Sei sulla pagina 1di 10

Correlation between Shear Wave

Velocity and SPT-N Values for North


Florida Soils
Milad Fatehnia
Geotechnical Engineer
Environmental and Geotechnical Specialists (EGS)
e-mail: mfatehnia@fsu.edu

Myron Hayden
Principal Geotechnical Engineer
Environmental and Geotechnical Specialists (EGS)
e-mail: myron.hayden@egs-us.com

Matthew Landschoot
Geotechnical Engineer
Environmental and Geotechnical Specialists (EGS)
e-mail: matt.landschoot@egs-us.com

ABSTRACT
Multichannel analysis of surface wave (MASW) is a seismic method that uses surface waves
(specifically fundamental mode Rayleigh Waves) to estimate shear wave velocities (Vs). Along
with shear wave, many design parameters of soil are associated with the standard penetration test
(SPT) as a dynamic in situ experiment. Both SPT-N data and geophysical data do not often exist in
the same area. Statistical analysis of correlation between these parameters is an alternate method
to estimate Vs conveniently and without additional investigations or data acquisition. In this
research, a new empirical formulae that can be used to correlate SPT-N and Vs values for the
typical soil materials encountered in north Florida is suggested. For this purpose, soil classification
information, shear wave velocity (derived from MASW method), and SPT-N values are gathered
from four geotechnical and geophysical investigations conducted in this region. By employing
M5 model tree algorithm, the relation between Vs and SPT-N values is predicted using the
collected dataset. The accuracy of the new proposed equation is determine by measuring the
correlation coefficient. The proposed correlation is also compared with the previously suggested
formulas of Vs determination by measuring Root Mean Square Error of each formula.

KEYWORDS:

MASW, SPT, shear wave velocity, M5 model tree

INTRODUCTION
Multichannel analysis of surface wave (MASW) is a seismic method that uses surface waves
(specifically fundamental mode Rayleigh Waves) for estimation of the shear wave velocities (Vs)
(Park et al., 1999). Vs is primarily a function of soil density, void ratio, and effective stress, with
secondary influences including soil type, age, depositional environment, cementation, and stress
history (Hardin and Drnevich 1972a, b). Considering the vertical velocity variation of the sites
(sand to clay to rock), different propagation velocity (phase velocity) for each frequency
component of the surface wave is expected. Shear wave velocity is an important soil parameter
- 12421 -

Vol. 20 [2015], Bund. 22

12422

that can be used in a broad range of applications, including mapping stratigraphic layers,
conducting pre-construction site characterization studies, calculating dynamic properties of soils,
evaluating the liquefaction potential, and detecting cavities, tunnels, and sinkholes (Seed et al.,
1983; Leparoux et al., 2000; Thitimakorn and Channoo, 2012).
Along with shear wave, many design parameters of soil are associated with the standard
penetration test (SPT) as a dynamic in situ experiment. SPT-N values are most widely used to
describe soil characteristics. Both SPT-N data and geophysical data do not often exist in the same
area. Statistical analysis of correlation between these parameters is an alternate method to
estimate Vs conveniently and without additional investigations or data acquisition (Akin et al.,
2011). This correlation can be used to estimate geotechnical parameters in areas where no
geophysical investigations exist. There is no theoretical relationship between destructive (e.g.
SPT) and non-destructive methods (e.g. seismic methods). However, several empirical
relationships exist for different soil types which appear to be site dependent. Jafari et al. (2002)
presented a detailed historical review on the statistical correlation between SPT-N versus Vs.
With an exception of Lee (1990), almost all of the studies focused on the relationships between
uncorrected SPT-N and Vs for all soils as well as sand and clay-type soils.
In this research, a new empirical formula that can be used to correlate SPT-N and Vs values
for the typical soil materials encountered in North Florida is suggested. For this purpose, soil
classification information, shear wave velocity (derived from MASW method), and SPT-N values
of four geophysical and geotechnical investigations conducted in North Florida were gathered. By
employing M5 model tree algorithm, the relation between Vs and SPT-N values is predicted
using the collected dataset. The accuracy of the new proposed equation is measured and a
comparison with the previously suggested formulas for Vs determination is also conducted.

