Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/272180342

Tactile Pressure Sensors in Centrifuge Testing

Article  in  Geotechnical Testing Journal · October 2013


DOI: 10.1520/GTJ20120061

CITATIONS READS

21 675

4 authors:

Hesham El Ganainy Anthony Tessari


LANGAN University at Buffalo, The State University of New York
14 PUBLICATIONS   147 CITATIONS    27 PUBLICATIONS   84 CITATIONS   

SEE PROFILE SEE PROFILE

Tarek H. Abdoun Inthuorn Sasanakul


Rensselaer Polytechnic Institute University of South Carolina
208 PUBLICATIONS   2,596 CITATIONS    42 PUBLICATIONS   201 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Dam and road safety project View project

Identification of nonlinear soil properties View project

All content following this page was uploaded by Anthony Tessari on 18 February 2015.

The user has requested enhancement of the downloaded file.


Geotechnical
Testing Journal
Hesham El Ganainy,1 Anthony Tessari,2 Tarek Abdoun,3 and Inthuorn Sasanakul4

DOI: 10.1520/GTJ20120061

Tactile Pressure Sensors in


Centrifuge Testing
VOL. 37 / NO. 1 / JANUARY 2014
Geotechnical Testing Journal

doi:10.1520/GTJ20120061 / Vol. 37 / No. 1 / JANUARY 2014 / available online at www.astm.org

Hesham El Ganainy,1 Anthony Tessari,2 Tarek Abdoun,3 and Inthuorn Sasanakul4

Tactile Pressure Sensors in Centrifuge Testing

Reference
El Ganainy, Hesham, Tessari, Anthony, Abdoun, Tarek, and Sasanakul, Inthuorn, “Tactile
Pressure Sensors in Centrifuge Testing,” Geotechnical Testing Journal, Vol. 37, No. 1, 2014, pp. 151–163,
doi:10.1520/GTJ20120061. ISSN 0149-6115

ABSTRACT
Manuscript received April 30, 2012; Assessing vertical and lateral earth pressure is important in geotechnical and foundation
accepted for publication August 23,
engineering. Vertical stresses are easy to calculate assuming a geostatic stress condition.
2013; published online October 28, 2013.
However, characterizing the lateral earth pressure is primarily based on judgment and
1
Project Engineering Associate, Paul C.
empirical correlations that assume a lateral earth pressure coefficient based on the shear
Rizzo Associates, Inc., 500 Penn Center
Blvd., Penn Center East, Suite 100, strength parameters of the soil. Effects of factors such as overconsolidation, geologic age,
Pittsburgh, PA 15235, United States of and pre-shaking on the lateral earth pressure magnitude and distribution are difficult to
America (Corresponding author), e-mail:
hesham.elganainy@rizzoassoc.com assess using only the empirical correlations found in literature. Direct measurements of the
2
lateral earth pressure in the field, full-scale, and centrifuge models are required to fully
Assistant Professor, Dept. of Civil,
Structural, and Environmental characterize these factors. This paper discusses the use of grid-based tactile pressure
Engineering, University at Buffalo, 212 sensors in centrifuge models to study the effects of overconsolidation and pre-shaking on
Ketter Hall, Buffalo, NY 14260, United
the lateral earth pressure. A new preparation and calibration procedure for the tactile
States of America, e-mail:
atessari@buffalo.edu sensors is discussed in detail. A series of five centrifuge tests were performed using dry and
3 saturated sand in normally consolidated, overconsolidated, and pre-shaken deposits. The
Iovino Chair Professor, Dept. of Civil and
Environmental Engineering, Rensselaer measured lateral earth pressure distributions are plotted and evaluated. Precautions and
Polytechnic Institute, 110 8th St., JEC recommendations for the use of tactile sensors in centrifuge experiments are given.
4049, Troy, NY 12180, United States of
America, e-mail: abdout@rpi.edu
4
Research Assistant Professor, Dept. of
Keywords
Civil and Environmental Engineering, tactile pressure sensors, centrifuge modeling, lateral earth pressure coefficient
Rensselaer Polytechnic Institute, 110 8th
St., JEC 4049, Troy, NY 12180, United
States of America,
e-mail: sasani@rpi.edu Introduction
Vertical and lateral effective stresses play a major role in characterizing the stiffness and strength
of soil deposits. These are produced by the self-weight of the soil in addition to any external loads
that act upon it. The vertical effective stresses are simple to characterize under the assumption of a
geostatic stress condition for the soil mass and a homogenous isotropic half-space assumption for
the stresses because of external loads. Conversely, the evaluation of lateral effective earth pressure

Copyright V
C 2014 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. 151
152 Geotechnical Testing Journal

