA, Washington, DC.
Paikowsky, S. G., Canniff, M. C., Lesny, K., Kisse, A., Amatya, S., and Muganga, R. (2010).
"LRFD Design and Construction of Shallow Foundations for Highway Bridge Structures."
NCHRP Report 651, Transportation Research Board, Washington, D.C.
Paikowsky, S. G., Fu, S., Amatya, Y., and Canniff, M. C. (2009). "Uncertainty in Shallow
Foundations Settlement Analysis and Its Utilization in SLS Design Specifications." Proc.,
Proceedings of the 17th International Conference on Soil Mechanics and Geotechnical
Engineering, Alexandria, Egypt, 1317 - 1320.
Roberts, L. A., Misra, A., and NeSmith, W. "Load and Resistance Factor Design (LRFD)
Application for Auger Cast-in-Place (ACIP) and Drilled Displacement (DD) Piles."
Contemporary Topics in In Situ Testing, Analysis, and Reliability of Foundations, GSP 186,
Reston, VA., 434-441.
© ASCE
Geo-Congress 2020
On Solid Ground: Preventative and Responsive Geotechnical and Structural Mitigation of

Geologic Hazards Impacting Oil and Gas Production
Charles D. Hubbard, P.E., P.G.1
1
Braun Intertec Corporation, Bloomington, MN. E-mail: chubbard@braunintertec.com
ABSTRACT
Braun Intertec Corporation is one of many geotechnical consulting firms supporting oil and
gas production by helping establish, maintain, and/or restore stability to production pads,
pipelines, and associated infrastructure. This is accomplished through a variety of means that
include site reconnaissance, exploration and material testing, geotechnical instrumentation and
monitoring, analytical modeling, and geotechnical and structural solutions. This paper highlights
work performed on sites both pending and in production. The featured sites lie within badlands
type terrain host to geologic hazards related primarily to slope instability, which have been
mitigated with traditional mass grading, and/or ground improvement, retention, buttressing, and
drainage techniques. The intent of this paper is to: (1) clarify the role of geotechnical consultants
in resource procurement efforts and operational efficiency; (2) describe the components of
effective site characterization and hazard identification/qualification programs; and (3) initiate
dialogue specifically around the inheritance or initiation of geologic hazards.
INTRODUCTION
This paper focuses primarily on failures and lessons learned from oil and gas projects in
western North Dakota. The projects featured herein lie within the same geologic setting whose
formations and failure mechanics remain challenging but are becoming better understood.
Project risk is only partly governed, however, by geologic setting. For these projects, risk is also
influenced by consultant/owner engagement. Permitting, design, construction, and post-
construction operation activities each may require consideration of the presence, significance,
and/or mitigation of geologic hazards imposed both on and by the project. Factors influencing
consultant/owner engagement include owner and consultant experience with the region’s
geology, past project performance, risk awareness, risk assumption, and project funding. The
timing of engagement can be particularly critical – early engagement offers more opportunity for
risk avoidance or mitigation without unfavorable impacts to infrastructure and operations.
CHALLENGES WITH OIL AND GAS PROCUREMENT IN WESTERN ND