CORRELATIONS BETWEEN SPT-N AND VS


A significant body of research can be found in the literature for correlations between SPT-N
and Vs. A summary of proposed empirical correlations is given in Table 1. The correlations are
established for three groups, namely all soil types, cohesionless soils and cohesive soils. The
variation between suggested correlations is mainly due to the various geotechnical conditions of
each studied site and also different equipment and methods of field investigations. In this
research, using all 300 collected data, a single equation for cohesive and cohesionless soils of
North Florida is suggested and its precision in estimating the shear wave velocity is compared
with the12 proposed correlations.

GEOPHYSICAL INVESTIGATIONS
Data acquisitions for all selected sites were accomplished using a Geometrics Geode, 24channel digital seismograph. A sledge hammer striking a polyethylene plate (shot point) was used
to generate a surface wave. The propagation velocities of that wave were recorded using an array
of low frequency geophones (4.5 Hertz). The geophones were spaced at 2 to 3-ft intervals using
3.5 inch coupling spikes and/or metal plates. For the MASW profile, after every shot, the shot
point and geophone array was advanced 1 geophone interval and a new shot was recorded. Data
for each shot point is recorded at the midpoint of the geophone spread. Due to this constraint, it is
not possible to collect subsurface data within 25 to 50 ft (approximate midpoint) of the actual
start or end of the line if the lines are abutting buildings or other obstacles. Photograph of one of
the MASW transect lines is illustrated in Figure 1.

Vol. 20 [2015], Bund. 22

12423

Table 1: Summary of proposed correlations between SPT-N and Vs


Author(s)

Soil Type

Kanai (1966)
Fujiwara(1972)

All Soils
All Soils
All Soils
Sand
Clay
All Soils
All Soils
All Soils
All Soils
All Soils

Imai (1977)
Imai and Tonouchi (1982)
Yokota et. al (1991)
Athanasopoulos (1995)
Iyisan (1996)
Kiku et. al (2001)
Anbazhagan and
Sitharam (2006)
Hasancebi and Ulusay
(2006)
Maheshwari et. al (2008)
Dikmen (2009)

All Soils
All Soils
Sand
Clay
All Soils
All Soils
Sand
Silt
Clay

Correlation between
SPT-N and Vs (m/Sec)
= 19 0.6
= 92.1 0.337
= 91 0.337
= 80.6 0.331
= 102 0.292
= 96.9 0.314
= 121 0.27
= 107.6 0.36
= 51.5 0.516
= 68.3 0.292
= 50(60 )0.41

= 90 0.309
= 90.82 0.319
= 97.89 0.269
= 95.64 0.301
= 58 0.39
= 73 0.33
= 60 0.36
= 44 0.48

During the processing of the seismic data, dispersion curves were calculated from the data of
each shot point (record). The dispersion curves show the phase velocity of the surface waves as a
function of frequency. 1-D profiles of the shear wave velocity (Vs) versus depth was then back
calculated from the dispersion curve. These data were used to record Vs values at the location of
each SPT. Finally, a series of 1-D shear wave velocity profiles were combined to produce a 2-D
cross-section of the shear wave velocities along the seismic line.
The MASW data were analyzed using SurfSeis, a computer processing program developed by
Kansas Geological Survey of Lawrence, Kansas. The program was used to calculate a twodimensional vertical cross section of the approximate shear wave velocities under the MASW
transect. One of the velocity cross sections is shown in Figure 2. The velocity cross section shows
the calculated shear wave velocity along the transect line.

Vol. 20 [2015], Bund. 22

12424

Figure 1: MASW transect line

Figure 2: MASW Calculated Velocity Cross Section

GEOTECHNICAL INVESTIGATIONS
In order to evaluate the relative strength and uniformity of the subsurface materials, standard
penetration tests (SPT) were conducted at the locations of geophysical investigations. The tests
were performed with an automatic hammer in accordance with ASTM D1586. 300 SPT data were
collected from 28 soil borings. In all soil borings, SPT data were conducted continuously from
the ground surface to a depth of fifteen feet using rotary drilling rig. From there, SPTs were
conducted on two and one-half (2 ) feet intervals to the termination depth.

