26 - in Situ Soil and Rock Tests

Source: GEOTECHNICAL ENGINEERING
26 In Situ Soil and Rock Tests
26.1 OVERVIEW
26.1.1 Probing for Answers

The first attempts at in-situ testing of soils probably involved jabbing the ground
with a pointed stick, and may have been the origin of the phrase, to ‘‘get the
point.’’ The stick eventually became standardized, modified, and improved,
evolving into a cone-tipped rod that became the ‘‘cone penetration test’’ or CPT.
This was further improved by adding a separate sleeve behind the tip to measure
side friction. More recent embellishments include measuring pore water pressure
and using wireless data transmission. It is difficult to comprehend the scientific
advances that are possible for a pointed stick. The multichanneled attack had
developed extensive empirical correlations that relate cone data to soil type and
bearing capacity.
As the cone test was being developed for relatively soft soils, particularly in
Holland, similar procedures were being developed in the U.S. for testing harder
soils. A sampler was devised that is driven into the soil instead of being pushed in.
The sampler and driving effort were standardized to become the Standard
Penetration Test or SPT, which still is the most widely used soil test in the U.S.
An advantage over the cone is that it measures penetration resistance and
simultaneously extracts a soil sample for identification and classification,
although the sample is too disturbed for meaningful strength tests. A
disadvantage is that the sampler must be removed from the boring after each
test, a boring instrument inserted, and the boring advanced by drilling to the next
test depth. A test is conducted by counting the number of hammer blows required
to drive the sampler prescribed distances.
Because both CPT and SPT test results are influenced by combined influences
from shearing, remolding, and compression of soil, other tests have been devised
780 Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

Copyright © 2007 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
In Situ Soil and Rock Tests
In Situ Soil and Rock Tests 781
to try and isolate and measure these separate parameters. For example, modulus
tests might be directly used to predict settlement, and shear tests to predict bearing
capacity and slope stability.
Another type of in-situ testing is nondestructive and substitutes electrical,

acoustic, or microwave (radar) energy for pushing force or driving energy. Many
of these tests were developed in the petroleum exploration industry for subsurface
mapping, and more recently adapted to measure properties such as a dynamic
modulus. They are discussed in Chapter 27.
The reliability of in-situ tests is evaluated by comparisons with more traditional

test methods, and by comparisons with field performance.
26.1.2 Exploration Drilling

At one time water wells were drilled with ‘‘cable tool rigs’’ that alternately lift and
drop a heavy drill stem with a percussion cutter attached. Drill cuttings were
intermittently bailed from the hole. Cable tool rigs are slow and simple, but can
drill through both soil and rock. They still are used for drilling shallow wells,
particularly in less developed areas.
More efficient drilling methods are required for drilling deep wells including
oil wells. Rotary drilling employs a continuous flow of water or a thin mud that
is circulated down the drill stem and carries cuttings back up between the
stem and the perimeter of the hole. There the fluid is directed to a ‘‘mud pit’’
where cuttings settle out and the mud is siphoned off and pumped back down
the hole.
The main component of drilling mud besides water is bentonite, or relatively pure
smectite, because the mix is thixotropic. Then if there is an interruption in drilling,
the mud holds cuttings in suspension instead of letting them settle to the bottom
of the boring and lock up the bit.
Rock drilling usually is performed with a Hughes ‘‘tricone bit,’’ which has three
rollers with teeth that impact the rock. In soft rocks and soils a ‘‘fishtail’’ or
similar hardened, steel bit is adequate to advance the boring. Rock cores are
obtained with a hollow diamond-laced coring bit.
The first step in exploration drilling is to plan and stake out boring
locations and do an elevation survey that will enable cross-sections
to be drawn. Every geotechnical report will have a disclaimer pointing
out that identifications of soils between borings are based on interpola-
tion and can be expected to vary from the interpretations—a matter of
considerable importance where variations are expected, for example in karst
areas.
Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

782 Geotechnical Engineering
26.1.3 Undisturbed Sampling

Soil samples are obtained with the Standard Penetration Test by pounding
a thick-walled sampler into the ground, and can hardly be considered
‘‘undisturbed.’’ An optional sampling method uses a thin-walled steel tube
called a Shelby tube, which is pushed into soil at the bottom of a boring.
This procedure allows laboratory tests to be performed on the soil samples,
including unconfined compression, triaxial, and consolidation tests. Special
‘‘piston samplers’’ draw the soil into the tube with partial vacuum, and ‘‘sleeve
samplers’’ have been developed that encase the soil in a Nylon tube to reduce
side friction.
A measure of sample disturbance is the ‘‘area ratio,’’ defined by Hvorslev (1949)

to represent the ratio of soil displaced to that which is sampled. The area ratio
is the cross-sectional area of the sample divided by the doughnut-shaped area
of the sampler. The larger the area ratio, the larger the degree of sample
disturbance:
ðo:d:Þ2 ði:d:Þ2
Area ratio ¼ ð26:1Þ
ði:d:Þ2
Example 26.1
Calculate area ratios for an STP sampler, and for a nominal 3 in. (75 mm) diameter 16
Gauge Shelby tube having a wall thickness of 0.065 in. (1.65 mm).
Answer:
STP: Area ratio ¼ ð22 1:52 Þ=ð1:5Þ2 ¼ 0:78
Shelby tube: ð32 2:872 Þ=2:872 ¼ 0:093
Each of the two sampling operations, SPT and Shelby tube, has advantages that
depend on the kind of soil, SPT being most useful for evaluating cohesionless
soils, and Shelby tube and laboratory testing most useful for cohesive soils.
Because of the unknown character of soil ahead of a boring, the two procedures
often are performed alternately, typically at 5 ft (1.5 m) depth intervals. A hole is
bored to the first testing depth, either STP or Shelby tube sampling performed, the
hole is drilled out to the next test depth, the alternate method used, and so on
through the depth of the boring. Boring depths are selected on the basis of the
application—deep enough to allow a realistic analysis for consolidation
settlement, foundation bearing capacity, and slope stability, deeper than the
longest anticipated piles or piers, or a sufficient distance into rock to ensure that
the rock is in place and not part of an old rockfall.
Shelby tube samples of sand are compressed during sampling and are not suitable
for laboratory tests, which has led to a continuing dependence on in-situ testing.
Even the relatively tranquil Shelby tube sampling of clay soils does not generate a
truly undisturbed sample because in-situ soil stress is not preserved. The confining

stresses existing in the field can be reinstated in a laboratory triaxial test in the
‘‘Lambe stress path method,’’ but only if the lateral stress that exists in the field is
known.
STP and Shelby tube sampling can be conducted in a boring made with or without
drilling mud. Another option to hold a boring open while avoiding the extra setup
and mud disposal is to use a ‘‘hollow-stemmed auger.’’ The auger has a central
plug that remains in place during drilling; then at the desired sampling or testing
depth the plug is removed and the sampler introduced down the center of the
auger.
26.1.4 Determining the Character of Soil Between

Sampling Depths
Sampling for laboratory tests or SPTs normally includes about 30 percent of a
soil column and leaves the rest for interpolation. It therefore is of critical
importance that the driller carefully examine the remaining 70 percent. An
advantage of some in-situ tests, in particular cone tests, is that they can be
continuous with depth.
Most soil drilling is performed with a helical screw-type ‘‘flight auger.’’ Drillers
whose main objective is to ‘‘make hole’’ (and money) will prefer to rotate the
auger rapidly so that pieces of soil are cut loose at the tip and pushed along up
the auger to emerge at the ground surface. In this case the only way to identify the
depth from which the soil was removed is to stop advancing the auger and wait for
cuttings from the bottom to come up.
Much more satisfactory for geotechnical exploration drilling is to rotate the bit
slowly while advancing the auger so that it literally screws itself into the ground
like a wood screw. The auger then is pulled vertically to shear off the soil and hold
it in place between flights of the auger. The soil then is exposed by cutting away
the surface with a knife, examined, and boundaries determined and depths
measured on the auger. This information is critical for a geotechnical evaluation.
For example, a settlement analysis requires knowing the thickness of the
consolidating clay layers and whether they are sandwiched between sand layers.
When using this procedure the driller must gauge each auger advance so that it
does not exceed the capability of the drilling machine to pull it out.
26.1.5 Groundwater Table

An important step in exploration drilling is measuring the depth to the
groundwater table. This conveniently is done with a line that completes an
electric circuit on contact with free water, and ideally should be done after
completion of drilling and 24 hours later. This cannot be done in mudded holes,
which is another reason to prefer auger-type drilling where possible.

26.2 STANDARD PENETRATION TEST (SPT)
26.2.1 Redefining the SPT

The Standard Penetration Test (ASTM Designation D-1586) uses a standardized
sampler shown in Fig. 26.1, and a driving effort that is standardized using a 140 lb
(63.5 kg) weight dropping 30 in. (0.76 m) for each blow. The number of blows is
counted and recorded for each of three 6 in. (152 mm) increments of advance. The
first count usually is lower than the last two, which are added to give N blow
counts per foot (per 0.3 m).
A ‘‘doughnut hammer’’ is a hollow cylindrical weight that fits over the drill rod
and hits on an expanded section of the rod. The hammer is lifted by looping a
rope two turns around a smooth pulley called a ‘‘cathead’’ that is continually
turning. When the end of the rope is pulled it grabs onto the cathead so that
friction raises the hammer. Then at the moment when the required height is
reached, the rope is forcibly released and the hammer drops. This constitutes one
blow. The drill stem is marked with chalk at 6 in. (152 mm) intervals, so the
number of blows can be determined for each interval. A typical rate is 40 to 60
blows per minute.
The ‘‘safety hammer’’ fits over the exposed end of a drill stem (Fig. 26.2)
and is the current standard. Measurements indicate that whereas with a
rope release the doughnut hammer has an efficiency of about 45 percent,
the safety hammer has an efficiency of about 60 percent. Results obtained
with a doughnut hammer are corrected to N60 on the basis of the ratio of
efficiencies.
Example 26.2
The blow count for 3 driving invervals with a doughnut hammer are 10, 14, and 16. What
is N60?
Answer: Adding the last two blow counts gives N45 ¼ 30 blows per ft (0.3 m). The energy
correction gives (45/60) 30 ¼ 22 blows/ft (0.3 m).
Figure 26.1
Standard
Penetration Test
sampler or ‘‘split
spoon’’ is driven
vertically by
repeatedly
dropping a large
weight.