The challenges with oil and gas procurement in western North Dakota relate primarily to the
topographic and geologic setting, the remoteness of that setting, and the challenges associated
with establishing and maintaining production sites as well as access, utility, and infrastructure
alignments to/from those sites. Considering their breadth, the topographic relief involved, and
the variable quality of the materials exposed in cuts, reused as fill, and ultimately supporting
infrastructure, these projects can be host to multiple, unfavorable geotechnical risks.
Constrained by limiting right-of-ways, required cuts and fills may need to be established at
gradients sufficiently steep to require structural reinforcement or retention to prove stable
analytically. Access road and pipeline alignments may need to purposely traverse unstable
ground to gain access to production sites, or to limit/balance cut and fill volumes in areas of
extreme topography. Pad surface water containment requirements elevate the potential for
© ASCE
Geo-Congress 2020
surface water to unfavorably impact site stability. Ultimately, “X” may mark the spot, and
owners (along with their consultants) must figure out how to procure resources from that spot
with appreciable risk.
Figure 1 provides context to the project development interests shared by owners and
consultants operating in the Williston Basin Provence, and the challenges associated with
recognizing and characterizing the hazards not only to which the project’s elements may be
exposed but also those that the project’s elements may impose on the site.
Figure 1. Well and battery pads with landslide impacts to primary access road.
There are many opportunities for failure, and as many seem to be a function of site
development as they are a function of in situ geologic conditions. To address and/or mitigate
hazards due to both sources, geotechnical consultants may provide a wide variety of
recommendations for cuts, fills, structural reinforcement and retention, seepage and surface
drainage control, and instrumentation and monitoring. How those recommendations are received
and weighed by owners based on their experience and perception of risk, however, may not align
with consultant assumptions or expectations, leading to an a la carte approach to implementation.
Continuation of consulting services through and beyond construction is also not guaranteed.
An example of a “built” slope failure hazard is shown in Figure 2, which identifies the range
of water sources impacting cuts and fills in the Willison Basin Provence, and suggests how such
sources of water impacted the stability of a fill providing access to a production site. The fill was
placed without geotechnical consultation or drainage considerations, and failed under the
influence of unfavorable seepage forces. The repair, which shut down product collection and
transportation for several months, involved removing the entirety of the failure mass, installing
drainage collection and disposal elements, and rebuilding the fill with free-draining material.
© ASCE
Geo-Congress 2020
Figure 2. Typical sources of water impacting cuts and fills in Williston Basin Provence.
Poor communication of project assumptions/expectations also influences project risk.
Though analytical modeling may demonstrate the need for 1 ¾:1 (Horizontal:Vertical) cut
slopes, permitted right-of-way boundaries may dictate those same cuts initially be made at a 1:1
(H:V) gradient. Though, long-term, provisions may be in place to trim fills back and use the
spoils to buttress and effectively flatten the cuts, in the short-term both cuts and fills may exist at
gradients steeper than advisable. (Buttressed cuts with no internal drainage are another source of
instability.)
Figure 3. Tilton DSU 2 Quad and 3 Quad well pad sites. Landslide Hazard Area in pink.
Many of the sites evaluated in western North Dakota also lie within landslide hazard areas
mapped by the North Dakota Geological Survey (NDGS), which triggers a directive from the
North Dakota Industrial Commission (NDIC) to qualify potentially impacted projects
geotechnically and provider owners with a post-construction monitoring plan. The hazard areas
are mapped primarily from aerial imagery, and so include both confirmable and suspect
© ASCE
Geo-Congress 2020
landslides. Awareness and confidence in the presence and impact of landslide hazards varies
based on the age of the maps, which may not identify recent landslides.
Figure 3 shows two production sites impacted by a mapped landslide hazard area. Figure 4
confirms ground movement occurred in the vicinity of the 2 Quad well pad, though development
was recommended due to lack of evidence that ground movement was recent, and the scope of
the development relative to the landslide feature. Figure 5 provides a view from the 3 Quad well
pad toward the landslide mass, which was determined to lie some distance beyond the pad
footprint, and so the 3 Quad pad, too, was considered eligible for development.
Figure 4. Angled/faulted bedding characteristic of mass historic failure, vic. 2 Quad site.
Figure 5. Historic failure mass atop undisturbed soft rock, vic. 3 Quad site.
ROLE OF THE GEOTECHNICAL CONSULTANT

Geology, and the engineering aspects of that geology, are critical to site and infrastructure
stability and performance. In the absence of site-specific information regarding geologic hazards,
© ASCE
Geo-Congress 2020
the value of geotechnical consulting services – which, including reconnaissance, exploration,

analyses, and construction and post-construction monitoring for a geologically unfavorable site
can be on the order of one hundred thousand dollars – can be underappreciated. A failure
impacting a pad, road, or pipeline, however, may result in a similar order of magnitude revenue
loss on a daily, weekly, or monthly basis. Failure mitigation costs may amount to between five
and ten times this value. The well pad access road shown in Figure 6 failed twice within a period
of only a few years, having been built and then re-built in the absence of geotechnical consulting
services. While it was determined that the re-built embankment was generally well compacted, it
was also confirmed that no provisions had been made to collect and dispose of groundwater
flowing toward and becoming trapped by the embankment fill.
There are numerous opportunities for owner engagement, many of which require limited
effort but potentially yield valuable findings. Engagement can occur by project stage,
commencing with permitting and progressing through design, investigation, construction, and
post-construction operations. We advocate investing as early in the project development process
as possible to confirm the soundness of higher level planning decisions and chart a path forward
that optimizes expenditures and limits risk to the built infrastructure.
Braun Intertec is most often retained after permits have been issued, right-of-way boundaries
have been established, and access and grading plans have been developed. Our requested scope
of services may also be limited to the testing of materials gathered by others for compliance with
particular material or performance specifications, or the inspection and reporting of conditions
exposed during construction or discovered post-construction.
Figure 6. Characterization of an embankment failure after it failed a second time.