Vol. 20 [2015], Bund. 22

12425

DATA ANALYSIS
In this research, M5 model tree algorithm was used to predict the relation between Vs and
SPT-N values. M5 model tree, is an improvement of M5 model trees. Model trees are kind of
machine-learning techniques with the idea of splitting the parameter space into areas (subspaces)
and building a linear regression model in each of them. M5 model has been previously applied in
different geotechnical and geoenvironmental areas including modeling water level-discharge
(Bhattacharya and Solomatina 2005), water movement in sandy soils (Samadianfard et al. 2014),
and saturated hydraulic conductivity estimation in double ring infiltrometer test (Fatehnia, 2015).
The main assumption that the trees-structured numerical prediction is built on, is that the
functional dependency is not constant in the whole domain but can be approximated on smaller
sub domains. These trees can be generalized to regression and modal trees and are capable of
dealing with continuous attributes (Naeej et al. 2013). Model trees, like regression trees, are
efficient for large data sets. M5 model tree was first introduced by Quinlan (1992). The method
was later improved to M5 by Wang and Witten (1997) in order to enhance its capability in
handling enumerated attributes and missing values of the data set. To solve a problem, M5 model
tree algorithm constructs a regression tree by recursively splitting the instance space. It uses a
splitting condition to minimize the intra-subset variability in the values down from the root
through the branch to the node. The standard deviation of the values that reach that node from the
root through the branch is used to measure the variability with calculating the expected reduction
in error as a result of testing each attribute at that node. The criteria used to choose the attribute is
maximizing the expected error reduction. The splitting process is stopped if the output values of
all the instances that reach the node vary only slightly or only a few instances remain (Naeej et al.
2013). The standard deviation reduction (SDR) is calculated as:
= ()

| |
||

( )

(1)

where is the set of data point before splitting, is data point that is the result of splitting
the space and falls into one subspace based on the chosen splitting parameter, and is the
standard deviation. Standard deviation is used to measure the error of the data points of a
subspace. M5' model tree tests different splitting points by measuring for sub-spaces before
splitting the space. As is maximized in a point, the point is selected as the splitting point
(node). When change is less than a certain value or a few data points remain in sub-space,
the splitting terminates. As tree grows, the accuracy of the model for training set uniformly
increases. However, over-fitting may be inevitable while a model tree is being built. Hence, in the
next step, pruning is adopted to avoid over-fitting (Witten and Frank 2005). Smoothing, as the
final step, is the regularization process to compensate any probable discontinuities within
adjacent linear models (Jang et al. 2009; Pfahringer et al. 2001).
In this research, WEKA program was applied for setting up the model based on existing 300
data by using 70 % of the data to predict the correlation. The results and error parameters are
presented in the Results and Discussion section.

RESULTS AND DISCUSSION


By adopting M5 model trees and using natural logarithm of all 300 Vs and SPT-N data, Eq.
2 was offered by the program. The program randomly selects 70% of the entire data for
correlation prediction and uses the remaining 30% to measure the error values. Correlation
coefficient of 0.893 for the proposed formula shows an accurate estimation of shear wave

Vol. 20 [2015], Bund. 22

12426

velocity. This equation can be used to predict Vs of both cohesive and cohesionless soils from
their respective SPT-N values. Since the data used in this analysis were gathered from the
projects performed in North Florida, they can better predict the local material properties.
= 77.1 0.355

(2)

In this equation, Vs is the shear wave velocity in (m/Sec) and N is the uncorrected SPT
values. Figure 3 illustrates the measured data of SPT-N and Vs values together with the proposed
correlation derived using M5 model tree algorithm.

400

Shear Wave Velocity (m/Sec)

350
300
250
200

Fatehnia et.
al (2015)

150

Measured
Data

100
50
0
0

10

20

30
40
SPT "N" VALUES

50

60

Figure 3: Field data and the proposed correlation derived using M5 tree model
The accuracy of equation 2 in predicting the shear wave velocity is compared with the other
12 previously suggested correlations. Figure 4 illustrates the Comparison of the proposed formula
with previous correlations between SPT-N and Vs values.
In order to compare the performance of equation 2 in predicting the shear wave velocity of
the collected data, Root Mean Square Error (RMSE) values of the proposed correlation was
compared with the other formulas. As shown in table 2, equation 2 with RMSE of 26.50 (m/Sec)
had lower error value than all the other suggested correlations studied in this research.