Figure 26.2
Hammers used to
drive a Standard
Penetration Test
sampler. (Left)
doughnut hammer,
(middle) safety
hammer, (right)
rotating cathead
and rope for lifting
and manual
release. Modern
‘‘trip hammers’’
have an automatic
release
mechanism.
The rope-and-cathead lifting procedure requires operator skill and introduces

variables that can be on the unsafe side. For example, if the weight is prema-
turely released or the rope drags on the cathead, the count is increased and
the soil is reported to be stronger than it actually is. Calibrations are used
to measure hammer efficiency. These variables are eliminated by use of a
‘‘trip hammer’’ that is raised and automatically released when gripping levers
ride up on a trip cone. The efficiency of trip hammers is considerably higher,
and data are provided by the manufacturer. Detailed investigations also will
include a correction for increasing inertia of the drill rod as the test proceeds
deeper.
26.2.2 Overburden Pressure

A second correction that is applied to SPT data involves overburden pressure. The
sensitivity of blow count to overburden pressure was first reported by Gibbs and
Holtz at the U.S. Bureau of Reclamation. Soil confined in a box was loaded
through a steel plate that supported various surcharge loads, and SPTs were
performed through holes in the plate. While the concept was valid, the test
arrangement may have included boundary conditions that exaggerated the
development of horizontal pressure on the sampler and gave correction factors
that are now believed to be unconservative, and factors suggested by Liao and
Whitman (1986) are based on studies by several investigators and are shown
in Fig. 26.3. The recommended correction factors may be calculated from the
following empirical formula:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
CN ¼ ð1=V0 Þ ð26:2Þ

Figure 26.3
Liao-Whitman (L-W) overburden correction to SPT blow count data. The earlier Gibbs-Holtz
correction is shown for comparison and should not be used. The equivalent depth is the sum of the
depth above the groundwater table plus one-half of the depth below the groundwater table, based
on ¼ 125 lb/ft3 (19.6 kN/m3). For a different soil unit weight, multiply depth by the appropriate ratio.
where CN is the correction factor and v0 is the effective overburden stress in

tons/ft2. Alternatively,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
CN ¼ ð1=100V0 Þ ð26:2aÞ
where 0v is in kPa.
In Fig. 26.3 the equivalent depth is based on a soil unit weight of 125 lb/ft3
(19.6 kN/m3). The effective thickness is defined as thickness above a
groundwater table plus one-half of the thickness below a groundwater table.
The Gibbs-Holtz (G-H) corrections are shown for comparison.
Example 26.3
The N data in Example 26.2 were obtained at a depth of 30 ft (9.1 m) and a
groundwater table exists at a depth of 8 ft (2.4 m). What is the blow count corrected for
overburden pressure?
Answer: The hammer correction remains valid and gave N ¼ 22. The effective depth is
8 þ (30 8)/2 ¼ 19 ft [or (2.4 þ (9.1 2.4)/2 ¼ 5.8 m] From Fig. 26.3, CN ¼ 1 and the blow
count remains at 22.
26.2.3 What Does the SPT Measure?

It sometimes is assumed that the resistance to driving the SPT sampler is mainly
from end resistance, but the incremental increase in resistance for successive
sections suggests that friction both on the inside and on the outside of the sampler
is at least equally important. Studies by Schmertmann (1979) indicate that friction
contributes from 45 to 85 percent of the resistance, the percentage being highest in
cohesive soils.
Other investigators have found that the N blow count relates more closely to side
friction measured with the cone than to cone end resistance. The SPT mainly
measures soil-to-steel friction and therefore is affected by lateral confining
stress and surface roughness of the sampler. The measured ‘‘friction’’ is higher for
cohesive than for granular soils, indicating a considerable influence from
adhesion of clay to the sampler. Side friction can be evaluated by subtracting
the second from the third increment, or a more precise measurement can be
obtained by simply twisting the sampler with a torque wrench (Kelley and
Lutenegger, 2004).
26.2.4 Relationships to Bearing Capacity and Settlement

of Shallow Foundations on Sand
Generally the higher the penetration resistance of sand, the higher the friction
angle, since part of the penetration resistance of sand comes from end bearing
that in turn depends on the relative density, and hence the angle of
internal friction. Excess pore water pressure is assumed to be negligible in
sand and coarser-grained soils. An approximation to the friction angle can be
obtained from Fig. 26.4, but it should be emphasized that these values
derive from empirical correlations and are only approximate. Because a
Figure 26.4
Approximate friction angles of sand from SPT data, adapted from data of de Mello (1971).
N values should be corrected to N60 but should not be corrected for overburden pressure.
The equivalent depth incorporates a groundwater correction as in Fig. 26.3.

depth factor is included, N values should not be corrected for overburden

pressure.
The friction angle and unit weight may be used to calculate the bearing capacity
and settlement of shallow foundations on sand according to the principles
discussed in Chapter 23.
Example 26.4
An SPT performed at 10 ft (3 m) depth in sand above the groundwater table gives N ¼ 3,
3, 4. Estimate the bearing capacity of a column footing 2 2 ft (0.6 m) at a depth of
1 ft (0.3 m).
Answer: N ¼ 3 þ 4 ¼ 7. From Fig. 26.6, ¼ 358. From Fig. 26.4 for a square
footing the bearing capacity factors N and Nq are 29 and 57, respectively. Then from
eq. (22.7),
qo ¼ ðB=2ÞN þ DNq ¼ ð125 lb=ft3 2 ft=2Þð29Þ

þ 125ð1Þð57Þ ¼ 3625 þ 7125 ¼ 10,750 lb=ft2 ð515 MPaÞ
With a factor of safety of 3, qa ¼ 3500 lb/ft2 (170 MPa). Multiplying by the base area gives
Qa ¼ 4 ft2 3500 lb=ft2 ¼ 14,000 lb ¼ 7 tons ð62 kNÞ
This is the bearing capacity against shear failure and does not ensure that there will not be
excessive settlement.
Settlement of foundations on sand is discussed in section 16.10.3. Schmertmann’s method is

shown in Fig. 16.17. SPT blow counts for sand may be converted to approximate cone
equivalents with Meyerhoff’s (1956) approximation, qc ¼ 4N, where qc is in tons/ft2.
As both cone and SPT values are affected by overburden pressures, this relationship should
apply to both corrected and uncorrected N values.
Example 26.5
Predict settlement of the footing in the preceding example.
Answer: In Fig. 16.17 for a square base, settlement is
S ¼ 0:25Bp=qc
where B is the base width, p is the foundation pressure, and qc is the cone bearing
value.
Let
qc ¼ 4N ¼ 4ð7Þ ¼ 28 tons=ft2 ð250 kNÞ

S ¼ 0:25ð2 ftÞð7 ton=ft2 Þ=28 ton=ft2 ¼ 0:125 ft ¼ 1:5 in: ð38 mmÞ
The allowable bearing pressure therefore must be reduced if settlement is to be less than
1 in. (25 mm).

26.2.5 Deep Foundations on Sand

Meyerhof (1956, 1976) showed empirical relationships between SPT blow counts
and pile end-bearing capacity and side friction. Representative data are as shown
in Fig. 26.5. It is not mentioned if the N values were corrected, so it must be
assumed that they were not. Depth correction factors in Fig. 26.3 vary at most by
a factor of 2 and probably average close to 1 for most tests.
Most of the data in Fig. 26.5 cluster around the line qp ¼ 4N, where qp is the point
bearing pressure in tons/ft2 (100 kPa). This relationship is exactly the same as that
which relates cone bearing pressure qc to N value, and is consistent with the
concept that the cone is a model pile.
From these and similar observations Meyerhof suggests the following formulas
for end-bearing and for side friction on piles driven in sand:
D
qp ¼ 0:4N 4N ð26:3Þ
B
fs ¼ Nav =50 < 1ton=ft2 ð100kPaÞ ð26:4Þ
2
where qp and fs are end-bearing and side friction stresses in tons/ft (100’s of kPa),
D is depth to the bottom and B is the bottom width, and Nav is the average blow
count for the length of the pile. An H-pile is assumed to have areas between the
flanges filled with soil. Taper can increase side friction by as much as 1.5.
Example 26.6
SPT data for a sand give an average blow count of 18 blows/ft (0.3 m). Estimate bering
capacity for a 20 ft (6.7 m) pile 1 ft (0.3 m) in diameter.
Figure 26.5
Data for
end-bearing
capacity of piles
versus SPT blow
count, summarized
by Meyerhoff
(1976) from many
sources.

Answer: End bearing cannot exceed 4N ¼ 4(18) ¼ 72 tons/ft2 (7 MPa)

qp ¼ 0:4ð18Þð20 ftÞ=ð1 ftÞ ¼ 144 tons=ft2 ð14 MPaÞ so
qp ¼ 72 tons=ft2 ð7 MPaÞ end bearing
Side friction: fs ¼ 18/50 ¼ 0.36 tons/ft251 ton/ft2.
Multiplying by the resective areas,

Q ¼ 72 ton=ft2 ðÞð0:5 ftÞ2 þ 0:36 ton=ft2 ðÞð1 ftÞð20 ftÞ ¼ 56:5 þ 7:2
¼ 63 tons ð567 kNÞ
With a factor of safety of 2,

Qa ¼ 31 tons ð280 kNÞ
26.2.6 SPT in Clay

Dynamic penetration resistance in saturated clay varies depending on sensitivity
of the clay and excess pore water pressures developed during driving, so SPT data
are considered even less definitive than for sand. The data therefore may correlate
to driving resistance of piles but not the final bearing capacity, which depends on
the setup factor.
For a preliminary estimate of end bearing in clay, Meyerhoff suggests eq. (26.3)
with a maximum value of 3N instead of 4N, and for side friction eq. (26.4).
26.3 CONE PENETRATION TESTS
26.3.1 Push vs. Drive-Cone Tests

Cone penetration tests measure the resistance of soil to penetration by a cone-
tipped rod. An advantage of cone tests over SPTs is that cone tests do not
require a breakdown and reassembly of the drill string after every test.
Another advantage is that instead of being driven with a hammer, the cone
is pushed, which better simulates a field loading condition. However, cone
tests can be difficult to perform from the back of a truck-mounted drill without
the use of screw-in ground anchors to hold the truck down. A drive cone can be
driven with an SPT hammer, but the test is not standardized and should not be
used without extensive correlation data. A hand-operated drive cone is useful for
inspecting footing excavations to ensure that they are in firm soil.
26.3.2 Static Cone Tests

Cone penetration tests (ASTM Designation D-3441) are most readily performed
with a hydraulic pushing apparatus located near the center of gravity of a truck
so that the reaction base is the entire weight of the truck. The system also can
be enclosed, which is amenable to the use of electronic and computer facilities.
Figure 26.6
Schematic
representations of
cone apparatus:
(a) mechanical
cone, (b) detail of
piezocone,
(c) electrical cone.
Tip and friction
sleeve dimensions
are standardized.
Cones have been standardized to have a 608 apical angle, which simulates the cone
of soil with a friction angle of 308 that is pushed ahead of a flat foundation during
a bearing capacity failure. The diameter is standardized at 35.7 mm (1.4 in.) in
order that the cone base area is 1000 mm2 (1.54 in.2).
Static cone penetrometers have an independently supported ‘‘friction sleeve’’ that

measures side friction on the sleeve. The standard sleeve area is 15,000 mm2
(23.25 in.2). As the sleeve is the same diameter as the cone, this corresponds to a
sleeve length of 133.7 mm (5.26 in.). Diagrams of mechanical and electronic cones
are shown in Fig. 26.6.
The standard rate of advance is 20 mm/s, and measurements are made by

mechanical, hydraulic, or electrical means. Electrical readouts allow continuous
testing whereby the cone and sleeve are pushed simultaneously, whereas mechanical
units require alternate pushing of the cone and of the sleeve. Cone penetration,
sleeve friction, and friction ratios can be plotted versus depth as a test proceeds
(Fig. 26.7).
Figure 26.7
Cone penetration
data for a layered
soil. A friction ratio
less than
2 percent indicates
sand. (Data
courtesy of
Terracon
Consultants, Inc.)
26.3.3 Influence of Depth and Lateral Stress

A gradual increase in cone resistance with depth is apparent in the left-hand graph
of Fig. 26.7, similar to that which occurs with the Standard Penetration Test.
Robertson and Campanella (1983) employed a depth function in relating friction
angle of sand to cone bearing value (Fig. 26.8). Note that this correlation applies
only to sand.
Combined influences of depth and lateral stress were examined theoretically

by Durgunoglu and Mitchell (1975) (Fig. 26.9). In most instances K is
between 0.5 and 1.0, but it can be many times higher in overconsolidated
soils and much lower in underconsolidated, collapsible soils. It can be seen
that a high K should increase the cone bearing value and a low K should
decrease it.
26.3.4 Relation of Soil Type to Cone Data

The friction ratio is the ratio of side friction to point stress on a unit area basis.
Because clay increases side frictional resistance, cone data are an indicator of
soil type. For sands the friction ratio is less than 2 percent, whereas for clays
it is from 2 to 10 percent. Intermediate soil types are not as well differentiated
and there is a considerable overlapping of soil types, as shown in
Fig. 26.10. The logarithmic scale for cone bearing suppresses but does
not preclude the role of a depth function. The overlapping of the various
fields is indicative of the accuracy of the determinations. Computer programs
used to interpret cone data can be misleading by specifically identifying soil
types, even though the correlations are not that specific, and the soil
identifications from cone data should be indicated as ‘‘suggested based on
average soil responses.’’

