Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Impact of a spacing reduction in a fall cone test
Technique d’essai répétée du cone de chute
R. Khalili, C. Jommi
Department of Civil and Environmental Engineering, Politecnico di Milano, Milan, Italy
W.T. Sołowski*
Department of Civil Engineering, Aalto University, Espoo, Finland
*wojciech.solowski@aalto.fi
ABSTRACT: The paper investigates an adjustment to the fall cone test, where the same soil sample is reused for four extra
tests. The analysis shows that the fall cone test inaccuracies are much higher than the effect of reusing the sample. Therefore,
the proposed procedure may help to establish the soil properties more accurately without much extra effort and reduce the
number of samples needed for testing.
RÉSUMÉ: L'article étudie un ajustement au test du cône de chute, où le même échantillon de sol est réutilisé pour quatre
tests supplémentaires. L’analyse montre que les imprécisions du test du cône de chute sont bien supérieures à l’effet de la
réutilisation de l’échantillon. Par conséquent, la procédure proposée peut aider à établir les propriétés du sol avec plus de
précision sans trop d’effort supplémentaire et à réduire le nombre d’échantillons nécessaires aux tests.
Keywords: Fall cone test; statistical analysis; marine clay; sensitive clay.
1 INTRODUCTION
The fall cone test was initially devised by the Swedish
state railways during the period from 1914 to 1922 and
subsequently gained widespread adoption in
Scandinavian countries (Hansbo, 1957). Currently, the
ISO 17892-6:2017 standard contains the details of the
procedure.
In general, the fall cone test requires that before its
release the metallic cone is in direct contact with the
surface sample. After the release, the cone penetrates
the specimen. The undrained shear strength is
correlated with the measured depth of penetration. The
test relies on the lack of friction in the apparatus, hence
the apparatus should be lubricated to minimise
friction. The test accuracy is related to a number of
factors, the interested reader may see Llano-Serna &
Contreras (2020) for discussion on cone roughness
influence and further references.
2 RESEARCH HYPOTHESIS
The ISO 17892-6:2017 standard requires that the test
points should be distributed so that the distance
between the outer boundaries of the cone penetration
in two tests is at least 14mm, and the cone penetration
closest to the perimeter is at least 7mm. For soft marine
clays, for 60 deg 60g cone the penetration can reach
9mm, leading to the spacing requirement of 24,4 mm
between centres of penetration and 12,2 mm from the
sample perimeter. For common core diameters 50, 54
and – used in this study – 58 mm, these limitations lead
to 3 sampling points on the single cross-section of the
sample. Three points are the minimum required by ISO
17892-6:2017. However, ISO 17892-6:2017 requires
that any test result deviating from the mean penetration
by more than 0.5mm should be excluded. This means
that often an extra point is needed, however, due to
distance restriction, no more sampling points are
allowed on the sample and another sample for testing
should be taken. This may be difficult and affects the
testing programme.
Recent research (e.g Mohapatra et al., 2023a;
Mohapatra et al., 2023b; Tran & Sołowski, 2019), has
indicated that the impact of the fall cone penetration
into a soft sensitive clay is contained to approximately
double the radius of the 60-degree fall cone
penetration mark, see Figure 1. Furthermore, the
results have shown little sensitivity to lateral support,
with no changes for full lateral support or lack of it.
Based on those results we decided to check the
hypothesis that a reduction of the spacing of the
sampling points to double the radius of the penetration
mark may have no influence on the results, while it
allows for the increase of the number of sampling
points on a standard 50 to 58mm sample cross-section,
potentially increasing the accuracy of the findings.
Additionally, the work aims to evaluate whether the
Proceedings of the XVIII ECSMGE 2024
GEOTECHNICAL ENGINEERING CHALLENGES
TO MEET CURRENT AND EMERGING NEEDS OF SOCIETY
© 2024 the Authors
ISBN 978-1-032-54816-6
DOI 10.1201/9781003431749-312
Open Access: www.taylorfrancis.com, CC BY-NC-ND 4.0 license
1679
Proceedings of the XVIII ECSMGE 2024
C – Risk analysis and safety evaluation
results of the test are statistically influenced by
removing the lateral support. If the denser spacing or
lack of lateral support has an effect, the depth of the
penetrations closer to the sides of the sample should be
higher compared to the first penetration in the middle
of the sample.
~10 mm
5-8 mm
24 mm
min 8 mm
Figure 1. Numerical replication of a fall cone test in soft
marine clay showing localised deformations around the
cone, Mohapatra et al. (2023b).
3 CLAY SAMPLES
In this research, tested soil samples had a diameter of
58 mm and a height of 10 cm. The penetration test was
conducted using a cone with a 60° angle and a mass of
60 grams. The observed penetration depths across all
samples range approximately from four to 12 mm.
In this study, a total of 39 normally consolidated
marine clay samples were tested with a fall cone. The
samples came from four locations at the offshore Kytö
site: KU2 (with depths ranging from 0.1 to 3.87
meters), KU3 (with depths ranging from 0 to 3.48
meters), KU1 (with depths ranging from 0 to 3.1
meters), and KU4 (with depths ranging from 0.15 to
3.75 meters). For a more detailed site description see
Saresma et al. (2023), while for more details on
properties and tests see Li et al. (2023).