Vol. 20 [2015], Bund. 22

12427

600
Kanai (1966)
Fujiwara (1972)

Shear Wave Velocity (m/Sec)

500

Imai (1977)
Imai and Tonouchi
(1982)
Yokota et. al (1991)

400

Athanasopoulos
(1995)
Iyisan (1996)

300

Kiku et. al (2001)


Anbazhagan and
Sitharam (2006)
Hasancebi and
Ulusay (2006)
Maheshwari et. al
(2008)
Dikmen (2009)

200

100

Fatehnia et. al
(2015)

0
0

20

40
60
SPT "N" VALUES

80

100

Figure 4: Comparison of the proposed formula with current correlations between SPT-N
and Vs
Table 2: Comparison of the RMSE values of the studied correlations with equation 2
Formula
Fatehnia et. al (2015)
Kanai (1966)
Fujiwara(1972)
Imai (1977)
Imai and Tonouchi (1982)
Yokota et. al (1991)
Athanasopoulos (1995)
Iyisan (1996)
Kiku et. al (2001)
Anbazhagan and Sitharam (2006)

RMSE (m/sec)
26.50
96.70
37.75
36.02
36.65
56.17
82.56
39.39
53.71
51.27

Hasancebi and Ulusay (2006)


Maheshwari et. al (2008)
Dikmen (2009)

28.60
31.69
40.22

Vol. 20 [2015], Bund. 22

12428

CONCLUSIONS
A new empirical formulae that can correlate SPT-N and Vs values for the typical soil
materials encountered in North Florida is suggested using M5 model tree algorithm. For this
purpose, 300 data of shear wave velocity (derived from MASW method) and respective SPT-N
values were gathered from four geotechnical and geophysical investigations conducted in this
region. The proposed formula with correlation coefficient of 0.893 had an acceptable estimation
of shear wave velocity. The accuracy of the new equation in predicting the Vs values of the
collected data was also compared with the previously suggested formulas by measuring the
RMSE value of each method. The proposed formula with RMSE value of 26.50 had better
performance in predicting Vs values compared to other equations.

ACKNOWLEDGEMENTS
The data used in this research was provided by Environmental and Geotechnical Specialists
(EGS).

REFERENCES
1. Akin, M.K., Kramer, S.L. and Topal, T., (2011). Empirical correlations of shear wave
velocity (Vs) and penetration resistance (SPT-N) for different soils in an earthquake
prone area (Erbaa-Turkey). Engineering Geology. 119, 117.
2. Anbazhagan, P. and Sitharam, T.G. (2006). Evaluation of dynamic properties and
ground profiles using MASW: correlation between V s and N 60 Proceedings of 13th
Symposium on Earthquake Engineering, Indian Institute of Technology, Roorkee,
December, 18-20.
3. ASTM D1586-11, Standard Test Method for Standard Penetration Test (SPT) and SplitBarrel Sampling of Soils, ASTM International, West Conshohocken, PA, 2011,
www.astm.org
4. Athanasopoulos, G. A. (1995). Empirical correlations Vs-NSPT for soils of Greece: a
comparative study of reliability, Proc. 7th Int. Conf. on Soil Dynamics and Earthquake
Engineering (Chania, Crete) ed. A S Cakmak (Southampton: Computational Mechanics),
1936.
5. Bhattacharya, B. & Solomatine, D.P. (2005). Neural networks and M5 model trees in
modelling water level-discharge relationship, Neurocomputing, 63, 381-396.
6. Dikmen, U. (2009). Statistical correlations of shear wave velocity and penetration
resistance for soils, Journal of Geophysics and Engineering, 6, 61-72.
7. Fujiwara, T. (1972). Estimation of ground movements in actual destructive
earthquakes, Proc. 4th European Symp. Earthquake Engineering (London), 12532.
8. Fatehnia, M. (2015). Automated Method for Determining Infiltration Rate in Soils,
Electronic Theses, Treatises and Dissertations, Paper 9327.
9. Hardin, BO, and VP Drnevich (1972a). Shear modulus and damping in soils:
measurement and parameter effects, J. Soil Mech. and Foundations Div., SM 6:603624.
10. Hardin, BO, and VP Drnevich (1972b). Shear modulus and damping in soils: design
equations and curves, J. Soil Mech. and Foundations Div., SM 7:667692.