Figure 26.8
Friction angle of
sand related to
cone bearing
value. The
equivalent depth is
depth above the
groundwater
table plus one-half
of the depth below
the groundwater
table. (From data
of Robertson and
Campanella,
1983.)
Figure 26.9
Relation of cone
bearing value to
depth and to lateral
stress ratio, K.
(From Durgunoglu
and Mitchell, 1975,
reprinted with
permission of the
American Society
of Civil Engineers.)
Example 26.7
In Fig. 26.7, a sand layer identified from samples has a cone bearing from 1.5 to 3 MPa and
a friction ratio of 2.5 to 5. How would this soil be identified from cone data?
Answer: Silty sand or sandy silt.

Figure 26.10
Soil identifications
from cone data.
(Modified from
Robertson and
Campanella,
1983.)
26.3.5 Thin Layers

It sometimes is assumed that because of the short length of the cone tip, 30.9 mm
(1.2 in.), the cone test will detect thin layers of dissimilar soils. However, the soil
zone influencing the cone bearing value extends over several tip diameters, and
layers thinner than about 100 mm (4 in.) may remain undetected (Meigh, 1987),
which is unfortunate because thin layers of sand can have a huge influence on the
rate of primary consolidation settlement. Such layers can be reliably detected only
by continuous sampling.
The friction cone measures an average response over the length of the sleeve,
which is 133 mm (5.2 in.). It also should be noted that the center of the friction
sleeve is about 100 mm (4 in.) above the cone tip with the electric cone, or 245 mm
(9 in.) with the mechanical cone, so measurements are not simultaneous for a
particular depth.
Stick-slip occasionally may produce a harmonic response or ‘‘chatter’’ in the

results. This tendency will increase as tests go deeper and the rod acts more
like a spring. Audible stick-slip is sometimes observed in the Borehole Shear test.
26.3.6 Other Uses of Cone Data

Cone data are used for predicting settlement of structures on sand in accordance
with procedures outlined in section 16.10.3, and to gain an approximation for the
friction angle of sand as indicated in Fig. 26.8, with a possible modification for the
influence of lateral in-situ stress (Fig. 26.9).

Cone tests are especially useful for quality control of compacted fill, being
complementary to near-surface density tests made during construction. An
illustration of how cone data were used to diagnose excessive settlement is shown
in Fig. 26.11.
As a rough guide, to prevent excessive settlement the allowable load on clay is

sometimes taken as 1/10 of the cone bearing capacity, and on sand 1/30 of the
cone value (Broms and Flodin, 1988).
The cone can be regarded as a model pile and used to predict pile bearing
capacity, particularly that part of the bearing capacity that is attributable to skin
friction. However, accuracy is influenced by scaling effects and inconsistent pore
water pressures, and is of the order of 50 percent (Schmertmann, 1975).
The cone resistance in saturated clays is a measure of the undrained or partially

drained shear strength, being influenced by the rate of advance and soil sensitivity.
By assuming an undrained condition a value for cohesion may be obtained for use
in the bearing capacity equation:
qc v
cu ¼ ð26:5Þ
NK
Figure 26.11
Cone bearing
values showing
layering in a 13 ft
(4 m) section of
compacted fill. The
regular variation is
described by
Allender as the
‘‘Oreo cookie
effect,’’ but some
of these cookies
had too much
filling. Under the fill
is alluvium and
glacial till.
(Courtesy of
Allender and
Butzke Engineers.)

where cu is the undrained cohesion, qc the cone bearing value, v the total
overburden pressure, and NK is a cone factor that varies from about 10 to 20,
averaging about 15. NK often is evaluated from local correlations with vane shear
measurements of cohesion (discussed below).
An important use of cone data is to determine the potential of sand for

liquefaction during an earthquake, which is discussed in the next chapter.
26.3.7 The Piezocone

Sensitivity of the cone test to layering is improved by monitoring pore water
pressures, which are higher in poorly draining soils. The pore water pressure
generated by probing with the cone depends on the position of the filter, being
highest when it is near the middle as shown in Fig. 26.6(b). The cone also can be
left at one depth to obtain an equilibrium pore water pressure, but this
information is more conveniently obtained with piezometers.
Pore pressure affects both the cone bearing and friction values, and bearing values
may be corrected to be on an effective stress basis (Robertson et al., 1986).
Empirical soil classification schemes also have been devised that incorporate pore
pressure data.
Use of the piezocone and correction for excess pore pressure also improves the
relationship between pile skin friction and cone data (Takesue et al., 1998).
However, this use requires careful de-airing of the system, and the porous stone is
susceptible to plugging.
26.4 VANE SHEAR
26.4.1 A Direct Approach

An alternative to a heavy reliance on empirical correlations is development of
other methods that more directly target specific soil properties. For example, to
determine the shear strength of soil, one can measure it directly instead of relating
it to driving resistance or friction against a steel surface.
The vane shear test was developed in Sweden in the 1940s and uses a vertical rod
with flat plates attached radially in the shape of an X (Fig. 26.12). The instrument
is pushed into the ground or into the bottom of a boring and twisted, and the
torque that is required to cause shearing around the cylindrical surface defined by
the blades is measured. The soil shear strength is calculated from the torque
measurement and the size and shape of blades of the vane.
A unique feature of the vane shear test is that after a peak strength value has been
attained, rotation can be continued to obtain a measure of the residual shear

Figure 26.12
Vane shear: (a)
and (c) standard
blade shapes;
(b) force being
directed to the
perimeter and
tending to mobilize
soil internal friction;
(d) an
experimental
shape used to
emphasize
shearing on
horizontal
surfaces.
strength and therefore of the sensitivity. In the laboratory this is possible only in a
direct shear test by a time-consuming cyclical reversal of the shear box.
26.4.2 Dimensions
The standardized height-to-diameter ratio is 2.0 (ASTM Designation D-2573).
Because, as shown in Fig. 26.12(a), part of the shearing surface is vertical and part is
horizontal, ends may be cut at 458 to prevent shearing along horizontal bedding
planes (Fig. 26.12(c)). Other specialized shapes such as Fig. 26.12(d) have been used
to vary the ratio of vertical and horizontal shearing resistance.
Question: Disregarding the shaft dimensions, what is the ratio of vertical to

horizontal shear areas for a standard rectangular vane?
Answer: With diameter D and length 2D, the ratio is (D 2D)/2(D/2)2 ¼ 4.
26.4.3 What Does Vane Shear Measure?

It often is assumed that the vane test measures only soil cohesion, but as shown in
Fig. 26.12(b), rotation causes a triangular distribution of pressure on a blade such
that the resultant force pushes outward on the soil and will mobilize friction if there
is any drainage. The vane is most suitable for testing slow-draining soft to medium
clay such as occurs in deltas or offshore. The relevant pore water pressures are at the
ends of the vanes where they are difficult to measure. By assuming full drainage,
Farrant in Australia showed that one should be able to evaluate the friction angle
by varying the number of blades, but the method was found to be imprecise.
Comparisons with field situations indicate two additional discrepancies that have
contributed to a declining use of vane shear (Schmertmann, 1975). First, failure is
progressive, starting at the leading edge of each vane and then extending until
failure involves the entire circumference. Therefore by the time shearing is
complete, part of the soil has lost strength due to remolding, with the amount of
strength lost depending in part on compressibility of the soil.
Second, the measured shearing strength is sensitive to the rate of application of

torque. This can be attributed to several factors: development of negative pore
pressure from dilatancy, development of positive pore pressure from compression,
and viscous behavior during measurement of a residual strength. Some of these
conditions give answers that are on the unsafe side.
The formula for reducing torque to shear strength for a vane of rectangular
section and a height/diameter ratio of 2 is
3T
¼ ð26:6Þ
28r3
where is the shearing resistance in lb/in.2 (N/m2 or Pascals), T is the torque in
inch-pounds (newton-meters), and r is the vane radius in inches (meters). The
equation is dimensionally homogeneous.
26.5 BOREHOLE SHEAR TEST

26.5.1 Controlling Normal Stress
The Borehole Shear Test was devised in the early 1960s to perform a direct shear
test on the sides of a borehole. The hole is 3 in. (75 mm) in diameter, preferably
made by pushing a Shelby tube, but it also can be hand-augered.
As shown in Fig. 26.13, two opposing, sharply grooved plates of the shear head
are pushed into soil along the sides of a borehole with a controlled constant
pressure, and after a time for consolidation the device is pulled upward and the
maximum pulling force measured. The normal and shearing stresses are calculated
from the respective forces and the plate areas:
N
n ¼ ð26:7Þ
A
F
f ¼ ð26:8Þ
2A
where n is the normal stress, N the normal force on each shear plate, A is the area
of one plate, f is the shearing stress at failure, and F is the pulling force acting on
two shear plates.

Figure 26.13
Schematic diagram
of the borehole
shear tester (BST).
The shear head is
expanded with a
constant normal
stress, time is
allowed for
consolidation, and
the shear head is
pulled upward.
(Courtesy of
Handy
Geotechnical
Instruments, Inc.)
26.5.2 The Question of Pore Water Pressure

The circumferential grooves and thin shear zone contribute to rapid dissipation of
excess pore water pressure, as shown in Fig. 26.14. The standard test procedure
allows 15 minutes consolidation time for the first data point and 5 minutes for
each successive data point; however, additional drainage occurs during applica-
tion of the shearing stress. Pore pressures in Fig. 26.14 were measured above the
midpoint of the shear plate where they were found to be highest. Because drainage
is unimpeded at the edges of the shear plates, pore pressure should decline to near
zero, so the average should be about one-third to one-half of the values in the
figure.
Pore pressures again increase as shearing stress is applied due to the increase in
major principal stress, so shearing rate is slow, 0.002 in. (0.05 mm)/sec, and is
further slowed down manually as the shearing stress approaches a maximum value.
Pore water pressures can be monitored during the test, or their effect can be more
conveniently determined by alternating shorter and longer consolidation times
to see if they influence the shear envelope (J. Schmertmann, personal
communication).
Figure 26.14
Maximum pore
water pressures
during
consolidation
phases of a
Borehole Shear
test in marine clay
(from data of
Lutenegger and
Tierney, 1986).
26.5.3 Borehole Shear Tests in Unsaturated Soils

Pore water pressures in unsaturated soils are negative and referred to as matric
suction. This usually is interpreted as moving the origin of a Mohr-Coulomb
failure envelop to the left, increasing cohesion and leaving the friction angle
unchanged (Fredlund and Rahardjo, 1993). However, Miller et al. (1998) show
that with a decreasing moisture content and increased suction prior to testing,
cohesion decreases and the friction angle significantly increases. Tests were
conducted in a 300 mm (12 in.) diameter rigid test chamber that may accentuate
changes in matric suction, which becomes more negative with increasing strain.
The phenomenon has not been observed in field tests, and in Miller’s tests the
uplift capacity of model drilled shafts was closely predicted from the Borehole
Shear test results.
Borehole shear tests preferably are conducted in holes left by a 75 mm (3 in.)

diameter thin-walled (Shelby) sampling tube. In caving soils sampling and testing
are conducted through a hollow-stemmed auger, Fig. 26.15.
26.5.4 Serendipity Will Out

Tests revealed an important feature of the Borehole Shear test, that it does not
require repositioning of the shear head between test data points. The reasons
appear to be the radial geometry and concentration of normal stress close to the
shear plates, coupled with rapid drainage of excess pore water pressure. As a
result the consolidated-drained shear strength after shearing and remolding
normally exceeds that of the surrounding undisturbed soil, so the shear surface
moves outward with each successive cycle of shearing. If this does not occur, as in
a quick clay, very low normal stress increments are used and drainage time is
increased. A complete Mohr-Coulomb shear envelope with five or more data
points normally is obtained in 30 minutes to an hour.