The composition of these marine sediments
exhibits some variations primarily attributed to the
different stages of the Baltic Sea's evolutionary
history. However, the stratification of these sediments
is relatively weak and the samples were very similar.
The clays are structured and sensitive. Their intact
undrained shear strength varies, mainly due to
differences in depth from which the samples originate.
4 METHODS
4.1 Fall cone test procedure
Figure 2 shows the position of each penetration for
each sample in the testing programme, with numbers
indicating the penetration sequence, leading to the
spacing between the sampling points smaller than
required by ISO 17892-6:2017. The sample used had
a 58 mm diameter and the points 2-5 were positioned
at least 10 mm from the side of the sample, with the
impact point approximately in the middle of the line
connecting the edge of the penetration 1 mark and
sample perimeter, i.e. around 10-14mm from the
sample perimeter. This means that the distance
between the edge of the penetration mark was around
5-8mm to the sample perimeter, but there could be
only about 8 mm between the edges of the penetration
mark at test 1 and the subsequent test, instead of the
required 14mm. Nonetheless, the spacing always was
approximately 2 penetration radii, complying with the
condition identified in numerical analyses. Besides the
spacing, the testing procedure followed ISO 17892-
6:2017, with tests done on thick 10cm extruded from
plastic tubes. There was no lateral support at the sides.
Figure 2. Position of each penetration.
4.2 Discussion of the testing procedure
An alternative would be to perform 3 tests according
to ISO 17892-6:2017 and then one extra test in the
middle with smaller distances to the other tests.
However, that would lead to a distance between the
middle test mark and any other test as low as 5mm, in
which case the penetration distortion zones would
likely overlap.
4.3 Statistical analysis
In case the spacing of the points was too close, the
subsequent penetrations should be affected due to the
disturbance of soil and the associated reduction in the
1680 Proceedings of the XVIII ECSMGE 2024
Impact of a spacing reduction in a fall cone test
undrained shear strength. Hence, in case the soil is
disturbed, we expect that there will be a statistically
significant difference between the results obtained in
the first penetration (middle point) and the subsequent
4 penetrations. Similarly, in case the lack of lateral
constraint affects the penetration depth, the statistical
analysis should reveal that.
Even though ISO 17892-6:2017 decides on the
acceptance or rejection of the test based on
penetration, in this contribution we decided to
concentrate on the statistical analysis of the undrained
shear strength. This is because the undrained shear
strength is the key value obtained in the test and the
important value for engineering. The results of the
analysis performed based on the penetration depths are
similar, but not included in this paper.
To analyse the data, first, all the undrained shear
strengths were normalised with respect to the ‘correct’
penetration in the middle of the sample.
where sn,i is the normalised value of the undrained
shear strength obtained from the i-th penetration with
respect to the value s1 obtained from the first
penetration. This assumes the middle point is the
correct one, with the four subsequent penetrations
possibly deviating from it. First, we tried analysing the
data for each penetration. Then, as the number of tests
was small, we decided to pool the data together.
In this case, all the tests besides the first one are
treated the same, leading to 39 sets of 4 deviating data
points. Then, we first assessed whether the obtained
data follows a normal distribution. Later, we checked
whether the undrained shear strengths from the
subsequent 4 penetrations deviated from the undrained
shear strength obtained from the test done at the centre
of the sample using the Student’s t-test. The null
hypothesis for the test is that the differences in the
averages obtained from the other penetrations are not
statistically significant when compared to the first
penetration.
Finally, we also assessed whether the data with
outliers removed (defined as data points with
undrained shear strength deviating from the value
obtained in the middle point by a certain percentage)
would lead to any statistically significant result. The
outlier cut-offs were 250% and 40% (marked 2.5),
200% and 50% (marked 2), 150% and 66.7% (marked
1.5) and 125% and 80% (marked 1.25).
We also assessed the data with tests following the
acceptance/rejection strategy similar to that in ISO
17892-6:2017. In this case, if in the first three
penetrations, there was a value deviating by more than
0.5mm, the value most deviating from the average was
discarded (excluding the test from the middle of the
sample) and another value was taken to compute the
average.
5 RESULTS
5.1 Statistical distribution of the normalised
data and testing of the hypothesis
The data is first normalized with respect to the data
from the first penetration. Assuming a normal
distribution for data, data obtained for each penetration
has a very high standard deviation, see Table 1. The
difference in mean value, vs the middle penetration, is
between 0,68% to 7.84%, while the standard deviation
is between 30,76% to 78.08%. These are very high
values, indicating much scatter in the data. In this case,
for each penetration point, we had 39 tests. The results
of the Student’s t-test show that the differences in
obtained undrained shear strengths for each test point
location are not statistically significant. The t-test
suggests that the mean values differences are not
statistically meaningful and cannot be taken as
evidence that the adjusted test procedure leads to
differences in the obtained values of undrained shear
strength. Unfortunately, the amount of data points for
each penetration location is not sufficiently large to
make a meaningful discussion about whether the
assumption of normal distribution is correct.