Vol. 20 [2015], Bund. 22

12429

11. Hasancebi, N. and Ulusay, R. (2006). Empirical correlations between shear wave
velocity and penetration resistance for ground shaking assessments, Bull Eng Geol
Environ, 66, 203-213.
12. Imai, T. (1977). P-and S-wave velocities of the ground in Japan, Proc.9th Int. Conf. on
Soil Mechanics and Foundation Engineering, vol 2, 12732.
13. Imai, T. and Tonouchi, K. (1982). Correlation of N-value with S-wave velocity and
shear modulus, Proc. 2nd European Symp. Of Penetration Testing (Amsterdam), 5772.
14. Iyisan, R. (1996). Correlations between shear wave velocity and in-situ penetration test
results, Tech. J. Chamber Civil Eng. Turkey, 7, 118799 (in Turkish).
15. Jafari, M.K. Shafiee, A. and Ramzkhah, A. (2002). Dynamic properties of the fine
grained soils in south of Tehran, J. Seismol. Earthq. Eng. 4 2535
16. Jung, N.C., Popescu, I., Kelderman, P., Solomatine, D.P., & Price, R.K. (2009).
Application of model trees and other machine learning techniques for algal growth
prediction in Yongdam reservoir, Republic of Korea. Journal of Hydroinformatics, 12,
262-274.
17. Kanai, K. (1966). Conf. on Cone Penetrometer The Ministry of Public Works and
Settlement (Ankara, Turkey), (presented by Y Sakai 1968).
18. Kiku, H., Yoshida, N., Yasuda, S., Irisawa, T., Nakazawa, H., Shimizu, Y., Ansal, A. and
Erkan, A. (2001). In-situ penetration tests and soil profiling in Adapazari, Turkey,
Proc. ICSMGE/TC4 Satellite Conf. on Lessons Learned from Recent Strong
Earthquakes, 25965.
19. Lee, S.H. (1990). Regression models of shear wave velocities, J. Chin. Inst. Eng. 13
51932
20. Leparoux, D., Bitri, A. and Grandjean, G. (2000). Underground cavities detection: a new
method based on seismic Rayleigh waves, European Journal of Environmental and
Engineering Geophysics. 5, 3353.
21. Maheswari U R, Boominathan A and Dodagoudar G R (2008) "Development of
empirical correlation between Shear wave velocity and Standard Penetration Resistance
in soils of Chennai City", Proceedings of the 14th World Conference on Earthquake
Engineerin ,Beijing, China
22. Naeej, M., Bali, M., Naeej, M.R., and Vaseghi Amiri, J., (2013). Prediction of Lateral
Confinement Coefficient in Reinforced Concrete Columns using M5 Machine Learning
Method, KSCE J. Civ. Eng., 17(7), 1714-1719.
23. Park, C.B., Miller, R.D. and Xia, J., (1999). Multichannel analysis of surface waves,
Geophysics. 64, 800808.
24. Quinlan, J.R., (1992). Learning with Continuous Classes, In: Proceedings AI'92 (ed
Adams & Sterling), pp. 343-348. World Scientific, Singapore.
25. Samadianfard, S., Nazemi, A.H., & Sadraddini, A.A. (2014). M5 Model Tree and Gene
Expression Programming Based Modeling of Sandy Soil Water Movement under Surface
Drip Irrigation, Agriculture Science Developments, 3, 178-190.
26. Seed, H.B., Idriss, I.M. and Arango, I. (1983). Evaluation of liquefaction potential using
field performance data, Journal of Geotechnical Engineering. 109, 458482.

Vol. 20 [2015], Bund. 22

12430

27. Thitimakorn, T. and Channoo, S. (2012). Shear Wave Velocity of Soils and NEHRP Site
Classification Map of Chiang Rai City, Northern Thailand, Electronic Journal of
Geotechnical Engineering. 17, 2891-2904.
28. Wang, Y., and Witten, I.H., (1997). Induction of model trees for predicting continuous
classes, In: Proceedings of Poster Papers of the European Conf. on Machine Learning
(Eds M. van Someren & G. Widmer), pp.128-137. University of Economics Faculty of
Informatics and Statistics, Prague, Czech Republic.
29. Witten, I.H., & Frank, E., (2005). Data Mining: Practical Machine Learning Tools and
Techniques, Morgan Kaufmann, San Francisco.
30. Yokota, K., Imai, T., Konno, M., (1991). Dynamic deformation characteristics of soils
determined by laboratory tests. OYO Tee. Rep. 3, p. 13.

2015 ejge