because of the soil weight usually involves the introduction of a Unlike their load cells counterparts, tactile pressure sensors
lateral earth pressure coefficient, K, defined as the ratio of the have the advantage of being thin, flexible and capable of meas-
lateral to vertical effective stress at any depth. The K value uring normal pressures over a large surface area. Several chal-
depends on various factors such as: soil type, relative density, lenges arise when considering their use in geotechnical
geologic age, overconsolidation, pre-shaking, and the lateral applications and these include sensor calibration, measuring
constraint condition (Jaky 1944; Terzaghi and Peck 1948; stresses in saturated soil deposits, and the effect of shear stresses
Lambe and Whitman 1969; Mayne and Kulhawy 1982; Kramer on the accuracy of the measured normal stress. These issues
1996; Sadrekarimi 2011). have previously impeded their use as reliable sensors in geotech-
Researchers and practicing engineers may calculate lateral nical physical testing.
earth pressures either from sophisticated numerical models or The objective of this paper is to discuss the reliability of
simple analytical solutions (Jaky 1944; Whittle et al. 1993; using tactile pressure sensors in centrifuge models and to intro-
Gourvenec and Powrie 1999; Nikolinakou et al. 2011). Numeri- duce a consistent preparation and calibration procedure for
cal models, such as those based on finite elements or other com- them. Using the proposed approach, the effect of overconsolida-
plex analyses, are expensive and require specialized experience. tion and pre-shaking on the lateral earth pressure will be
This limits their use to research and high-end applications. The investigated.
merit of simplicity that analytical solutions possess makes them
the most favorable approach to practicing engineers.
Analytical solutions are predominantly based on an
Sensor Description
assumption that the lateral constraint condition for the soil stra- Tactile sensors consist of two thin, flexible polyester sheets that
tum is at rest, active, or in a passive state. This problem has contain either rows or columns of resistive ink. The sheets are
been extensively studied in the past either mathematically or joined together such that the superimposed ink traces form a
empirically and many simple analytical equations for the lateral grid. The intersection of the rows and columns produces sens-
earth pressure coefficient can be found in common geotechnical ing cells, or “sensels.” The sensors operate by measuring
engineering textbooks (e.g., Terzaghi and Peck 1948). These sol- changes in the electrical resistance at the grid intersections
utions mainly evaluate the K value utilizing a soil shear strength upon application of load, which are linked to the amount of
parameter, e.g., the phi angle (Jaky 1944). Factors such as over- applied pressure through a calibration process (Palmer et al.
consolidation and pre-shaking cannot be precisely accounted 2006). Tactile pressure sensors are commercially available from
for using these simple expressions. various manufacturers (e.g., Pressure Profile Systems, Inc., Sen-
Direct measurements of the lateral earth pressure in the sor Products, Inc., and Tekscan, Inc.) in many sizes, shapes, and
field, full-scale, and centrifuge experiments permit engineers pressure ratings. Tactile pressure sensors manufactured by
and researchers to identify different elements affecting the K Tekscan were used in this study.
value and the distribution of the lateral earth pressure. More- The electrical resistance between the internal traces varies
over, this approach does not require simplifying mathematical when pressure is applied to the sensor. This resistance change is
assumptions, which may not correctly simulate field conditions, measured and converted to an 8-bit number that is transmitted
to obtain a closed form solution for the problem (Michalowski to a data acquisition device. Proprietary software converts the
2005). number to a pressure value based on the sensor calibration.
Stresses in soils have been conventionally measured using Finally, the resulting array of numbers is transformed into a
load cells. This has always been a challenge because the physical colorized distribution of pressure (Palmer et al. 2006). Selection
presence of a load cell in the soil affects the measured stresses, of a sensor should not be based solely on its physical dimen-
stimulates soil arching and creates a non-uniform stress field sions. Both the maximum anticipated pressure and the mini-
around the cell that will likely skew the measured stresses mum measureable pressure increment should be identified.
(Weiler and Kulhawy 1982; Selig 1989; Paikowsky and Hajduk Saving the pressure value as an 8-bit number results in 256 fi-
1997; Dave and Dasaka 2011). Additionally, load cells only pro- nite divisions for the entire range. Thus the minimum change
vide a localized pressure measurement. Determining the pres- in pressure of an individual sensel, which can be recognized by
sure distribution inside the soil requires the use of a dense array the system, is 1/256 of its pressure range. This is defined as the
of load cells, which is not feasible in centrifuge models. code-width and is of particular importance when selecting sen-
Tactile pressure sensors, originally developed for the analy- sors for measuring events with small or dynamic variations. To
sis of dental occlusions (Podoloff and Benjamin 1989), are cur- allow more flexibility in selecting sensors, the gain of the system
rently being used by various geotechnical researchers in may be modified by a factor of three. For example, if using a
measuring soil contact stresses (Paikowsky and Hajduk 1997; sensor rated for 100 kPa, the gain can be increased such that the
Paikowsky et al. 2000, 2003, 2006; Stith 2005; Palmer et al. sensor saturates at approximately 33 kPa. Conversely, the gain
2006, 2009; Ha et al. 2008; Tessari et al. 2010; El Ganainy 2012). can be reduced so that the maximum measurable pressure is
EL GANAINY ET AL. ON TACTILE PRESSURE SENSORS 153

300 kPa. The gain change will propagate through all other sen- magnitude of drift and hysteresis in the sensor readings and
sors attached to the system and will negate previous calibra- improves the repeatability of the results. Traditionally, the con-
tions. Furthermore, the code-width will also change with ditioning load is applied as a uniform pressure by either pneu-
respect to the new maximum pressure. Figure 1 shows the com- matic or hydrostatic devices.
ponents of the tactile pressure sensors measurement system. To compensate for the differences in sensitivity between
sensels because of manufacturing tolerances or repeated use, an
Use in Centrifuge Testing: equilibration step is performed. Equilibration focuses on the
scaling of the individual sensels and is achieved by summing
Feasibility and Limitations their readings and dividing by the total number of sensels. This
In the past, tactile sensors have been used in various applica- value is subtracted from each sensel to recognize those that are
tions within the automotive, robotics, and medical industries more or less sensitive (“hot” or “cold”). A correction factor is
(Luo et al. 1998; Sumiya et al. 1998; Girão et al. 2007). In geo- then applied so that the average reading is constant throughout
technical engineering applications they have been used in meas- the sensor. Equal and even application of pressure is very im-
uring soil pressures around pipelines crossing fault line (Ha portant for equilibration. Observations suggest that the slightest
et al. 2008; Palmer et al. 2009), contact stresses under rigid foot- variation of pressure on the sensor can produce compounding
ings (Paikowsky et al. 2000), and beneath a pile of sand error during equilibration. That is, slight variations in the equil-
(Paikowsky et al. 2003, 2006, 2010), and the vertical stresses ibration pressure can cause hot sensels to be equilibrated as
under railroad tracks (Stith 2005). cold. Therefore, a uniform pressure applied through a pneu-
In the past two decades, the use of centrifuge experiments matic device to the entire active area of the sensor is recom-
in testing soil-structure systems has increased extensively mended for equilibration. The software allows equilibration at a
because of its feasibility, reliability, and low cost compared to maximum of five points. However, Tekscan, Inc. recommends
full-scale testing. Tactile sensors have been used in many equilibrating the sensors at only three loading levels: 20 %,
research projects at the Geotechnical Centrifuge at Rensselaer 50 %, and 80 % of the sensor’s pressure range.
Polytechnic Institute (Ha et al. 2008). Because the sensors are Finishing the conditioning and equilibration steps, the sen-
not waterproof, the use of tactile sensors in reliably measuring sor is ready for calibration. Calibration focuses on the force-to-
stresses in saturated sands has been a challenge. In the next sec- resistance relationship for the entire sensor. In this step, a uni-
tions a new approach for waterproofing, preparing and calibrat- form pressure is applied to the sensor causing changes in the re-
ing the sensors will be introduced. The sensors will be then used sistance of the loaded sensels. The analog reading from the
to measure the lateral pressure of water and soils in centrifuge sensels is converted to a digital value, referred to as raw sensor
models. The results will be compared with the theoretical solu- data units. This value is then correlated to engineering units
tions found in literature to validate the proposed approach. based on the magnitude of the applied pressure through the
software.
Conventional Sensor Preparation Sensors are typically calibrated using a one-load or two-
load calibration scheme (Tekscan, Inc. 2003). During the one-
and Calibration Procedure load calibration, it is assumed that the sensor has zero output
Tactile pressure sensors should be prepared via a three-step under no load. A known load is then applied to the sensor to
procedure (Tekscan, Inc. 2003). The first step is to condition obtain a single calibration point. The calibration line is obtained
the sensor through a load-unload cycle, three to five times, to a by connecting the zero point to the calibration point on a sensor
pressure level of 120 % of its maximum rating. This reduces the output versus load graph. A two-load calibration uses an initial