This test procedure is called a ‘‘stage test,’’ and results in high degree of reproduc-
ibility and linearity because essentially the same depth of soil is repeatedly tested.
A representative data plot is shown in Fig. 26.16. The bilinear failure envelope is
similar to those obtained with a laboratory direct shear test (Fig. 18.8), and it will
be seen that the normally consolidated leg of the shear envelope extrapolates
Figure 26.15
Borehole Shear
tester shear head
being lowered into
a hollow-stemmed
auger for testing at
the bottom of a
boring. (Photo
courtesy of
R. Failmezger, In
Situ Testing, L.C.,
and Dr. David
White, Iowa State
University.)
Figure 26.16
Borehole shear test data. This test was continued beyond the preconsolidation pressure to establish
a bilinear failure envelope. (Courtesy of Dr. Evert Lawton, University of Utah.)

Figure 26.17
Rock Borehole
Shear test
comparative data.
RBST strengths
are slightly lower
because of
chipping of the
rock by the shear
plate teeth. The
test can be
performed in rock
that is too
fractured to core.
(Data courtesy of
the Oyo
Corporation,
Japan.)
through the origin. Because of its speed and resolution compared with routine
triaxial testing, the test sometimes is adapted for laboratory use.
25.5.5 Rock Borehole Shear Test

In the 1970s the principle of the Borehole Shear test was adapted for testing
sedimentary rocks because of the difficulty of obtaining intact cores of fractured
rocks for laboratory testing. This device uses smaller shear plates in order to
administer normal pressures up to 12,000 lb/in.2 (80 MPa). This is a single-point
test as the shear plates engage and break off a chip of rock, after which the shear
head is removed from the boring, rotated 458, and the test repeated at a different
normal stress. As in the case of the BST, suppression of the sampling variable
results in highly linear failure envelopes that are closely comparable to results
from laboratory tests, as shown in Fig. 26.17.
The test went full circle when the rock tester shear plate was adapted to the soil
tester for testing intermediate rocks such as shale or soft limestone, where it often
can be used as a stage test instead of a single-point test (Handy et al., 1976).
26.6 PRESSUREMETERS
26.6.1 Applying Only a Normal Stress

A pressuremeter consists of an expanding rubber bladder that fits inside a
borehole. The bladder is expanded with water, and the expansion pressure and

Figure 26.18
Advances in
pressure-
expansion
devices, or
‘‘pressuremeters.’’
corresponding increases in volume are measured. The pressure at failure is the

‘‘limit pressure.’’
A predecessor to the pressuremeter was invented in the 1930s by Köhler, and is

shown in Fig. 26.18. However, inflation was with gas pressure, and results were
difficult to interpret because the soil compresses vertically as well as laterally. In
the 1950s, Loius Menard in France cleverly remedied the latter situation by
adding two ‘‘guard cells’’ that are inflated at the same time, thereby giving
essentially a horizontal plane-strain expansion of the center cell. The Menard
pressuremeter and its derivatives are in common use today.
An important modification developed simultaneously in England and in France is

a ‘‘self-boring’’ feature that is designed to slide the instrument into a boring
without allowing relaxation of the soil pressure. Data obtained with self-boring
pressuremeters give additional insight into the expansion mechanics of the
Menard pressuremeter.
26.6.2 Theory
Lamé’s theory for expansion of a hole in an elastic medium indicates that radial
compressive stress should decrease with the square of radial distance. Tangential
stress is tensile, and increases (algebraically) with radial distance, shown at the left
in Fig. 26.19.
In an ideal plastic material, when the radial pressure reaches the proportional
limit, the material in theory will continue to expand so long as the pressure is
maintained. This is the basis for the concept of a pressuremeter limit pressure.
Soils, however, have internal friction that increases as a result of compression,
Figure 26.19
Ideal elastic zone
stresses are
shown on the left,
and changes that
occur when a limit
pressure is
exceeded are
shown on the
right. Tension
cracks reduce
tangential stress
to zero. The rate
of change in radial
stress in the
plastic zone
depends on the
angle of internal
friction.
so the limit pressure is arbitrarily defined as the pressure at which the hole volume
is doubled.
The relation of a plastic zone where the limit pressure has been exceeded and a
surrounding elastic zone is shown at the right in Fig. 26.19. Radial tension
cracking can occur in the elastic zone if tangential tension exceeds the lateral
in-situ soil stress. The existence of these stress zones was confirmed by lateral
stress measurements near Rammed Aggregate Piers shown in Fig. 24.7.
26.6.3 Instrumentation
In the Menard pressuremeter, gas pressure is applied to water in a standpipe and
the water level monitored to indicate changes in the expansion cell volume.
Simultaneously a second standpipe shows inflation of the two guard cells at close
to and slightly below the main cell pressures.
In self-boring pressuremeters instead of the volume being monitored, the diameter

is measured with internal flat springs that carry strain gauges and bear against
external walls of the cell. Another alternative is load cells attached to the exterior
of the device and bearing directly against the soil. Self-boring pressuremeters
normally do not have guard cells, as there should be no initial gap between the
device and walls of the boring.

A ‘‘push-in pressuremeter’’ is similar to the self-boring devices except that the hole
is cut with a Shelby tube-like cutter and soil remains in the center core of the
device until a test is completed and the pressuremeter is removed from the boring.
Another type of pressuremeter is an expansion cell that is arranged to follow a
cone to supplement the cone data with lateral expansion data, but in this case soil
disturbance becomes a major factor.
An external shield of overlapping metal strips is commonly used to protect the

exposed pressuremeter membrane from abrasion or puncturing.
26.6.4 Performing a Test

All pressuremeters must be calibrated by expanding inside a steel tube having
known expansion characteristics. This step is essential to account for expansion of
the hoses, standpipes, etc., under pressure.
A hole is bored and the Menard device lowered to the desired test depth, or the
self-boring or push-in devices are taken to the desired depth. Pressure is raised in
increments with each pressure held constant for 1 minute. The volume change is
measured and pressure-volume expansion curves developed as shown in
Fig. 26.20.
The initial upward curve obtained with the Menard device is sensitive to hole
disturbance and does not occur with the self-boring pressuremeter. This part of
the curve is attributed to repair of disturbed soil with the perimeter of the boring.
French geotechnical engineers who make extensive use of pressuremeter tests
recommend testing in hand-augered borings made with drilling mud that main-
tains a positive pressure on the sides of the hole. The use of a circulating mud
Figure 26.20
Self-boring and
Menard
pressuremeter
curves.

means that the auger is not withdrawn to empty it, and depths of 60 to 90 ft (20 to
30 m) are possible.
26.6.5 In-Situ Stress

The starting point of the linear pressure-volume relationship may reasonably be
assumed to represent the pre-existing lateral stress, and as can be seen in
Fig. 26.20, the starting point from a self-boring device generally is more definitive
than the reading with the Menard device because of hole disturbances.
26.6.6 Modulus
The linear portions of pressure-volume curves generally are well defined,
particularly for a reloading cycle, so it is reasonable that they should be useful
for predicting settlement of foundations. Difficulties are that the drainage state is
not known, compression is horizontal instead of vertical, and is radial instead of
uniaxial.
According to elastic theory,

P
G¼V ð26:9Þ
V
where G is the shear modulus, V the volume of the expanding hole, and P/V is
the slope of the pressure-volume curve after the calibration corrections are
applied. The volume used in eq. (26.9) is the average of V0 and V1 that bracket the
linear response range. With the Menard device the modulus determination may be
made from an unload-reload cycle prior to reaching a limit pressure.
The shear modulus is converted to Young’s modulus E by

E ¼ 2 ð1 þ vÞG ð26:10aÞ
where v is Poisson’s ratio. Menard recommended assuming a standard value of
v ¼ 0.33, which gives
EM ¼ 2:66G ð26:10bÞ
where EM is the ‘‘Menard modulus.’’ However, the modulus differs from that
obtained with a self-boring device, and is horizontal instead of vertical. Menard
developed a complex, semiempirical equation to predict settlement from this
modulus while taking into account footing dimensions and rheological and shape
factors.
26.6.7 Limit Pressure

The limit pressure is a measure of soil strength, but does not differentiate between
internal friction and cohesion. By making certain assumptions the limit pressure
can be used to estimate cohesion by assuming that the friction angle is zero, which
represents an undrained shear strength. Alternatively, a limit pressure can be used

to estimate friction angle by assuming that cohesion is zero, as for a sand.

However, the limit pressure is influenced by simultaneous radial compression and
tangential tension, which is appropriate for soil around driven piles but not under
shallow foundations. Stress may or may not be partly relieved by radial tension
cracks, which are difficult to detect.
Menard related bearing capacity to limit pressure through an empirical bearing

capacity factor k based on grain size and depth of embedment. Other k factors are
defined for pile foundations. Comparisons between predicted and measured
bearing capacities are shown by Briaud (1986) to be variable, with a factor of
safety of 3 required to be on the safe side. Data reduction programs are provided
by the manufacturers.
26.6.8 Role of Empiricism

Most in-situ test methods rely for their interpretation on empirical relation-
ships developed from data. One danger is that there is a world of conditions
that are outside of those used to establish the relationships and that
remain untested. On the other hand, design that is based on experience should
not be discounted if there is a long and continuous record of success in a
particular area.
26.7 THE DILATOMETER
26.7.1 Overview
The soil Dilatometer, shown in Fig. 26.21, was developed in the late 1970s by
Dr. Silvano Marchetti at the University of Turin in Italy. The device is spade-
shaped, 14 mm (0.55 in.) thick, and on one face has a 60 mm (2.4 in.) diaphragm
that is expanded by gas pressure for a distance of 1 mm (0.039 in.). The
Dilatometer is pushed in to the next test depth, pressure readings are obtained
with the diaphragm relaxed and expanded, and the instrument pushed to the next
test depth where the operation is repeated.
The diaphragm is expanded with gas pressure, and the closed and open positions
of the diaphragm are indicated by an electrical contact and buzzer. Because the
diaphragm has an innate stiffness, reference pressure readings are made after the
diaphragm is ‘‘exercised’’ to obtain a uniform response. Subsequent readings are
corrected by reference to these values.
The device is pushed downward for a distance of 100–200 mm (4–8 in.) and a
reading taken, which is the ‘‘A’’ reading. The diaphragm then is expanded 1 mm,
and the expansion pressure is the ‘‘B’’ reading. After correction the two readings
are referred to as p0 and p1. A ‘‘C’’ reading may be taken when the diaphragm is
closed as an indication of static pore water pressure.