It is important to point out that the number of failures we evaluate is small relative to the
number of pads in operation, the number of miles of access road being traveled upon, the number
of feet of pipeline in place, etc. For those engaged with forensic projects, however, the amount of
work devoted to failures can be high, with the results of that work often suggesting the failures in
question could have been avoided with some, or a greater degree of, geotechnical support. Many
owners in the oil and gas industry employ geologists and engineers, and they, like other
© ASCE
Geo-Congress 2020
proposed well pad (green) straddling the limits of a much larger historic landslide. A site
reconnaissance confirmed the aerial imagery interpretation and also revealed recent movement,
allowing the owner to opt out of pad construction.
Figure 8 is an excerpt from a plan summarizing the scope of field work performed to
characterize the failed access road embankment shown in Figure 6. The scope of geotechnical
consulting work for this failure included slope inclinometer casings, survey monuments, and
multiple back-calculation and mitigation analyses that cost the owner much more than a design-
phase evaluation (which was not performed), in both dollars and time. The mitigation work for
this failure, which has not been completed, may exceed the cost of the geotechnical consulting
work by at least one order of magnitude.
Figure 8. Plan of embankment failure with field work highlights and other site features.
We have catalogued shear strength data by formation and lithology. We have analyzed slope
stability using averages, means and standard deviations, and upper and lower bound values. We
remind ourselves that the long-term post-construction condition may not be the most critical for
the project. We may need to design based on a conceptual worst-case failure geometry and/or
subsurface geologic profile when we can’t gather enough data to confirm the actual failure
geometry and geologic profile.
Topographic and right-of-way constraints limited access to the embankment failure shown in
Figures 6 and 8. With borings and a slope inclinometer casing confined to the embankment crest
and immediate shoulder, and with little visual evidence defining the downslope failure limits, an
iterative analytical approach was taken to determine a worst-case failure geometry to which
mitigation alternatives could be “fitted” and designed. The results of our iterations culminated in
the analytical model shown in Figure 9, which shows a layer of alluvium, buried beneath the
embankment fill, whose optimized (but undeterminable) thickness proved quite impactful to our
evaluation of stabilization element properties, particularly in regard to pipeline support (the
pipeline located downslope of the embankment crest and buried within the embankment fill).
Successful mitigation required confidence in the capacity of the stabilization elements, which
© ASCE
Geo-Congress 2020
included tied-back H-piles, drilled shafts, and micropiles.