FIG. 1
Tactile pressure sensors measurement system.
154 Geotechnical Testing Journal

50 %–60 % of the expected values). This outcome highlights the


FIG. 2 Lateral hydrostatic pressure measured using tactile sensors calibrated importance of choosing the right equipment and procedure for
with an air bladder.
calibrating the sensors. Paikowsky and Hajduk (1997) have
developed a calibration device that applies pressure to the sen-
sor through a granular material that can be either soil or glass
beads. Although using this device in calibrating the sensors
showed good results, it has not been standardized yet. Further-
more, its use requires independent manufacturing because the
sensor distributing companies do not produce it.

Proposed Sensor Preparation and


Inflight Calibration Procedures
As discussed earlier, the preparation procedure for tactile sen-
sors is made up of three steps: conditioning, equilibration, and
calibration. Conditioning and equilibrating are performed by
applying uniform pressure over the active sensor area via a
pneumatic pump system. In this study, a rubber air bladder was
used to apply pressure. Calibrating the sensors should be done
using the same material that the sensor will be used to measure
load and a second higher load. The calibration points are then its stresses during testing. This is to account for the material
connected using a power law equation (Palmer et al. 2009). compliance at the interface (Tekscan, Inc. 2003). In geotechnical
Tekscan, Inc. recommends calibrating the sensors using the applications, tactile sensors are used mainly to measure soil
two-load calibration scheme at 20 % and 80 % of the sensor’s pressures; thus, calibrating the sensors should be done by pres-
pressure range. surizing the sensors through a layer of soil at the interface.
Sensor calibration is a challenging task. Figures 2 and 3 The proposed approach is to calibrate the sensors inflight
show the lateral pressure of water and dry Nevada sand deposits and this can be achieved as follows. If it is necessary to measure
at relative densities of 45 % and 75 % and OCR ¼ 1.0. The data the pressure of a given material, the sensor may be calibrated by
was measured in two centrifuge models using tactile sensors placing it horizontally at the bottom of a model box and loading
that were calibrated using the air bladder shown in Fig. 1. As can it with a known height of the same material at a constant mass
be seen in the figures, the measured pressures are considerably density. Spinning this model to higher g-levels, the sensor can
lower than the theoretical lateral pressures (approximately be calibrated by calculating the vertical load that acts upon it at
each specific g-level. Knowing the mass density of the overlying
material and its height at 1 g, the target calibration g-levels can
be back calculated. Calibrated sensors can be then used to mea-
FIG. 3 Lateral earth pressure of nevada sand measured using tactile sensors
calibrated with an air bladder. sure the contact pressure of the same material either in the free
field, against other surfaces, or at any lateral constraint
condition.

Centrifuge Testing Program


The experimental program presented herein consists of five cen-
trifuge experiments that were designed to address the proposed
sensors preparation procedure, refine the calibration, measure
lateral pressure of Ottawa and Nevada sand, and to investigate
the effect of overconsolidation and pre-shaking on the meas-
ured pressures. Table 1 summarizes the experiments performed,
highlights the different parameters tested, and lists the major
purpose of each experiment. In the subsequent sections, these
experiments will be dissected and discussed in detail. Table 2
gives the basic properties of Ottawa and Nevada sand used in
the experiments.
EL GANAINY ET AL. ON TACTILE PRESSURE SENSORS 155

TABLE 1 Lateral stresses measurements centrifuge testing program.

Test Sand OCR Pore fluid Model boxa amax (g) Purpose
b
T1 Ottawa/Nevada 1 Dry test Narrow rigid box No shaking Calibrating sensors using known soil
Dr ¼ 75 % Weight (horizontal sensors)
c
T2 Ottawa 2 Dry test Wide rigid box No shaking Measuring water and soil lateral stresses
Dr ¼ 75 % (Vertical sensors)

T3d Ottawa 4 25cp 1-D Large 10 Cycles 0.01 g Measure soil lateral stresses
Dr ¼ 75 % Viscous fluid Laminar box 20 Cycles 0.15 g (Vertical sensors)

T4 Nevada 1 Dry test Narrow Rigid box No shaking Measuring soil lateral stresses
Dr ¼ 75 % (Vertical sensors)

T5 Nevada 1 Dry test 1-D Large No shaking Measuring soil lateral stresses
Dr ¼ 75 % Laminar box (Vertical sensors)

Note: In all tests: measurements were recorded at 25 and 50 g; Dr (%), relative density of the soil deposit; OCR, overconsolidation ratio; amax (g), prototype peak
input acceleration at model base; f, prototype frequency of input motion ¼ 20 Hz.
a
Rigid boxes dimensions: wide rigid box (length 0.91 m  width 0.61 m  height 0.36 m); narrow rigid box (length 0.88 m  width 0.39 m  height 0.36 m).
b
Two calibration models were built; one with Ottawa sand and one with Nevada sand.
c
Prior to building the soil model, the box was filled with only water and spun to 50 g. The lateral pressure was measured at 25 and 50 g, and compared with the
lateral hydrostatic pressure for calibration validation.
d
Test was done in collaboration with Lucas (2010).