Figure 26.21
The Marchetti soil
Dilatometer. Two
pressure readings
are obtained,
one with the
diaphragm flat
and one with it
expanded 1 mm
into the soil.
26.7.2 Evaluating K0
Marchetti defines two dimensionless parameters plus one that has the dimensions
of a modulus. The first parameter is in the nature of a lateral stress ratio:
p0 u0
KD ¼ ð26:11Þ
v0
where KD is the Dilatometer modulus, u0 is the static pore water pressure that is
calculated from the depth below the groundwater table, and v0 is the vertical
effective stress that is calculated from the test depth and groundwater level. If the
p0 reading were an undistorted lateral in-situ stress, KD would equal K0. However,
insertion of the device increases the lateral stress, so a correction is necessary. On
the basis of correlations in clays for a range of K0 ¼ 0.4 to 2.8, Marchetti suggested
the following empirical relationship:
K0 ¼ ðKD =1:5Þ0:47 0:6 ð26:12Þ
Because the response to lateral displacement of different soils is not uniform, this
expression has been modified by different investigators as shown in Fig. 26.22,
and according to the various correlations the ratio of K0 to KD value varies from
about 1/20 to 3. A correlation for sands by Baldi et al. (1986) gave a standard
error of estimate of 0.119. This means that if K0 ¼ 0.5, about two-thirds of the test
values will be in a range 0.38 to 0.62. Dilatometer K0 values appear reasonable as
they are based on averaged soil responses, and they are useful if this level of
reliability for K0 can be tolerated. As discussed in the next section, the

Figure 26.22
Some empirical
relationships
between K0 and
the Dilatometer KD
(From Baldi et al.,
1986, reprinted
with permission of
the American
Society of Civil
Engineers.)
Dilatometer thickness/width ratio is such that p0 probably represents a limit

pressure similar to that which is obtained with the pressuremeter.
26.7.3 Material Index

A second dimensionless parameter defined by Marchetti is the ‘‘material index,’’
which depends on the change in pressure from expansion of the diaphragm in
relation to pressure on the expanded diaphragm. The index is described as a ratio
of soil stiffness to soil strength:
p
ID ¼ ð26:13Þ
p1 u
Low ID values (0–0.5) indicate clayey soils, medium values (0.5–1.5) silty soils, and
high values (41.5) sandy soils. The material index also is used to help define other
soil parameters including overconsolidation ratio, which also depends on KD, and
in turn is used to define preconsolidation pressure. A value for cohesion is
obtained from KD if the friction angle is zero, or of internal friction from ID if
cohesion is zero. The various measures are readily calculated and printed out, but
their number should not defy the principle of ‘‘degrees of freedom,’’ that the
number of answers cannot exceed the number of input parameters. In this case
there are two measurements, p0, p1, plus two other fixed variables, overburden
pressure and pore pressure, plus optional additional measurements such as
penetration resistance and the ‘‘C’’ pressure. If the number of answers exceeds the
number of variables the answers cannot be independent but must be interrelated.
26.7.4 Predicting Settlement

The most successful use of the Dilatometer appears to be to predict settlement.
The Dilatometer modulus, ED, is based only on the increase in pressure needed to
expand the diaphragm:
ED ¼ 38:2p ð26:14Þ

where the constant was derived on the basis of elastic theory for stress near a
circular hole (Marchetti, 1980). A vertical compression modulus M is defined
from elastic theory as
M ¼ E=ð1 2 Þ ð26:15Þ
where E is the elastic modulus and is Poisson’s ratio. As the limits on for an
elastic material are 0 and 0.5, the corresponding limits on M/E are 1.0 to 1.33.
However, when the analogous equation is written using the Dilatometer modulus
and evaluated experimentally, the range is much larger (Fig. 26.23). Let
M ¼ RM ED ð26:16Þ
where RM was found to vary with KD, which appears to be influenced by the limit
pressure, and to a much lesser degree on the material index. Except for very soft
soils the Dilatometer modulus underestimates the elastic modulus, perhaps related
to disturbance from insertion of the instrument. Additional corrections that
further increase M with depth for normally consolidated soils are suggested by
Schmertmann (1986). Settlement is
V H
H ¼ ð26:17Þ
M
where H is the change in thickness of a layer having a thickness H and vertical
compression modulus M, as a result of a change in vertical stress v.
In 15 comparisons with measured settlements, the discrepancy from Dilatometer

predictions varied from 29 percent to þ67 percent, with an average over-
prediction of about 10 percent (Schmertmann, 1986), lending credibility to the
empirical approach. The soils in that investigation varied from a highly
overconsolidated clay (where the overprediction was 67 percent) to silty clay,
silt, silty sand, sand, and peat. The last two showed the largest variability. (The
prediction for a quick clay is not included in this summary as settlement was
overpredicted because of remolding.)
Figure 26.23
RM values to
convert
Dilatometer
modulus ED to a
compression
modulus M.
(Plotted from
equations
presented in
Dilatometer
Manual by
Marchetti and
Crapps.)

Example 26.8
A layer of silty clay 2.5 m (8.2 ft) thick will sustain an added load of 2 bars (2 ton/ft2,
191 kPa). Adjusted A and B readings give p0 ¼ 1.01 bar and p1 ¼ 1.43 bar on a total stress
basis. The average pore water pressure is 0.14 bar, and effective overburden pressure at the
middle of the layer is 0.46 bar. Predict compression of this layer.
Answer: On an effective stress basis, p00 ¼ 0.87 bar and p10 ¼ 1.29 bar. p ¼ 0.42 bar.
KD ¼ p00 =v0 ¼ 0:87=0:46 ¼ 1:90:
ID ¼ p=p00 ¼ 0:42=0:87 ¼ 0:48
ED ¼ 34:7p ¼ 34:7ð0:42Þ ¼ 14:6 bar
From eq. 26.27, RM ¼ 0.9 so M ¼ 0.9(14.6) ¼ 13.1 bar.
Then H ¼ v H/M ¼ 2 bar (2.5 m)/13.1 bar ¼ 0.4 m (1.2 ft).
26.8 LATERAL STRESS WITH THE K0 STEPPED BLADE
26.8.1 The Challenge

Lateral in-situ stress is one of the most difficult soil properties to measure, while
being one of the most important in engineering, not only as the basis for retaining
wall design, but also for foundation bearing capacity and skin friction on piles.
The most accurate means for measuring lateral stress in situ in soils probably is
with a self-boring pressuremeter, but the test is expensive and time-consuming and
the instrument is easily clogged or damaged.
Another approach is to push in a device such as a cone or Dilatometer that

displaces soil and increases lateral stress according to shear or compression
behavior, and attempt an empirical correction. A modification is to push in a thin
load cell such as a ‘‘Gloetzl’’ cell and leave it in place until the soil pressure
equilibrates, but there is no assurance that the final measured pressure represents
that which existed prior to insertion of the instrument.
A load cell also can be buried in soil or attached to a solid surface to measure
changes in pressure in response to some type of loading. This approach has been
widely used in retaining wall and underground conduit research, but caution still
must be used in interpreting the data because of a difference in modulus between the
load cell and the soil, such that pressures may tend to concentrate on the load cell.
A fourth approach is to pump in a fluid and cause hydraulic fracturing, indicated

by a sudden reduction in pressure, but in an overconsolidated soil the fractures
may occur horizontally so that the pressure is measured vertically.
In the 1970s the U.S. Federal Highway Administration sponsored research

to develop a better method for measuring lateral in-situ stress in soils.

The ‘‘K0 Stepped Blade’’ introduces different levels of disturbance and extra-
polates measurements to obtain a hypothetical pressure on a zero-thickness blade.
A similar approach is used in chemical analysis and X-ray diffraction—to add a
measured amount of a chemical or mineral and make two measurements.
26.8.2 Operation of the Blade

As shown in Fig. 26.24, the Stepped Blade has four steps incrementally increasing
in thickness from 3 to 7.5 mm. Each step carries a pneumatic pressure cell that is
flush with the surface and has a cover membrane that is in contact with the soil.
The cell has two chambers that are inflated with slightly different gas pressures so
that when the soil pressure is reached, the membrane is lifted and allows a cross-
leak that is indicated by a differential pressure gauge, at which time the gas
pressure is read. As the gas pressure equals the soil pressure, no additional
calibration is needed. The cell then is immediately vented to prevent bulging of the
membrane.
A test is conducted by boring to the test depth, then removing the drilling
instrument. The first, thinnest step of the blade is pushed into the bottom of the
boring and the pressure is read. Then the second step is pushed in, and the
Figure 26.24
Lateral pressures
are measured
sequentially at the
same depths with
the K0 Stepped
Blade and
extrapolated to
obtain a
hypothetical
pressure on a
zero-thickness
blade. (Courtesy
of Handy
Geotechnical
Instruments, Inc.)

pressure is read on both the second and first steps, in that order. Then the third
step is pushed in and the pressure read on cells 3, 2, and 1, and so on, until all
steps have been pushed and read. An additional blank section of the blade is
pushed to obtain an additional set of readings. The result is 4 pressure readings at
the shallowest depth, 4 at the next, 3 at the next, 2 at the next, and 1 at the deepest
depth, giving 14 pressure readings in all. This operation normally requires about
15 minutes. The instrument then is removed and the hole advanced to the next test
depth, where the operation is repeated.
The device measures total stress, so it is important that the level of the
groundwater table be determined in order to reduce the data to effective stresses.
26.8.3 Interpretation
The logarithm of pressure was found to relate to blade thickness, which is
analogous to a consolidation e-log P plot. This implies that for a test to be valid
the soil must be consolidating. The sequence of reading is intended to allow
approximately the same drainage time for each step. If consolidation is prevented
by poor drainage, all steps should measure a limit pressure.
26.8.4 Results
Some representative test results are shown in Fig. 26.25. An important step in
interpretation is to discard data that do not show an increase in pressure with
increasing step thickness. The two most common causes are (1) in stiff soils the
first step often creates an elastic response because the soil structure is not broken
down, which gives a high reading that is a measure of in-situ stress plus a
Figure 26.25
K0 Stepped Blade
test data at three
subdepths showing
extrapolation to
obtain lateral
in-situ stress. The
question mark
indicates elastic
response prior to
the soil structure
breaking down.

modulus; (2) in soft soils the thicker steps can cause a lateral bearing capacity
failure that is analogous to the limit pressure in the pressuremeter test. Data that
do not meet the requirement that pressures must increase with increasing
thickness therefore are omitted from the interpretation.
In Fig. 26.25 one measurement is omitted because of an elastic response and

several are omitted because the pressure reached a limiting pressure and stayed
constant or decreased. The change in slope of the data relates to drainage times,
and excess pore pressures also appear to extrapolate out and give the same zero-
thickness extrapolation for total lateral stress.
26.8.5 Accuracy of the Stepped Blade Test

A comparison of Stepped Blade test data with lateral stresses measured with a
self-boring pressuremeter in a saturated, highly plastic expansive clay is shown in
Fig. 26.26. In this series 44 Stepped Blade determinations were made that allowed
calculation of a statistical confidence interval, discussed later in this chapter.
Another series of tests conducted in a passively displaced silt indicated a test
Figure 26.26
Blade averages
(open circles) and
90 percent
confidence limits
from 44 tests
bracket six
determinations
with a self-boring
pressuremeter.
Pressuremeter
data are courtesy
of Dr. Michael
O’Neil, University
of Houston. (From
Handy et al.,
1982.)

Figure 26.27
Thickness/width
ratios of various
push-in soil
measurement
devices, and (top)
soil responses
identified from
Stepped Blade
tests.
precision of 5 percent, with pressures that were consistent with Rankine theory
(Handy and White, 2006).
26.8.6 Relation to Thickness/Width Ratio

Relationships between various push-in soil-testing instruments are shown
in Fig. 26.27, with soil responses identified from Stepped Blade tests. The
thickness/width ratio of a Gloetzl cell varies from about 0.05 to 0.025, so insertion
normally will cause either an elastic or consolidating response. An elastic response
will not result in a time-related relief of pressure, and a consolidating response
eventually should reach an equilibrium pressure after completion of primary and
secondary consolidation.
26.9 STATISTICAL TREATMENT OF DATA
26.9.1 Overview
Statistics can be applied to any data, but statistical methods are particularly
appropriate for soils because of their large variability, and for in-situ tests because
of the abundance of data that can be generated in a relatively short time. These
two characteristics are complementary, because a large variability requires many
evaluations to establish a reliable average that may be used for design. For
example, if the strength of a soil deposit varies between 20 and 40, it is very
unlikely that a single measurement will give an answer that is representative of
the true average. If two tests are performed and averaged, the estimate will
be improved, and if 10 tests are performed and averaged, the estimate will be
improved even further. A statistical analysis provides a means for establishing
reliability of an average (or mean), based on variability and the number of
evaluations.