Figure 9. Worst-case failure geometry. Measured depth of shear marked with red arrow.
We promote risk-based as much as deterministic property assessments. While numerical
modeling and factor-of-safety determinations are typical from a compliance and permitting
perspective, material properties and potential failure mechanisms are often too complex to be
resolved solely from analyses. It is not always possible to satisfactorily replicate historic failures
analytically nor, as mentioned, is it always possible to access and characterize many broad and/or
complex project sites to begin with. Practices including Potential Failure Mode (PFM)
assessments can help us understand both how and why failure occurred, and which of many
possible mitigating design and/or construction elements or processes may prove most
practicable.
We also developed and promote a post-construction inspection form we can use, or direct
owners to use, to track post-construction project performance over time. We have used this form
to help educate and empower owners to recognize and become familiar with geotechnical items
of importance they might not have been aware of in the past.
Stabilization measures designed by our staff for project sites in the Williston Basin Provence
have included excavation and replacement, buttressing, retention, and drainage, often in
combination. While we develop many plans, profiles and details for such measures, proprietary
contractors are also engaged to help owners budget and solicit bids for the work.
Figure 10 shows an anchored H-pile wall designed to stabilize a landslide impacting one of
the access roads shown in Figure 1. The landslide was modest in size and could also be removed
and replaced with a buttress fill. Given the unfavorable terrain surrounding the landslide,
however, the quality of material available for reuse, and a need to drain the buttress fill
internally, our confidence in the performance and longevity of the buttress did not match our
confidence in the retention option despite the qualifications of contractors with experience in the
area (on which we were wary of placing unreasonable performance expectations). Not only did
the retention option not require removal of the landslide mass, it reduced the volume of required
fill through the impacted road segment by approximately 80 percent.
Failures that occur during the course of stabilization work are unfortunate and can be costly
but cannot be dismissed simply as a result of human error or considered of no value to the
experienced practitioner. We are often attempting to mitigate failures perhaps better described as
© ASCE
Geo-Congress 2020
complex systems of interconnected geologic and/or structural elements whose mobilization and
movement are similarly complex. Unfavorable conditions can also reveal event-, time-, or
weather-dependent vulnerabilities not apparent or discovered during a project’s design phase. In
any case, experience, more than anything, reminds us there is always something to learn.
Figure 10. Modeling of tie-back wall supporting access road through landslide.
Figure 11. Retention structure remnant and buttressed fill slope supporting well pad.
Figure 11 shows a well pad that experienced multiple failures, two of which occurred
subsequent to original pad preparation, a third during repair activities, and a fourth subsequent to
repair activities (no geotechnical work was performed to evaluate site conditions prior to
construction). The initial and largest failure impacted the north pad slope (lower half of Figure
© ASCE
Geo-Congress 2020
11) and occurred along the interface between the pad fill and the surface of the ravine into which
the pad fill was placed. Site characterization work suggested the ravine contained landslide
debris that was not removed in advance of filling.
A tied-back H-pile wall was designed and under construction to retain and isolate the pad
from the adjacent, failed slope when rain shut the project down and remobilized the partially
retained failure mass. Excavations to remove the damaged structural elements were ultimately
extended back into the pad to create a “root” for a compacted 2:1 (H:V) fill slope that currently
supports the bulk of the pad. To discourage long-term destabilization of the compacted fill slope,
a culvert was placed within a channel passing perpendicular to the ravine bottom, and the culvert
was buried beneath a substantial amount of additional fill to effectively flatten that portion of the
ravine downslope from the 2:1 compacted fill slope.
It is easy (and natural) to believe our past experience and current tools will always be
sufficient to solve the problem at hand. With experience, however, comes an awareness that
there will always be something we have not seen, something that will render our process
vulnerable to omission and perhaps failure. Applying the “beginner’s mind” to each new
problem, discovering what it is we don’t know about the problem, expands the boundaries
possibility from the start, guiding us to use our intuition and determine if an alternative set of
tools may be needed to solve the problem.
EDUCATION, RELATIONSHIPS, AND RISK MANAGEMENT

To further our profession, geotechnical consultants need to set and uphold expectations not
only for ourselves but also for the project owners and owner-consultants that retain us. We must
strive to come prepared, and be prepared to learn. Every environment we enter, every project we
contract for, and every owner, consultant, and contractor we work with, has something to
broaden our knowledge and experience. To our peers:
 Know your geology and geomorphology.
 Stratigraphic profile
 Geologic origin and depositional environment
 Material properties
 Groundwater and surface water patterns
 Weathering and failure mechanisms
 Put yourself in your clients’ shoes
 Find out what is (and isn’t) important
 Leverage their experience and knowledge
 Find out how things have failed (and have been repaired) in the past
 Features and causes
 Mitigation methods and elements
 Learn your industry’s language
 Materials and procedures
 Quality oversight
 Pause – trust is being extended among parties with different experiences and knowledge
Owners and other owner-consultants should come prepared as well, and geotechnical
engineers can assist with that preparation. Consultants can – and should – also educate, helping
others learn their language, think more critically, and make better decisions on their own.
Educating and empowering owners and other owner-consultants can only help all parties reach
© ASCE
Geo-Congress 2020
the same goal with a shared appreciation for outcomes, risk, and value. To others:
 All project phases are important
 Invest more up front
 Chart the course before taking the plunge
 Identify natural and critical transitions and checkpoints
 Scope and budget each segment separately
 Design reports are guides, not gospel; the design is not complete until the project is built
 Plan on changing the plan
 Review, reflect, respond