Sensors Lamination a successful and reliable lamination, the sensor should have
vents to release this air.
Tactile sensors are water-resistant but not waterproof and the Tekscan sensors are either internally or externally vented.
joint adhesive will rapidly dissolve when submerged. Thus using Internally vented sensors have a path for the entrapped air to go
them, without waterproofing, to measure the contact pressure through the sensor handle that can be eventually vented by
in saturated sands can result in the destruction of the sensor punching a small hole, as shown in Fig. 5-1. Externally vented
and will definitely yield unreliable results. To solve this prob- sensors have micro tubes along the side of the sensor that act as
lem, the sensors are laminated with clear plastic sheets to make air vents. Laminating these sensors will close the vents. To
them waterproof. Two Teflon sheets are then glued to the faces extend these vents outside of the lamination, two shrink tubes
of the laminated sensors to reduce friction with the soil at the can be attached to the vents, as shown in Fig. 5-2. Proper venti-
interface. lation is extremely important because the sensors cannot oper-
Tactile sensors are made of thin polyester sheets; thus, they ate as a sealed gauge.
cannot withstand high temperatures. The majority of commer- Figures 6 and 7 show the lamination procedure for inter-
cial lamination devices use heat to bond the laminate material. nally and externally vented sensors, respectively. In the current
Therefore, a cold adhesive lamination process is used instead.
Figure 4 shows the tools used in the lamination process. During
lamination, the sensor passes between the drums of the laminat-
FIG. 4 Cold lamination process tools.
ing machine forcing any entrapped air between the two polyes-
ter sheets, constituting the sensor, to be squeezed out. To ensure

TABLE 2 Basic properties of Ottawa and Nevada sand used in the


centrifuge testing program.

Property Ottawa sand Nevada sand

D10 (mm) 0.155 0.09


D50 (mm) 0.258 0.15
Specific gravity, Gs 2.665 2.67
Minimum void ratio, emin 0.61 0.511
Maximum void ratio, emax 0.80 0.887
Minimum dry unit weight, cmin (kN/m3) 14.75 14.85
Maximum dry unit weight, cmax (kN/m3) 16.53 16.77
156 Geotechnical Testing Journal

study, tactile pressure sensor model No. 5250 was used. These
FIG. 5 (1) Internally vented sensors; and (2) externally vented sensors. are internally vented and consist of two 0.0001-m thick polyes-
ter sheets with a square active sensing grid of 0.2459 m length
and a pressure range of 25 psi (172 kPa) (Tekscan, Inc. 2003).
Two sensors were laminated, conditioned, and equilibrated
using a rubber air bladder. Two Teflon sheets were then glued
to the sensor faces using a thin layer of vacuum grease. Figure 8
shows the tactile sensor after preparation.

Sensor Calibration Model


A rigid container, of length 0.88 m, width 0.39 m, and height
0.36 m, was used for the calibration model. Load was applied to
the sensor through the self-weight of a uniform layer of dense
Ottawa sand (test T1). A 1 in. (0.0254 m) thick layer of sand
was placed inside the box by dry pluviation and evenly com-
pacted to a relative density of 75 %. The sensor was placed hori-
zontally, as shown in Figs. 9-1. Finally, a layer of 2 in.
(0.0508 m) thick sand was pluviated over the sensor and evenly
compacted. The thickness of the top layer was limited to 2 in.
(0.0508 m) because of the limitation of the sensor handle length
to enable placing the acquisition handle on the sand surface
away from the sensor. The acquisition handle was then attached
to the sensor and placed on the sand surface beside it. Figure 9-2
shows the calibration model ready for spin up.
The sensor was calibrated at 30 and 60 g using the two-
point calibration scheme. These correspond to vertical stresses
of about 21 and 48 kPa, respectively, and were chosen to be rep-
resentative of the range of maximum stresses that the sensor is

FIG. 6
Lamination procedure for internally vented sensors: (1) feed
a lamination sheet; (2) peel the laminate back and pull to
expose the adhesive sheet; (3) slowly feed the leading edge
of the sensor while rotating the laminator drum; (4)
continue rotating the drum while pulling the adhesive sheet;
(5) after finishing this side, pinch a small hole in the sensor’s
handle; (6) cut a strip of about 1 in. (0.0254 m) from the
lamination; (7) repeat steps 1–4 for the other side of the
sensor; (8) after finishing this side, peel the protective paper
sheet of the lamination; and (9) trim the lamination around
the sensor leaving a 1 in. (0.0254 m) wide strip.
EL GANAINY ET AL. ON TACTILE PRESSURE SENSORS 157

FIG. 7
Lamination procedure for externally vented sensors: (1) tape
two shrink tubes to the external air vents; (2) feed a
lamination sheet, peel the laminate back and pull to expose
the adhesive sheet; (3) slowly feed the leading edge of the
sensor while rotating the laminator drum and pulling the
adhesive sheet; (4) cut a strip of about 1 in. (0.0254 m) from
the lamination; (5) repeat steps 2 and 3 for the other side of
the sensor then peel the protective paper sheet; and (6) trim
the lamination around the sensor leaving a 1 in. (0.0254 m)
wide strip.