The engineer often must make a choice between performing a few highly
sophisticated tests or performing more tests that are less accurate and less costly.
Which approach is best, or can they be combined, a survey test to establish an
average and more detailed tests on a sample that is representative of the average?
Variability is expressed by a ‘‘standard deviation,’’ which is determined by

summing departures of data from an average. However, as the average will vary
depending on the number of tests, the standard deviation is only estimated and
depends on the quality of the average. The standard deviation, which once
required hours to determine with a hand-crank calculator, now is determined
almost instantly with a computer or a small electronic calculator that has statistics
functions.
The standard deviation calculated from test results actually represents the sum of
two standard deviations, one representing the soil variability and the other a test
variability. This relationship is as follows:
s2 ¼ s2s þ s2t ð26:18Þ
where s is the standard deviation that is measured, ss is that of the soil, and st that
of the test.
This equation is easily visualized because it describes the hypotenuse of a right

triangle (Fig. 26.28). Soil variability is represented by the long leg of the triangle,
test variability by the shorter leg. The objective of testing should be to minimize
the length of the hypotenuse, but it can be seen that if the soil is highly variable,
little can be gained by using highly refined test procedures that have a low
variability. This may help explain the continuing popularity of rapid test
procedures, such as SPT, CPT, and unconfined compressive strength, even though
they only indirectly address basic soil parameters. However, these procedures can
be valuable to screen and select representative samples that can be subjected to
more detailed tests, thus giving the best of both worlds.
The relationship in Fig. 26.28 also indicates that a way to reduce variability is to
target the soil variability by subdividing the soil on the basis of distinguishing
Figure 26.28
Statistical variability s is the square root of the sum of squares of test variability s t and soil variability
ss. For most efficient evaluations s should be minimized, so the test should fit the soil variability.
characteristics. For example, if sand layers are interbedded with clay layers,
it would be foolish to determine critical engineering properties based on a single
composite sample, as there should be at least two samples.
26.9.2 Measuring Variability

The simplest statistic is the average or arithmetic mean, which is readily calculated
and easily understood. However, the mean of 20, 25, 30 is the same as the mean of
5, 25, and 45, so an expression also is needed to describe variability. The standard
deviation includes the number of samples or determinations:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P
ðX XÞ 2
s¼ ð26:19Þ
n
where s is the estimate for the standard deviation, X represents each individual
value, X is the mean of the X values, and n is the number of X values.
(An alternative procedure is to divide by n – 1 instead of n in recognition of the
degrees of freedom.) The estimate of s becomes more reliable with an increasing X
or number of measurements, but there should be no systematic change in s related
to the number of measurements. (In statistical literature s usually is represented by
, but the Roman version is used here to avoid confusion with stress. Also, s can
signify an estimate of the standard deviation.)
One use of the standard deviation is to calculate how much of a population

probably exceeds certain limits. This usually is done by assuming an ideal normal
or Gaussian distribution of data, which is illustrated in Fig. 26.29. As indicated in
the figure, 67 percent of the area under the curve, which represents 68 percent of
the population, is bounded by X s.
Example 26.9
The following Standard Penetration Test blow counts were obtained in 10 tests:
N ¼ 4, 12, 18, 9, 16, 6, 13, 10, 8, 14
Calculate the standard deviation (without using a program) and, assuming that this is an
accurate representation of the true population spread, estimate the N value exceeded by
95% of the deposit. (Use 90% area boundaries since these will define 5% ‘‘tails’’ at both
ends of the distribution.)
Answer: The sum is 110 so the mean is X ¼ 110/10 ¼ 11.0.
Values of (X – X) are as follows: 7, 1, 7, – 2, 5, – 5, 2, – 1, 3, 3. Therefore

2 ¼ 49 þ 1 þ 49 þ 4 þ 25 þ 25 þ 4 þ 1 þ 9 þ 9 ¼ 176. Then
(X – X)
rffiffiffiffiffiffiffiffi
176
s¼ ¼ 4:2
10
Therefore X ¼ 11.0 4.2 where the entry signifies one standard deviation. If these are
accurate estimates, from Fig. 26.29, 90% of the values will be between 11.0 þ 1.645(4.2),

Figure 26.29
A statistical
normal distribution
relating the
number of tests
(vertical axis) to
deviations from a
mean value
(horizontal axis).
The s values in
this case are
standard
deviations (s in
the text).
or 4.1 to 17.9. This means that if the distribution is symmetrical, 5% will be less than 4.1
and 95% will exceed 4.1.
26.9.3 Certainty of an Average

Because engineering decisions often are based on average properties instead
of the spread, a more useful application for standard deviation is to establish
the reliability of an average. This procedure establishes confidence limits on
the mean and is a function of s and n. This range refers only to the
mean, and not to the population or ‘‘spread.’’ The larger the sample, the
closer the confidence limits. This is analyzed according to a ‘‘t distribution’’
(Table 26.1).
Example 26.10
Calculate the 95% confidence limits on the mean in the previous example.
Answer: The 95% confidence limits are X 0.75(s) ¼ 11.0 0.75(4.2) ¼ 11.0 3.2, or
between 7.8 and 14.2.
If the average value of 11.0 is used for design without a factor of safety, the
likelihood that the true average is lower than that value is 50 percent, so if
the design procedures are accurate the failure rate would be 50 percent, which
is unacceptable. If the lower value of 7.8 is used for design, the failure rate
is reduced to 2.5 percent, which represents one ‘‘tail’’ of the distribution
No. of observations 90% confidence 95% confidence Table 26.1
4 1.36 1.84 Confidence limits on

5 1.06 1.39 a mean value,
X a s
6 0.90 1.15
7 0.79 1-00
8 0.72 0.89
9 0.66 0.82
10 0.61 0.75
12 0.54 0.66
14 0.49 0.60
16 0.45 0.55
18 0.44 0.51
20 0.40 0.48
25 0.35 0.42
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
425 1:645 ðn 3Þ 1:96 ðn 3Þ
Adapted from ASTM (1951).
and normally would be unacceptable even though the nominal ‘‘factor of safety’’
is 11 7.8 ¼ 1.4.
The statistical ‘‘t-test’’ therefore can allow one to evaluate the reliability of a
factor of safety. The lower the standard deviation represented in Fig. 26.28, the
better the reliability.
26.9.4 How Many of Anything Is Enough?

In the above example if the number of tests had been 20 instead of 10 and
had generated a comparable set of numbers, the 95 percent confidence limits
on the mean would have been 11.0 0.48(4.2) ¼ 11.0 2.0 instead of 3.2,
thereby reducing the range by more than one-third. The larger the number
of tests, the greater the certainty of the mean value. However, the tests must be real
and not fictitious, which would be a fraudulent misrepresentation of data.
A guide to the number of tests is presented in ASTM Designation E-122:

n ¼ ð3s=EÞ2 ð26:20Þ
where n is the number of tests, s is the standard deviation, and E is the acceptable
error. In theory this still leaves a 3 in 1000 chance of being out of bounds based on
a normal distribution. However, in engineering the normal distribution has its
wings clipped because it is impossible to have negative values for blow counts,
cone resistance, factor of safety, etc., so 3 in 1000 usually can be regarded as
virtual certainty.

Example 26.11
How many tests will be required for an allowable error of 20% of the mean of the
preceding example? (This allowable error would be covered by increasing the factor of
safety by 20%.)
Answer: The estimate of E ¼ 0.20(11.0) ¼ 2.2. Then n ¼ (3 4.2/2.2)2 ¼ 33 tests.
As this calculation is based on preliminary evaluations of s and X from 10 tests,

those estimates could be repeated after 20 tests to determine if there are any
changes.
26.9.5 Two Populations

The most effective way to reduce data variability is to avoid the fruit salad of
mixed data, such as not discriminating between sand and clay layers. In some
cases differences may not be obvious until the data are plotted as a bar graph,
which is readily accomplished with a computer spreadsheet.
A bar graph shows the number of data points falling within prescribed data
intervals, and in theory will approach a statistical normal distribution if there is
sufficient data. Figure 26.30 shows a bar graph for 123 standard penetration tests
conducted in glacial till at a hospital site. To plot such a graph an arbitrary
interval is assigned on the horizontal axis, and each test value is added to the bar
having the corresponding interval. In Fig. 26.30 three tests have N values from 10
to 14, seven from 15 to 19, and so forth.
Figure 26.30
Standard
Penetration Test
blow counts in a
glacial deposit
consisting of
brown (oxidized)
till over gray
(unoxidized) till.
(The heights of
the bars only
reflect the number
of tests and not
the strength.)

Even though the soil is glacial till, the distribution in this case is obviously bimodal
because it has two modes, or population peaks. The modes correlate with the field
description, one soil being gray and unoxidized and the other brown and oxidized.
At one time the depth of oxidation was attributed to weathering time, but this
graph indicates a more fundamental reason, that the gray till is harder and more
dense so it restricted entry of oxygen-charged water. The gray till probably is
‘‘subglacial,’’ meaning that it was compressed under the weight of the glacial ice,
whereas the brown till probably was deposited during the final retreat of the ice
front. This information obviously can be helpful for interpreting data from other
similar sites.
The two populations therefore are treated separately to calculate means and
standard deviations, which show that not only is the brown till layer much weaker
than the underlying gray till (N ¼ 27 compared to 73), it also is more variable
(s ¼ 12 compared to 5). This indicates a greater likelihood that the brown
variation may contain weaker zones that were not detected in the testing. The
building therefore was founded on deep foundations that extend into the
unoxidized gray till.
26.9.6 Other Statistical Measures

t-Tests
A t-test is used as illustrated above to define confidence limits on the mean, and
also may be used to compare means of two samples to determine the likelihood
that they represent two distinct populations.
Variance
Variance is the square of the standard deviation. An ‘‘analysis of variance’’
(ANOVA) is used to compare means of three or more samples.
Standard Error
The standard error sometimes is included as a notation after the mean to relate
the reliability of the mean to the number of observations. Standard error is
smaller than the standard deviation, and is defined as
" #
1
se ¼ s 1 pffiffiffiffiffiffiffiffiffi ð26:21Þ
ð2nÞ
Coefficient of Variation
How can one compare apples with oranges? Statistics at least allow a comparison
of the variability of two populations. This is useful to compare variability of
different test methods. The coefficient of variation is
s
Cv ¼ ð26:22Þ
X

where s is the standard deviation and X is the mean. The coefficient of variation
often is expressed as a percent. In order to determine the coefficient of variation
for a particular test method, the test should be repeated many times in a
uniform soil.
Example 26.12
Calculate the coefficient of variation in the above example.
Answer: Cv ¼ 4.2/11.0 ¼ 0.38 or 38%. This includes contributions from both soil and test
variability.
Linear Regression and Correlation

Linear regression, or fitting a straight line to X and Y data, is frequently used in
engineering and can readily be performed with a computer or programmable
calculator. In applying this method it is critical to assign X values to an
independent variable that has little or no error, because is it assumed that all
random error is in the Y variable. If that is not possible, the regression can be
repeated, first assuming that the X variables are independent, and then that the
Y variables are independent, discussed below.
Linear regression investigates the data and fits a linear relationship that minimizes
the sums of squares of deviations in the Y direction. The result has two
coefficients, a and b, for a linear equation having the form
Y ¼ aX þ b ð26:23Þ
2
Most computer or calculator programs also will give a value for R , which is the
correlation coefficient. If all data are on line, R ¼ 1 or – 1, and there is no deviation
of data from the line. A negative value only means that X increases as Y decreases.
Squaring R takes away the sign. R2 also is affected by the number of data points,
n, which therefore should be stated.
No Independent Variable
In engineering both X and Y data may have random variability, in which case a
simple regression of one on the other is not correct and will give the wrong a and b
coefficients. If both random errors are unavoidable and approximately equal,
two regressions, Y on X and X on Y, will give two lines that intersect at
the mean values of both variables, and an average can be sketched in. Both
relationships will have the same R2 and the lines will bracket the correct
relationship. If R2 ¼ 0, the two regression lines will be at right angles and there is
not a relationship.
There are many other statistical methods that can be useful in engineering,
for example to determine confidence limits for a linear regression, and
engineers are encouraged to take a course in statistics, if only to gain a better
appreciation of the inevitable variability of data and the presence of experimental
error.