Table 1. Mean value, standard deviation and t-test result for
each penetration location.
Location
Mean
Standard
deviation
T-test
Significant?
2
1,0186
0,3869
0,3005
NO
3
1,0784
0,7808
0,6267
NO
4
0,9932
0,3076
0,1389
NO
5
1,0730
0,4614
0,9874
NO
To increase the amount of data, we decided to pool
the data from all the points together, leading to 155
data points, due to cutting one outlier (see Figure 3).
This full data, as well as data with cut outliers, is
assessed for the normal distribution based on
histograms and QQ plots, see Figures 4 and 5. The
plots are similar and suggest that the normal
distribution is an acceptable assumption. The results of
a statistical analysis based on the assumption that the
data follows normal distribution are in Table 2. The
same table contains also mean and standard deviation
values, as well as the Student‘s t-test values.
An alternative is to use lognormal distribution, with
the QQ plots, see Figure 6. Both distributions are
visually confirmed based on histograms, see Figure 7,
1681
Proceedings of the XVIII ECSMGE 2024
C – Risk analysis and safety evaluation
with perhaps lognormal distribution being a slightly
better assumption.
The full data, as well as data prepared according to
ISO 17892-6:2017, and data with outliers cut, were
tested against the null hypothesis using Student’s t-
test. We checked both normal and lognormal
distributions for statistical significance for the
hypothesis that the undrained shear strength results
obtained from tests done at the sides of the sample are
statistically meaningfully different to the undrained
shear strength values obtained from the tests done in
the middle of the sample, Tables 1-3 contain the
analysis results.
Table 2. Mean value, standard deviation and t-test values
for the combined data, normal distribution assumption.
Cut-
off
Data
points
Mean
Stand-
ard de-
viation
T-test
Sig-
nifi-
cant?
No
155*
1.041
0.5127
0.3221
No
ISO
77
1.033
0.2349
0.2150
No
2.5
153
1.001
0.3389
0.0505
(No)
2
146
0.979
0.2713
0.0114
No
1.5
121
1.007
0.1881
0.5530
No
1.25
88
1.006
0.1153
0.6464
No
Cutoff – preparation of the data, with the cut-off indicating
cut-off values – 2.5- values below 40% and above 250% of
the average value cut, 2 – 50% and 200%, 1.5 – 150% and
66,7%, ISO – according to ISO standard/ N- assumed
normal distribution, L – lognormal. T-test confirms the
hypothesis when p<0.05. *one point cut out, see Figure 3
for an explanation.
5.2 Analysis of the results
The results confirm that there is no statistically
meaningful effect indicating that the undrained shear
strength obtained for the samples at the sides is
different from those in the middle. The confirmation
of the hypothesis for the cut-off value of 2.0 is due to
the different number of points cut from both sides of
the distribution. The results for this case, see Figure 5,
indicate that the penetration depths at the sides are
smaller than in the middle, which has no physical
sense. This perhaps shows how tricky it is to cut the
outliers in the data and that such action should be
avoided, as it may lead to incorrect results.
Figure 3. Histogram of all the data. The outlier pointed out
by the arrow is cut off for analysis, as it is a clear testing
error (insufficient penetration due to excessive friction in
the apparatus).
Figure 4. Histogram of data with outliers cut off (cut of
value 40% and 250% of the average value), shown vs
normal distribution (top) and lognormal distribution
(bottom) In lognormal distribution numbers correspond to
standard deviations.
1682 Proceedings of the XVIII ECSMGE 2024
Impact of a spacing reduction in a fall cone test
Figure 5. QQ plots of data assuming normal distribution:
full data (top), data prepared according to ISO 17892-
6:2017 (middle) and with outliers cut off (cut of value 40%
and 250% of the average value)(bottom).
Figure 6. QQ plot of data vs lognormal distribution line with
outliers cut off (cut of value 40% and 250% of the average
value).
Figure 7. Histogram of data with outliers cut off at 50% and
200% of the average value, shown vs lognormal
distribution, with numbers indicating standard deviations.
Data is skewed towards lower values of undrained shear
strength due to the cut-off value selected.
Table 3. Student’s t-test values, lognormal distribution.
Cut-off
t-test p-value
Confirmed (Y/N)
No
0.2484
N
ISO
0.7120
N
2.5
0.0505
(N)
2
0.0114
Y
1.5
0.5530
N
1.25
0.9479
N
Cutoff – preparation of the data, with the cut-off indicating
cut-off values – 2.5- values below 40% and above 250% of
the average value cut, 2 – 50% and 200%, 1.5 – 150% and
66,7%, ISO – according to ISO standard/ N- assumed
normal distribution, L – lognormal. T-test confirms the
hypothesis when p<0.05.
6 CONCLUSIONS
The data we have does not support the hypothesis that
the undrained shear strength values from tests on the
1683
Proceedings of the XVIII ECSMGE 2024