expected to measure during testing (upper bound of approxi- static pressure at two different g levels: 25 and 50 g. The com-
mately 103 kPa). parison shows excellent agreement between the measured and
theoretical lateral hydrostatic pressure distribution with model
depth.
Measuring Lateral Hydrostatic
Pressure Measuring the at-Rest Lateral Earth
The calibrated sensor was first used to measure the lateral
Pressure Coefficient of Ottawa
hydrostatic pressure by filling the model box with only water,
while fastening it to one of the walls using duct tape (test T2). Sand
The box was filled until the water level coincided with the top- The calibrated sensor is now ready for use in a soil model. The
most row of sensels. This is to measure the lateral hydrostatic at rest lateral earth pressure coefficient, Ko, of an Ottawa sand
pressure distribution starting with a zero value at the top of the deposit dry pluviated at 75 % relative density, was measured in
sensor. This experiment was essential for two reasons: (1) to a centrifuge experiment (test T2). The model was built in a wide
validate the proposed approach of calibrating sensors in flight, rigid container of length 0.91 m, width 0.61 m, and height
because the comparison with the theoretical lateral hydrostatic 0.36 m. The sensor was taped to a thin flexible aluminum plate
pressure will involve no assumption for the lateral pressure at its top edge, using duct tape, and then clamped in the middle
coefficient, and (2) to validate the assumption that sensors cali- of a wide rigid box to eliminate boundary effects and to simulate
brated with dry sand can be used to measure the contact stresses the at rest condition. The purpose of the aluminum plate is to
in saturated sand. This assumption need to be verified because hold the sensor completely vertical while pluviating and com-
of the practical difficulties associated with placing the acquisi- pacting the sand around it. The sand was placed in 1 in.
tion handle in calibration models where water or saturated sand (0.0254 m) thick lifts and compacted to the target relative den-
is used. Figure 10 shows the measured and theoretical hydro- sity. The sensor was temporarily detached and the aluminum
plate was slowly pulled up from behind the sensor by 1 in.
FIG. 8 Tekscan tactile sensors model No. 5250 after lamination.
(0.0254 m) for the first two lifts. This was followed by placing
the next sand lift and pulling up the plate from behind by 1 in.
(0.0254 m) every placed sand lift. This two-step process was
continued successively until the sand level coincided with the
top row of sensels. Again, this is to measure the lateral earth
pressure distribution starting with a zero value at the top of the
sensor. Figure 11-1 shows the experimental setup of test T2 and
Fig. 11-2 shows the completed model ready for spin up.
The model was spun to a maximum g-level of 50 g. Read-
ings of the lateral soil pressure were taken at 25 and 50 g. The
model was then spun down to 1 g and then re-spun again to
25 g to assess the effect of overconsolidation on the lateral earth
pressure coefficient. Figure 12 shows the measured distributions
158 Geotechnical Testing Journal

FIG. 9
Calibration model: (1) tactile sensor placed over 1 in.
(0.0254 m) of sand; and (2) tactile sensor placed
sandwiched in 3 in. (0.0762 m) of sand ready for spin up.

of the lateral earth pressure with depth for the normally and vertical and lateral straining. This Ko value compares well with
over-consolidated soil deposits at 25 and 50 g. the simplified Jaky’s formula for the lateral earth pressure coeffi-
Fitting the measured lateral earth pressure distribution cient at rest of similar normally consolidated granular soil
using a linear regression analysis, the at rest lateral earth pres- deposits (Jaky 1948; Schmidt 1966; Michalowski 2005), which
sure coefficient of the 75 % relative density normally consoli- simplified in a widely accepted form as:
dated Ottawa sand deposit was found to be very close to 0.5.
(1) Ko ¼ 1  sin /0
Lateral soil pressure measured at both 25 and 50 g showed
excellent agreement with the Ko ¼ 0.5 distribution line except where /0 is the effective angle on internal friction, which was
near the bottom of the sensor where a slight decrease in the lat- determined to be in the range of 30 to 32 degrees for the tested
eral pressures were measured, as shown in Fig. 12. This is a Ottawa sands.
direct consequence of the boundary effect imposed by the bot- Following the same fitting approach for the measured over-
tom of the rigid box on restraining adjacent soil elements from consolidated data, overconsolidating the soil deposit to an OCR
of 2 increased the lateral earth pressure coefficient at rest by
about 12 % to 0.56. Similar to the normally consolidated case,
the measured lateral pressures of the overconsolidated sand fits
FIG. 10 Measured and theoretical hydrostatic pressure at 25 and 50 g. well the Ko ¼ 0.56 distribution line except near the bottom
boundary where a slight decrease in the soil lateral pressures
were measured, as shown in Fig. 12. Although Schnaid and
Houlsby (1991) reported values for Ko in the range between 1.0
and 2.0 for “heavily” overconsolidated deposits, the measured
Ko value of 0.56 for the “lightly” overconsolidated soil deposit of
OCR ¼ 2 agrees well with other empirical correlations that is
applicable for a wide ranges of OCR. Worth (1972) proposed an
empirical correlation for calculating the coefficient of lateral
earth pressure at rest for overconsolidated sand as a function of
OCR and Poisson’s ratio () as follows:
h  i
(2) Ko ¼ ð1  sin /0 ÞOCR  ðOCR  1Þ
1

where Poisson’s ratio () ¼ 0.1 to 0.3 for loose sand, and 0.3 to
0.4 for dense sand. Equation 2 predicts Ko to be about 0.6 for a
dense sand deposit overconsolidated to an OCR of 2. Mayne
and Kulhawy (1982) proposed a correlation between Ko and
OCR irrespective of Poisson’s ratio that builds on Jaky’s simpli-
fied formula as follows:
0
(3) Ko ¼ ð1  sin /0 ÞOCRsin /

which yields a Ko of about 0.7 for overconsolidated sand deposit


of OCR ¼ 2.
EL GANAINY ET AL. ON TACTILE PRESSURE SENSORS 159

FIG. 11
(1) Experimental setup of a tactile sensor during model
preparation; and (2) rigid box model ready for spin up.