Example 26.13
The following data were obtained from calibration of a test instrument:
Dynamometer: 0 3.4 6.5 9.5 13 16.4 19.5 22.6 25.6 29
Gauge pressure: 0 20 40 60 80 100 120 140 160 180
Determine the regression coefficients and correlation coefficient with the dynamometer as
the independent variable.
Answer: a ¼ 0.5230, b ¼ 6.222, R2 ¼ 0.9998, n ¼ 10.
Beta Distribution
The Gaussian distribution reaches into prohibited territory, for example by
indicating a positive probability for a negative factor of safety, which is physically
impossible. This has led to increasing use of a beta distribution, which is more
versatile because end points can be defined. Usually a beta distribution is selected
that is closest to a normal distribution but is subject to an experimental validation.
A beta distribution has two shape functions, and , that are equal if the distri-
bution is symmetrical. Some examples of beta distributions are shown in Fig. 26.31.
R. Failmezger (personal communication) applied a beta distribution to show a
linear relationship between standard deviation and a required factor of safety.
Normalizing Data
In statistics to ‘‘normalize’’ literally means to make data more normal, that is, to
more closely fit a normal distribution such as shown in Figs. 26.29 and 26.30.
Plotting particle sizes to a logarithmic instead of a linear scale is an example of
this kind of normalization, and is justified because sedimentation rate depends on
the square of the particle diameter and not on the diameter, and most engineering
soils are sediments. Any use of a logarithmic scale to obtain a linear relationship,
as in interpretation of consolidation tests or Stepped Blade data, is in fact
Figure 26.31
Examples of
statistical beta
distributions.
Advantages of
beta are that it is
adjustable and
can avoid any
implication of
negative values.

normalizing the data. This definition can be broadened to include any

rearrangement of data that will either yield or take advantage of a linear
relationship.
Another kind of transform is if data variability increases with size—the larger the
size, the larger the variability. The coefficient of variation is an example of this
kind of normalization, and allows comparisons between entirely different objects
or occurrences.
In computer science, data are normalized to fit within a prescribed range that is
easier to compute. ‘‘Floating point’’ arithmetic uses numbers expressed with an
exponential notation, such as 12 105 instead of 0.00012.
Researchers sometimes normalize data by expressing it in dimensionless

terms. For example, pressure can be expressed relative to atmospheric
pressure and expressed in ‘‘bars.’’ An advantage is that this eliminates the
system of units, as a dimensionless term is the same regardless whether units are in
SI, English, or some other system. The Reynolds number is a dimensionless term
that empirically has been shown to separate laminar from turbulent flow of fluids.
Dimensionless terms are the ‘‘pi terms’’ of dimensional analysis.
Sometimes a transform to a dimensionless number can introduce unintended bias,

which defeats the purpose of normalization. Two examples that are common in
geotechnical engineering are the overconsolidation ratio (OCR) and coefficient of
lateral stress (K0). Both of these ratios involve vertical stress that is zero at the
ground surface. It therefore is impossible to calculate an OCR or K0 value for
soil at the ground surface because it means dividing by zero. It also is not
appropriate to describe a soil as having a particular value or even an average
value of OCR or K0, any more than it is to describe a soil as having a particular
overburden pressure. The popularity of these terms may relate to their use
for laboratory triaxial tests where the reference vertical stress is known or
controlled.
26.9.7 Rejection of Data

It is not permissible to omit data simply because points do not fit a relationship.
Bad data can be worse than no data, so care is required in making measurements.
Data points that are far off the mark are called outliers. An outlier should be
rechecked to make certain there has been no mistake in calculating or entering the
numbers. Nevertheless outliers are unavoidable, as when a penetration test hits a
rock or a void.
One guideline is the probability of occurrence of an outlier, for example that the
probability should not be less than 1/2n. For example, if the number of
observations is 10, an outlier point may be rejected and labeled with a question

mark if its probability of occurrence is less than 1/20 or 0.05. This and other
probabilities are presented in ASTM Designation E-178.
A simple calculation is required to test an outlier; its departure from the estimated
mean value is divided by the standard deviation:

X X
D¼ ð26:24Þ
s
The criteria for rejection are shown in Fig. 26.32.
Considerable caution should be used in omitting outliers, particularly if they are

on the unsafe side, because they could be real. Preferably the test should be
repeated. A low CPT or SPT value could flag a void or a soft zone.
Example 26.14
The following SPT blow counts are obtained from tests in a uniform clay. Estimate
the mean and standard deviation, examine for outliers, and if necessary recalculate the
mean.
N ¼ 18, 12, 46, 13, 16
Answer: The mean and standard deviation are 21.0 14.2, n ¼ 5. The questionable data
entry is 46, which is on the high side and will raise the mean value. From eq. (26.24),
D ¼ j21:0 46j 14:42 ¼ 1:73
From Fig. 26.32, with n ¼ 5 there is less than a 5% probability that this is from the same
population as the rest of the values. It therefore is reasonable that this data point should
be rejected. The recalculated mean and standard deviation are 15 2.8 instead of
21.0 14.2, giving a lower and therefore a safer and more reliable estimate that can be used
for design.
Figure 26.32
Criteria for
rejection of data
(adapted from
ASTM Designation
E-178).

Figure 26.33
The borehole
shear test is one
of the few in-situ
tests that can be
performed without
a drilling machine,
and is the only
field test that
measures soil
cohesion and
friction angle on
an effective stress
basis. The shear
head is being
expanded into the
soil with air
pressure and will
be pulled upward
to cause shearing.
26.10 SUMMARY
26.10.1 Tests with an Attitude

The pioneers of geotechnical (then called soil) engineering had specific objectives in
mind when they devised tests to measure soil properties such as particle sizes,
plasticity, compressibility, and friction angle and cohesion (Fig. 26.33). This is in
contrast to the early developments in in-situ testing, where primary goals were
speed and ease of use with a drilling machine. Tests such as the standard penetration
test are effective as survey tests, but do not discriminate between compressibility
and shear strength. Important improvements include development of cones that
are pushed instead of driven, and separate determinations of side friction and
end bearing, Pore water pressures can be measured as an aid to soil identification
since there is no sample. Nevertheless the ambiguity of empirical correlations has
opened journal pages to hundreds of papers that delve into the details. A unique
and positive aspect of the standard penetration test is that it provides a sample.
26.10.2 The Next Stage

The next generation of in-situ tests was more targeted and includes vane shear, the
pressuremeter and Dilatometer. However, these tests still measure combined
factors. Users also should be cautioned that if many different parameters can
appear on the printed page, they cannot be independent if their number exceeds
the number of measurements. Thus a soil may be evaluated for friction or for
cohesion but not for both, which leads to overdesign. This is permissible for small
structures but can become very expensive for large ones, so a trend continues to
develop in-situ tests that are directed towards specific objectives such as internal
friction and cohesion and lateral in-situ stress.
26.10.3 A Useful Combination

Soils are so variable that many tests are required for statistical reliability. The
speed and simplicity of tests such as the cone and SPT therefore are appropriate
for initial survey purposes, and can be followed with more definitive tests that
focus on particular soil properties. For both safety and economy this can be a
winning combination.
Problems
26.1. Prepare a table showing advantages and disadvantages of the cone
penetration test compared with the Standard Penetration Test. Include
level of skill of the technicians.
26.2. How does geotechnical exploration drilling differ from drilling for water?
For oil?
26.3. Assume that a soil loses one-half of its strength by remolding. A standard
cone push rod is 36 mm (1.4 in.) o.d. and 16 mm (0.63 in.) i.d. Calculate
the amount of extension of 10 m (30.5 ft) of rod as a soil goes from a
peak strength of 2 MPa (20 tons/ft2) to its residual strength. The modulus of
elasticity of steel is 200(10)6 kPa (30 106 lb/in.2). Explain the significance.
26.4. What is the vane shear strength of sand? Explain.
26.5. As a geotechnical engineer you are asked to observe exploration drilling
and testing at a job site where they are performing a Standard Penetration
Test. What do you look for in terms of the test, samples, and
characterization of soil between the test depths?
26.6. The following SPT data were obtained at a depth of 5 ft in soil above the
groundwater table with a doughnut hammer: 15, 23, 26. What is N in
blows per foot (0.3 m) What is N60 with a depth correction?
26.7. The soil in the preceding problem is a sand. Which value of N is used to
estimate the friction angle? What is the estimated friction angle?
26.8. Use the answer from the preceding problem to estimate the bearing
capacity of the footing in Example 26.4.
26.9. Predict settlement for the footing in Problem 26.8.
26.10. From averaged cone data in Figs. 26.7 and 26.10, estimate the grain
size of the Cary till and of the sand. Which identifications are more
reliable?
26.11. The apical angle of the cone test simulates what soil friction angle?
26.12. Estimate total settlement from compression of the weaker layers of fill in
Fig. 26.11 under a broad load of 1 ton/ft2, assuming that the layers behave
as a sand.
26.13. The stiff layer immediately underneath the fill in Fig. 26.11 is a clay
paleosol. Estimate its cohesion and bearing capacity. How does NT
compare with the bearing capacity Nc for a circular foundation? How do
you explain the differences, if any?
26.14. The maximum torque measured with a standard rectangular vane is
35 ft-lb (47.5 N-m). What is the shear strength?
26.15. The test in Fig. 26.16 was conducted at a depth of 16 ft (4.88 m). The soil
unit weight is 105 lb/ft3 (16.5 kg/m3). The groundwater table is at 4 ft
(1.2 m) depth. What is the OCR?
26.16. A Borehole Shear test develops the following data in unsaturated loess:
Normal stress, lb/in.2 (kPa): 5.5 (38.5) 12 (84) 17.7 (124) 24 (168)
Shear failure ’’ ’’ 4 (28) 4.6 (32) 11.6 (81) 14.7 (1.3)
(a) Plot a Mohr-Coulomb graph and failure envelope, and explain if any
data are omitted. Estimate cohesion and angle of internal friction. (b)
Perform a linear regression for linear portions of the graph and determine
the correlation coefficient, cohesion, and friction angle.
26.17. The following Stepped Blade measurements were obtained at two
subdepths:
Step thickness (mm): 3 4.5 6 7.5
2
Pressure lb/in. (kPa): 32 (224) 24 (168) 26.5 (186) 29 (200)
45 (315) 29 (186) 34 (238) 30 (210)
(a) Plot the data with pressure on a log scale (or plot log pressure),
examine the data for acceptability, and extrapolate to obtain lateral in-situ
stresses. (b) Perform linear regressions on the thickness-log pressure data
with thickness as the independent variable, calculate the correlation
coefficient, lateral in-situ stress, and slope of the lines.
26.18. K0 Stepped Blade measurements in a basal stratum of loess give K0 ¼ 1.0.
Explain. What are the engineering implications for the overlying loess soil?
26.19. Give two explanations why lateral stress measurements in a residual soil
consistently give K041. Will K0 be constant with depth? Explain. (Hint:
What is K0 at the ground surface?)
26.20. In Example 26.8 a clay layer 1 m (3.3 ft) thick gives a Dilatometer
A reading ¼ 1.14 þ 0.22 bar and B ¼ 2.39 – 0.43 bar. The pore water
pressure is 0.475 bar, and the effective overburden pressure is 0.70 bar.
Calculate settlement from compression of this layer.
26.21. In Fig. 26.30 there are several data points that occur in the wrong popu-
lation. Perform a statistical test to determine if they should be omitted.
26.22. In the glacial till in Fig. 26.30, what laboratory test might give an estimate
of the thickness of the glacier? What in-situ test?
26.23. In Example 26.9 the standard deviation is 4.2. (a) What is the coefficient
of variation? (b) Estimate the limits on the mean that will encompass 50%
of the data (known as the ‘‘probable error’’). (c) What is the standard
error?
26.24. In Example 26.9 predict how many tests will be required for an acceptable
error of 10%.
26.25. In Fig. 26.30, which curve most nearly simulates a symmetrical normal
distribution? How could one confirm that this is the best curve?
26.26. SPTs are performed in a glacial till that contains gravel particles.
(a) Calculate the mean and standard deviation for N ¼ 17, 16, 25, 20, 18,
37, 45, 26, 18, 22 blows/ft (0.3 m). (b) Plot a bar graph of the data. Would
you be justified in deleting any value(s)? If so, suggest a possible explanation
and recalculate part (a). (c) The tests start at a depth of 5 ft (1.5 m) and are
performed at 5 ft (1.5 m) depth intervals. Apply a depth correction as
necessary. (d) Calculate 95% confidence intervals on the mean.
26.27. What tests are most appropriate to determine soil variability across and
under a site? What tests are most appropriate to establish design parameters
such as compressibility and effective stress internal friction and cohesion?
References and Further Reading