 Information in hand at any point can alter next-steps
 Construction rarely fails to surprise
 Pause – trust is being extended among parties with different experiences and knowledge
Collectively, our goal should be to optimize all phases of our project scope so investment is
commensurate with risk and uncertainty. This involves (1) focusing on the design, construction,
and geologic elements driving stability, performance, and risk; (2) having both agreement on and
understanding of those critical elements; and (3) proceeding with confidence that the proper
logistical and technical measures are in place to move our projects toward a successful
conclusion.
Finally, effort is commensurate with risk. No site is the same. There is no such thing as a
“standard” scope. Decisions are as much judgmental as technical. Knowledge and confidence in
the situation at hand should be tempered by the degree of uncertainty associated with that
knowledge, and the complexities of the path ahead. Where the degree of uncertainty is high, and
the risks associated with unfavorable stability and performance are substantial, clarity is critical
to define what should be done, what should not be done, and what the assumed risk is for all.
© ASCE
Geo-Congress 2020

Geo Congress 2020 University of PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Geo Congress 2020 University of PDF

Uploaded by

Copyright:

Available Formats

GEOTECHNICAL SPECIAL PUBLICATION NO.

GEOTECHNICAL ENGINEERING CONFERENCE

February 25–28, 2020

Published by the American Society of Civil Engineers

Published by American Society of Civil Engineers

Photocopies and permissions. Permission to photocopy or reproduce material from ASCE

Errata: Errata, if any, can be found at https://doi.org/10.1061/9780784482841

Copyright © 2020 by the American Society of Civil Engineers.

Joseph Labuz, Ph.D., P.E., F.ASCE, University of Minnesota

Geo-Congress 2020 Conference Program Committee

 Derrick D. Dasenbrock, P.E., D.GE, F.ASCE, Minnesota Department of Transportation

Technical Program Committee

 Roman D. Hryciw, Ph.D., M.ASCE, University of Michigan

 James P. Hambleton, Ph.D., M.ASCE, Northwestern University

Pouyan Asem Ivan Contreras Megan Hoppe

Steve Olson David Saftner Brent Theroux

Lessons (Re)learned from Geotechnical Failures .................................................................. 14

Twenty-Year Performance of a Mixed LLRW/RCRA Waste Disposal Facility................... 29

Gas Explosion Analysis ........................................................................................................... 57

Supporting a Bridge between Countries Case Study: Construction of Baudette

Computational Tools for the Analysis of Stability of Embankments in

Eisenhower Bridge North Abutment and Approach Settlement: A Case History

Washington Park Reservoir Improvements: Accommodating Ancient Landslide

A Retrospective on the Evolution of Geotechnical Sensing and Instrumentation

Experimental Study of Forces Induced in Mechanical Excavation of Rock ....................... 156

An Overview of Performance Monitoring for Drilled Full Displacement Type

Sky Harbor Airport Runway Realignment .......................................................................... 187

Kennedy Bridge Instrumentation: A Pier Review ............................................................... 210

On Solid Ground: Preventative and Responsive Geotechnical and Structural

The Art of Numerical Modeling in Geomechanics

Figure 3. Incremental displacement vectors and contours of horizontal displacement, due

Figure 4. Contours of excess pore pressure at steady-state fluid conditions.

MISCONCEPTIONS, MISTAKES AND MISUNDERSTANDINGS

JOB TITLE : (*10^3)

FLAC (Version 7.00)

Boundary plot 2.500

Contour interval= 2.50E-06

0.500 1.500 2.500 3.500 4.500

JOB TITLE : (*10^3)

FLAC (Version 7.00)

Boundary plot 2.500

Contour interval= 5.00E-07

0.500 1.500 2.500 3.500 4.500

Figure 11. Conditions as Figure 10 but with dilation angle of 20°.

Petroleum Science and Engineering, 38, 177-186. Elsevier.

Lessons (Re)learned from Geotechnical Failures

Figure 1. Self-sinking caisson

Figure 2. Subsurface Conditions

Figure 3. Conditions in caisson after failure

Figure 4. Relative Stiffness of Wished-in-place Caisson

away from it in others to cause corresponding decreases in pressure. Furthermore, development

Figure 5. Conditions during construction

SLOPE AND RETENTION SYSTEM FAILURE

location of the failure are shown in Figure 5.

Figure 6. Subsurface Conditions

Figure 7. Cross Section at Open Cut

Figure 8. Failure of Open Cut and Retention System

cited as a demonstration of adequacy. Despite acknowledging the lack of supporting calculations,

SERVICEABLITY FAILURE OF RETENTION SYSTEM

Figure 9. Excavation Support and Typical Damage in Adjacent Building

Figure 10. Subsurface conditions

Figure 12. Collapse of Type M3 excavation adjacent to Nichol Highway

RETENTION SYSTEM FAILURE