Overconsolidation and rings. More detailed information about this container can be
found on the facility website (www.nees.rpi.edu).
Pre-Shaking Effects An Ottawa sand deposit was placed in the laminar box by
To investigate the effects of overconsolidation and pre-shaking dry pluviation at 75 % relative density and saturated with 25 cP
on the lateral earth pressure coefficient, K, of Ottawa sand, a viscous fluid. The tactile sensor was placed against one of the
centrifuge model (test T3) was built in the 1-D large laminar two longitudinal sides of the box, parallel to the direction of
box available at the geotechnical centrifuge facility at Rensselaer motion of the laminar rings. The sensor was first taped to the
Polytechnic Institute. The laminar box is composed of a stack of box side, using duct tape, to restrain it while building the model.
rectangular rings separated by linear roller bearings and As the sand layers were placed, the tape was removed gradually
arranged to permit relative movement in the longitudinal direc- until it was completely gone after finishing building the model.
tion, during base shaking, with minimal friction between the This allowed the laminar rings to move freely during base shak-
ing without being laterally restrained by the sensor. The box
was placed on an inclined table at 2 degrees with the horizontal
(model units) along the direction of movement of the rings and
installed over a base shaking simulating machine mounted on
FIG. 12 Lateral earth pressure distribution for NC and OC dry Ottawa sand. the centrifuge basket. More details about this experiment can be
found in El Ganainy (2012).
The model was first spun to 25 g and then to 100 g for con-
solidation. After, it was spun down to 25 g to simulate the stress
conditions in an overconsolidated soil deposit with an OCR ¼ 4.
The model was then shaken with 10 sinusoidal cycles of 0.01 g
(nondestructive motion) followed by 20 sinusoidal cycles of
0.15 g, both at 2 Hz excitation frequency (all prototype units).
Figure 13 shows the measured effective lateral earth pressure dis-
tributions for the normally consolidated (NC), overconsolidated
(OC), and pre-shaken soil deposits at 25 g, together with the
theoretical linear fit of the data and the lateral earth pressure
coefficient, K, back calculated using a linear regression analysis.
As shown in Fig. 13, the lateral earth pressure coefficient, K,
of the NC saturated Ottawa sand deposit was found to be about
0.5. Overconsolidating the soil deposit to an OCR ¼ 4 increased
the K value by 22 % to K ¼ 0.61. Pre-shaking has similar effect
of increasing the K value. An increase of about 33 % to
K ¼ 0.81, from the pre-shaken K was measured after applying a
strong base shaking of 20 sinusoidal cycles of 0.15 g at 2 Hz ex-
citation frequency.
It should be noted that tactile sensors measure total stresses.
The effective stresses reported herein were estimated by sub-
tracting the hydrostatic pressure from the total stresses. Total
160 Geotechnical Testing Journal

standard procedure. It was then calibrated inflight, in a separate


FIG. 13 Effective lateral earth pressure distribution for NC, OC, and pre- experiment similar to test T1, using Nevada sand as the calibra-
shaken saturated Ottawa sand at 25 g.
tion load.
The main difference between the two tests was the model
box. Test T4 was built in a narrow rigid container of length
0.88 m, width 0.39 m, and height 0.36 m, while test T5 was built
in the 1-D large laminar box. In each of the two experiments,
the sensor was placed in a vertical position at the middle of the
model box parallel to its longitudinal side. The same technique
used in test T2 for placing the sensor while building the model,
was used in tests T4 and T5. Figure 14 shows the experimental
setup of the two tests.
Each model was spun to 50 g and the lateral earth pressure
distribution was recorded. Figures 15 and 16 show the measured
lateral earth pressure distributions for tests T4 and T5, respec-
tively. Shown on the same figures are the theoretical lateral
earth pressure distribution of normally consolidated Nevada
sand at 75 % relative density and a distribution line representing
the lower bound of the measured data at 60 % of the theoretical
stresses reported after shaking were measured after excess pore pressure distribution. A theoretical value of Ko ¼ 0.45 was
pressure built up during shaking has completely dissipated, selected based on direct shear box and triaxial test results con-
which was confirmed by pore pressure measurements. Thus, if ducted at Rensselaer Polytechnic Institute on the same sand
tactile sensors were to be used in applications where pore pres- with the same relative density. Furthermore, it was consistent
sures are unknown (e.g., undrained shearing), pore pressures with data reported in the literature on the same sand by the
should be measured simultaneously to estimate the effective VELACS project team (Krstelj and Prevost 1992; Popescu and
stresses. Prevost 1993, 1995).
Comparing the lateral earth pressure distributions meas-
ured in these two different containers, shown in Figs. 15 and 16,
Boundary Condition Effects on Ko with those measured in the wide rigid container, shown in Fig.
Another set of two experiments were prepared to investigate the 12, yields the following remarks:

effects of the boundary conditions of the model (i.e., the compli- (1) The lateral earth pressure distribution measured in a
ance of the box walls), if any, on the measured lateral soil pres- wide rigid container shows excellent agreement with a
sures (tests T4 and T5). For the purpose of these tests, dry linear distribution, as presented in Fig. 12.
Nevada sand was used for model preparation and dry pluviated (2) For the narrow rigid container, the relative flexibility of
to 75 % relative density. A Tekscan tactile sensor model No. the box combined with the boundaries being closer to
5250 was laminated, conditioned, and equilibrated using the the sensor has a large impact on the measured lateral

FIG. 14
Experimental setup: (1) test T4—narrow rigid box model;
and (2) test T5—1-D large laminar box model: (2.1) 1-D large
laminar box, and (2.2) tactile sensor installation.
EL GANAINY ET AL. ON TACTILE PRESSURE SENSORS 161

laminar rings are allowed to independently deform later-


FIG. 15 Lateral earth pressure distribution for NC dry Nevada sand at 50 g
(test T4—narrow rigid box).
ally without being restrained at the base of the box. As
illustrated in Fig. 16, the stress relief effect extends over
the entire model height.