Almar, S., Baguelin, F., Canépa, Y., and Frank, R. (1998). ‘‘New Design Rules for the
Bearing Capacity of Shallow Foundations Based on Menard Pressuremeter tests.’’ In
Geotechnical Site Characterization 2, 727–733. Balkema, Rotterdam.
American Society for Testing and Materials (Annual Book of Standards). ‘‘Standard
Practice for Dealing with Outlying Observations,’’ Designation E-178.
American Society for Testing and Materials (1951). ASTM Manual for Quality Control of
Materials, Spec. Tech. Publ. 15-C. ASTM, Philadelphia.
Baldi, G., Bellotti, G., Ghionna, V., Jamiolkowski, M., Marchetti, S., and Pasqualini, E.
(1986). ‘‘Flat Dilatometer Tests in Calibration Chamber.’’ Use of In Situ Tests in
Geotechnical Engineering, ASCE Geotech. Spec. Pub. No. 6, 431–446. ASCE
New York.
Briaud, J. L. (1986). ‘‘Pressuremeter and Foundation Design.’’ Use of In Situ Tests in
Geotechnical Engineering. 74–115. ASCE, New York.
Broms, B., and Flodin, N. (1988). ‘‘History of Penetration Testing.’’ In Penetration Testing
1988. Balkema, Rotterdam.
Brouwer, J. J. M. (2002). ‘‘Guide to Cone Penetration Testing,’’ http://
www.conepenetration.com.
de Mello, V. (1971). ‘‘The Standard Penetration Test—a State-of-the-Art Report.’’ Proc.
4th Int. Conf. on Soil Mechanics and Foundation Engineering 1, 1–86.
Deacon, J. (2006). ‘‘The Really Easy Statistics Site’’ http://helios.bto.ed.ac.uk/bto/
statistics/tress3.html
Durgunoglu, H. T., and Mitchell, J. K. (1975). ‘‘Static Penetration Resistance of Soils.’’
In Situ Measurement of Soil Properties I, 151-188. ASCE, New York.

Failmezger, R. A., and Anderson, J B., ed. (2006). Flat Dilatometer Testing. Proc. of the
Second Intern. Conf. on the Flat Dilatometer. In-Situ Soil Testing, L. C., Lancaster, Va.
Fredlund, D. G., and Rahardjo, H. (1993). Soil Mechanics for Unsaturated Soils.
John Wiley & Sons, New York.
Gibbs, H. J., and Holtz, W. G. (1957) ‘‘Research on Determining the Density of Sands by
Spoon Penetration Testing.’’ Proc. 4th Int. Conf. on Soil Mechanics and Foundation
Engineering 1, 35.
Handy, R. L., Pitt, J. M., Engle, L. E., and Klockow, D. E. (1976). ‘‘Rock Borehole Shear
Test.’’ Proc. 17th U.S. Symp. on Rock Mechanics 486, 1–11.
Handy, R. L., Remmes, B., Moldt, S., Lutenegger, A. J., and Trott, G. (1982). ‘‘In Situ
stress determination by the Iowa Stepped Blade.’’ ASCE J. Geotech. Eng. Div. 108(11),
1405–1422.
Handy, R. L., and White, D. J. (2006). ‘‘Stress Zones Near Rammed Aggregate Piers:
I. Plastic and Liquefied Behavior; II. Radial Cracking and Wedging.’’ ASCE J.
Geotech. and Geonenviron Eng. 132(1), 54–71.
Hvorslev, M. J. (1949). Subsurface Exploration and Sampling of Soils for Civil Engineering
Purposes. Waterways Experiment Station, U.S. Army Corps of Engineers, Vicksburg,
Miss.
Kelley, A. M., and Lutenegger, A. J. (2004). ‘‘Unit Skin Friction from the Standard
Penetration Test Supplemented with the Measurement of Torque.’’ ASCE J. Geotech.
and Geoenviron. Eng. 130(5), 540–3.
Liao, S., and Whitman, R. V. (1986). ‘‘Overburden correction factors for SPT in sand.’’
ASCE J. Geotech. Eng. 112(3), 373–7.
Lutenegger, A. J., and Tierney, K. F. (1986). ‘‘Pore Pressure Effects in Borehole Shear
Testing.’’ Use of In Situ Tests in Geotechnical Engineering, ASCE Geotech. Spec. Pub.
No. 6, 752–764. ASCE, New York.
Marchetti, S. (1980). ‘‘In situ by flat Dilatometer’’J. Geotech. Eng. Div. ASCE. 106(3),
299–321.
Meigh, A. C. (1987). Cone Penetration Testing. Butterworths, London.
Meyerhoff, G. G. (1956). ‘‘Penetration Tests and Bearing Capacity of Cohesionless Soils.’’
ASCE J. Soil Mechanics and Foundation Eng. Div. 182(SM1), 1–19.
Meyerhoff, G. G. (1976). ‘‘Bearing Capacity and Settlement of Pile Foundations.’’ ASCE
J. Geotech. Eng. Div. 102(GT3), 195–228.
Miller, G. A., Azad, S., and Hassell, C. E. (1998). ‘‘Iowa Borehole Shear Testing in
Unsaturated soil.’’ In Geotechnical Site Characterization 2, 1321–1326. Balkema,
Rotterdam.
Robertson, P. K., and Campanella, R. G. (1983). ‘‘Interpretation of Cone Penetration
Tests.’’ Canadian Geotech. J. 20, 718–745.
Robertson, P. K., Campanella, R. G., Gillespie, D., and Greig, J. (1986). ‘‘Use of
Piezometer Cone Data.’’ Use of In Situ Tests in Geotechnical Engineering. ASCE
Geotech. Spec. Pub. No. 6, 1263–1280. ASCE, New York.
Schmertmann, J. (1975). ‘‘Measurement of in Situ Shear Strength.’’ In Situ Measurement of
Soil Properties II, 57–138. ASCE, New York.
Schmertmann, J. (1979). ‘‘Statics of SPT.’’ ASCE J. Geotech. Eng. Div. 105(5), 655–670.
Schmertmann, J. H. (1986). ‘‘Dilatometer to compute foundation settlement.’’ Use of In
Situ Tests in Geotechnical Engineering, 303–321. ASCE, New York.
Takesue, K., Sasao, H., and Matsumoto, T. (1998). ‘‘Correlation Between Ultimate Pile
Skin Friction and CPT Data.’’ Geotechnical Site Characterization 2, 1177–1182.
Balkema, Rotterdam.


26 - in Situ Soil and Rock Tests

Uploaded by

Copyright:

Available Formats

26 - in Situ Soil and Rock Tests

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

26 - in Situ Soil and Rock Tests

Uploaded by

Copyright:

Available Formats

Source: GEOTECHNICAL ENGINEERING

26 In Situ Soil and Rock Tests

26.1.1 Probing for Answers

780 Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

In Situ Soil and Rock Tests 781

Another type of in-situ testing is nondestructive and substitutes electrical,

The reliability of in-situ tests is evaluated by comparisons with more traditional

26.1.2 Exploration Drilling

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

782 Geotechnical Engineering

26.1.3 Undisturbed Sampling

A measure of sample disturbance is the ‘‘area ratio,’’ defined by Hvorslev (1949)

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

In Situ Soil and Rock Tests 783

26.1.4 Determining the Character of Soil Between

26.1.5 Groundwater Table

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

784 Geotechnical Engineering

26.2 STANDARD PENETRATION TEST (SPT)

26.2.1 Redefining the SPT

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

In Situ Soil and Rock Tests 785

The rope-and-cathead lifting procedure requires operator skill and introduces

26.2.2 Overburden Pressure

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

786 Geotechnical Engineering

where CN is the correction factor and v0 is the effective overburden stress in

26.2.3 What Does the SPT Measure?

In Situ Soil and Rock Tests 787

26.2.4 Relationships to Bearing Capacity and Settlement

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

788 Geotechnical Engineering

depth factor is included, N values should not be corrected for overburden

qo ¼ ðB=2ÞN þ DNq ¼ ð125 lb=ft3 2 ft=2Þð29Þ

Qa ¼ 4 ft2 3500 lb=ft2 ¼ 14,000 lb ¼ 7 tons ð62 kNÞ

Settlement of foundations on sand is discussed in section 16.10.3. Schmertmann’s method is

qc ¼ 4N ¼ 4ð7Þ ¼ 28 tons=ft2 ð250 kNÞ

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

In Situ Soil and Rock Tests 789

26.2.5 Deep Foundations on Sand

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

790 Geotechnical Engineering

Answer: End bearing cannot exceed 4N ¼ 4(18) ¼ 72 tons/ft2 (7 MPa)

Side friction: fs ¼ 18/50 ¼ 0.36 tons/ft251 ton/ft2.

Multiplying by the resective areas,

With a factor of safety of 2,

26.2.6 SPT in Clay

26.3 CONE PENETRATION TESTS

26.3.1 Push vs. Drive-Cone Tests

26.3.2 Static Cone Tests

In Situ Soil and Rock Tests 791

Static cone penetrometers have an independently supported ‘‘friction sleeve’’ that

The standard rate of advance is 20 mm/s, and measurements are made by

792 Geotechnical Engineering

26.3.3 Influence of Depth and Lateral Stress

Combined influences of depth and lateral stress were examined theoretically

26.3.4 Relation of Soil Type to Cone Data

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)

In Situ Soil and Rock Tests 793