Conclusions
The proposed preparation procedure of laminating tactile pres-
sure sensors with waterproof plastic sheets combined with Tef-
lon at the interface, proved to be an effective way for producing
waterproof, durable, and reliable sensors suitable for measuring
soil contact stresses in dry and saturated soil deposits. Calibrat-
ing tactile sensors inflight, by pressurizing them with the same
material of the production test, proved to be a promising
approach because it utilizes the same stress conditions and envi-
ronment in which the sensors are going to be used for testing.
Furthermore, sensors calibrated using dry sand can be effec-
tively used in measuring the contact stresses in saturated sand
deposits.
The lateral earth pressure coefficient at rest, Ko, of a nor-
mally consolidated Ottawa sand deposit at 75 % relative density
earth pressure distribution. The lateral deformation of
the model box walls during spin up induces lateral stress was found to be about 0.5. Overconsolidation has the effect of
relief on the sensors, reducing the measured lateral increasing the lateral earth pressure coefficient. An increase of
stresses. This effect is clearly observed in the measured approximately 12 % from the normally consolidated value was
lateral pressures, shown in Fig. 15, especially near the recorded for an overconsolidated Ottawa sand deposit having
ground surface. Near the bottom boundary of the box, an OCR ¼ 2, whereas for an OCR ¼ 4, an increase of about
this stress relief effect diminishes as a result of the walls 22 % was measured. As expected, pre-shaking has the similar
being fixed at the base. Thus, the measured lateral pres- effect and an increase of 33 % from the pre-shaken value was
sures approach the theoretical values near the bottom of recorded after a base shaking of 20 sinusoidal cycles of 0.15 g at
the box. 2 Hz excitation frequency.
(3) The stress relief effect is even more noticeable in the
Finally, the flexibility of the model box used in the centri-
more flexible 1-D large laminar container. This is an
fuge experiment tend to reduce the recorded lateral earth pres-
effect of the container design wherein the individual
sures from the at rest values because of the lateral stress relief
associated with the lateral deformation of the container during
spin up. The recorded lateral earth pressures in flexible contain-
FIG. 16 Lateral earth pressure distribution for NC dry Nevada sand at 50 g ers are more comparable with active lateral earth pressures.
(test T5—1-D large laminar box). Therefore, it is recommended that to achieve the lateral earth
pressure at rest in centrifuge models is to build them in a wide
rigid container to minimize the effects of the boundary condi-
tions. The container flexibility has relatively no effect on the lat-
eral earth pressure where the active or passive earth pressure
will be measured behind retaining walls, around tunnels, and
pipelines crossing active fault lines, etc.

ACKNOWLEDGMENTS
Many individuals have contributed to the work done in this pa-
per, either by advice or by facilitating and helping in the centri-
fuge experimental work. The writers would like to thank the
staff at the Center for Earthquake Engineering Simulation
(CEES) at Rensselaer Polytechnic Institute, Professor Ricardo
Dobry (RPI), Daisy Lucas (RPI), Mauricio Miguel Garcia
Theran (UNCC), and Professor Mark Muszynski (Gonzaga).
162 Geotechnical Testing Journal

References Rock Mechanics Symposium, VGE Publications, Cambridge,


MA, pp. 1171–1177.
Dave, S. M. and Dasaka, S. M., 2011, “A Review on Pressure Paikowsky, S. G., Palmer, C. J., and Rowles, L. E., 2006, “The
Measurement Using Earth Pressure Cell,” Int. J. Earth Sci. Use of Tactile Sensor Technology for Measuring Soil Stress
Eng., Vol. 4, No. 6, pp. 1031–1034. Distribution,” Proceedings of the GeoCongress 2006—Geo-
El Ganainy, H., 2012, “Lateral Spreading and Liquefaction technical Engineering in the Information Technology Age,
Assessment of Cohesionless Soils in the Free Field: An Inves- ASCE, Atlanta, GA.
tigation Using Centrifuge Testing,” Ph.D. dissertation, Paikowsky, S. G., Tien, H. S., and Rolwes, L. E., 2010,
Rensselaer Polytechnic Institute, Troy, NY. “Observing and Measuring Arching in Ideal and Natural
Girão, P. S., Postolache, O., and Pereira, J. M. D., 2007, “Tactile Granular Materials,” Proceedings of the 7th International
Sensors and Their Use in Industrial, Robotic and Medical Conference on Physical Modeling in Geotechnics (ICPMG
Applications,” Proceedings of the IMEKO 20th TC3, 3rd 2010), Zurich, Switzerland, pp. 373–378.
TC16 and 1st TC22 International Conference Cultivating Palmer, M. C., O’Rourke, T. D., Stewart, H. E., O’Rourke, M. J.,
Metrological Knowledge, Merida, Mexico, Nov 27–30, 2007. and Symans, M., 2006, “Large Displacement Soil-Structure
Gourvenec, S. M. and Powrie, W. 1999, “Three Dimensional Fi- Interaction Test Facility for Lifelines,” Proceedings of the 8th
nite Element Analysis of Diaphragm Wall Installation,” Geo- US National Conference on Earthquake Engineering and
technique, Vol. 49, No. 6, pp. 801–823. 100th Anniversary Earthquake Conference Commemorating
Ha, D., Abdoun, T., O’Rourke, M. J., Symans, M. D., O’Rourke, the 1906 San Francisco Earthquake, EERI, San Francisco,
T. D., Palmer, M. C., and Stewart, H. E., 2008, “Centrifuge CA.
Modeling of Earthquake Faulting Effects on Buried HDPE Palmer, M. C., O’Rourke, T. D., Olson, N. A., Abdoun, T., Ha,
Pipelines Crossing Fault Zones,” J. Geotech. Geoenviron. D., and O’Rourke, M. J., 2009, “Tactile Pressure Sensors for
Eng., Vol. 134, No. 10, pp. 1501–1515. Soil-Structure Interaction Assessment,” J. Geotech. Geoen-
Jaky, J., 1944, “The Coefficient of Earth Pressure at Rest,” J. Soc. viron. Eng., Vol. 135, No. 11, pp. 1638–1645.
Hung. Archit. Eng., Vol. 78, No. 22, pp. 355–358. Podoloff, R. M. and Benjamin, M., 1989, “Tactile Sensor for
Jaky, J., 1948, “Pressure in Silos,” Proceedings of the 2nd Interna- Analyzing Dental Occlusion,” SOMA Eng. Hum. Body, Vol.
tional Conference on Soil Mechanics and Foundation Engi- 3, No. 3, pp. 1–6.
neering, Vol. 1, Rotterdam, The Netherlands, pp. 103–107. Popescu, R. and Prevost, J. H., 1993, “Centrifuge Validation of a
Kramer, S. L., 1996, Geotechnical Earthquake Engineering, Pren- Numerical Model for Dynamic Soil Liquefaction,” Soil Dyn.
tice-Hall, Englewood Cliffs, NJ. Earthq. Eng., Vol. 12, pp. 73–90.
Krstelj, I. and Prevost, J. H., 1992, “Dynamic Effects in a Satu- Popescu, R. and Prevost, J. H., 1995, “Reliability Assessment of
rated Layered Soil Deposit: Centrifuge Modeling,” Soil Dyn. Centrifuge Soil Test Results,” Soil Dyn. Earthq. Eng., Vol. 14,
Earthq. Eng., Vol. 11, pp. 485–496. pp. 93–101.
Lambe, T. W. and Whitman, R. V., 1969, Soil Mechanics, John Sadrekarimi, A., 2011, “Seismic Displacement of Broken-Back
Wiley and Sons, New York. Gravity Quay Walls,” J. Waterway, Port, Coastal, Ocean
Lucas, D., 2010, “Effect of Soil Compressibility on Sand Eng., Vol. 137, No. 2, pp. 75–84.
Response to Liquefaction and Lateral Spreading,” M.Sc. the- Schnaid, F. and Houlsby, G. T., 1991, “Measurement of the
sis, Rensselaer Polytechnic Institute, Troy, NY. Properties of Sand by the Cohesion Pressuremeter Test,” Soil
Luo, Z. P., Berglund, L. J., and An, K. N., 1998, “Validation of Mechanics Report Number 113/91, University of Oxford,
F-Scan Pressure Sensor System: A Technical Note,” J. Rehab. Oxford, U.K., 35 p.
Res. Dev., Vol. 35, pp. 186–191. Schmidt B., 1966, “Earth Pressures at Rest Related to Stress His-
Mayne, P. W. and Kulhawy, F. H., 1982, “Ko—OCR Relation- tory,” Can. Geotech. J., Vol. 3, No. 4, pp. 239–242.
ships in Soil,” J. Geotech. Eng. Div., Vol. 108, No. 60, pp. Selig, E. T., 1989, “In Situ Stress Measurements,” Proceedings of
851–872. the Symposium on the State- of-the-Art of Pavement Response
Michalowski, R. L., 2005, “Coefficient of Earth Pressure at Rest,” Monitoring Systems for Roads and Airfields, Report 89-23,
J. Geotech. Geoenviron. Eng., Vol. 131, No. 11, pp. U.S. Army Cold Regions Research and Engineering Labora-
1429–1433. tory, Hanover, NH, pp. 162–168.
Nikolinakou, M., Whittle, A., Savidis, S., and Schran, U., 2011, Springman, S., Nater, P., Chikatamarla, R., and Laue, J., 2002,
“Prediction and Interpretation of the Performance of a Deep “Use of Flexible Tactile Pressure Sensors in Geotechnical
Excavation in Berlin Sand,” J. Geotech. Geoenviron. Eng., Centrifuges,” International Conference on Physical Modeling
Vol. 137, No. 11, pp. 1047–1061. in Geotechnical Engineering, St. John’s, Canada, pp. 113–118.
Paikowsky, S. G. and Hajduk, E. L., 1997, “Calibration and Use Stith, J. C., 2005, “Railroad Track Pressure Measurements at the
of Grid-Based Tactile Pressure Sensors in Granular Materi- Rail/Tie Interface Using Tekscan Sensors,” M.S. thesis, Uni-
al,” Geotech. Testing J., Vol. 20, No. 2, pp. 218–241. versity of Kentucky, Lexington, KY.
Paikowsky, S. G., Palmer, C. J., and Dimillio, A. F., 2000, Sumiya, T., Suzuki, Y., Kasahara, T., and Ogata, H., 1998,
“Visual Observation and Measurement of Aerial Stress Dis- “Sensing Stability and Dynamic Response of the F-Scan in-
tribution under a Rigid Strip Footing,” Geotechnical Special Shoe Sensing System: A Technical Note,” J. Rehab. Res. Dev.,
Publications 94, American Society of Civil Engineers, Reston, Vol. 35, pp. 192–200.
VA, pp. 148–169. Tekscan, Inc., 2003, Tekscan I—Scan Equilibration and Calibra-
Paikowsky, S. G., Rowles, L. E., and Tien, H. S., 2003, tion Practical Suggestions, Tekscan Inc., South Boston, MA.
“Visualization Measurements of Stress around a Trap Door,” Terzaghi, K. and Peck, R. B., 1948, Soil Mechanics in Engineer-
Proceedings of the 12th Pan American Conference and 39th ing Practice, John Wiley and Sons, New York.
EL GANAINY ET AL. ON TACTILE PRESSURE SENSORS 163

Tessari, A., Sasanakul, I., and Abdoun, T., 2010, “Advanced Whittle, A. J., Hashash, Y. M. A., and Whitman, R. V., 1993,
Sensing in Geotechnical Centrifuge Models,” Proceedings of “Analysis of Deep Excavation in Boston,” J. Geotech. Eng.,
the 7th International Conference on Physical Modeling in Geo- Vol. 119, No. 1, pp. 69–90.
technics (ICPMG 2010), Zurich, Switzerland, pp. 395—400. Worth, C. P., 1972, “General Theories of Earth Pressure and
Weiler, W. A. and Kulhawy, F. H., 1982, “Factors Affecting Deformation” Proceedings of the 5th European Conference
Stress Cell Measurements in Soil,” J. Geotech. Eng. Div., Vol. on Soil Mechanics and Foundation Engineering, Vol. 2, Ma-
108, No. 12, pp. 1529–1548. drid, Spain, pp. 33–52.

Copyright by ASTM Int’l (all rights reserved); Thu Jan 2 10:37:47 EDT 2014
Downloaded/printed by
Hesham El Ganainy (Paul C. Rizzo Associates, Inc., Pittsburgh, Pennsylvania, United States)
Pursuant to License Agreement. No further reproduction authorized.
View publication stats

You might also like