You are on page 1of 11

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

net/publication/308903756

An Evaluation of T-Z Analysis Methods

Conference Paper · October 2015

CITATIONS READS
2 2,881

4 authors, including:

Kevin Stanton Sherif Elfass


Mott MacDonald Group University of Nevada, Reno
23 PUBLICATIONS   73 CITATIONS    47 PUBLICATIONS   123 CITATIONS   

SEE PROFILE SEE PROFILE

Kirk Ellison
Arup
19 PUBLICATIONS   134 CITATIONS   

SEE PROFILE

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

Analysis Procedures for Evaluating Superheavy Load Movement on Flexible Pavements View project

Deep Foundations View project

All content following this page was uploaded by Kevin Stanton on 06 October 2016.

The user has requested enhancement of the downloaded file.


AN EVALUATION OF T-Z ANALYSIS METHODS

Kevin V. Stanton, University of Nevada, Reno, NV, USA, kevinstanton@nevada.unr.edu


Ramin Motamed, University of Nevada, Reno, NV, USA, motamed@unr.edu
Sherif Elfass, University of Nevada, Reno, NV, USA, elfass@unr.edu
Kirk Ellison, Arup, San Francisco, CA, USA, Kirk.Ellison@arup.com

ABSTRACT

The determination of accurate load transfer curves are essential for estimating deep foundation load-
settlement response using t-z style analyses. While recently developed semi-empirical methods from
Ashour et al. (2010) and Ashour & Helal (2012) have shown promise in preliminary studies, they have
not yet been thoroughly benchmarked against a more widely accepted approach. To address this issue,
these recent procedures are implemented in the computer program CGI-DFSAP to predict the results of
four Osterberg-cell (O-Cell) load tests. These predictions are compared to outcomes obtained following
the recommendations from Reese & O’Neill (1988) for t-z analyses of drilled shafts in sand and clay
which are carried out with the program TZ-PILE. Overall, the results suggest that the new approach offers
improved accuracy regarding predictions of drilled shaft response to axial loading.

Keywords: deep foundations, axial load test, t-z analysis

INTRODUCTION

The behavior of axially loaded deep foundations is often analyzed using a technique known as the t-z
method. In this method, deep foundation elements are broken into multiple segments which are supported
by discrete nonlinear springs to capture the contribution of both side-friction (t-z springs) and end-bearing
(Q-z springs). Hence, the t-z and Q-z relationships dictate the predicted load-settlement response. A
number of different approaches have been developed to estimate load-transfer curves with some relying
solely on empirical data from laboratory and field experiments (i.e. Coyle & Reese, 1966; Vijayvergiya,
1977) and others based mainly on analytical frameworks (i.e. Kraft et al., 1981; Randolph, 1994).

Reese & O’Neill (1988) provide recommendations which are widely used for developing t-z and Q-z
curves for analyzing drilled shafts. This method is hereafter referred to as Method 1 (M1). It should be
noted that these empirically derived relationships are only applicable for problems associated with drilled
shafts and are therefore not appropriate for developing load-transfer curves for other types of deep
foundations (i.e. driven piles). This is because of the difference in the frictional resistance which can be
developed between a circular concrete-soil interface as opposed to that of other materials (i.e. steel)
and/or different cross sectional geometries.

The M1 load-transfer curves form the basis for the t-z analysis of drilled shafts implemented in the
computer program TZ-PILE (Ensoft, 2014). In light of this, TZ-PILE is employed herein as a reference to
compare against the capabilities of another method of implementing t-z style analyses. This method is
based on a semi-empirical framework for soil behavior developed by Norris (1986) and is implemented
using the program CGI-DFSAP (Computers & GeoEngineering Inc., 2011). More information concerning
this approach, hereafter referred to as Method 2 (M2), is given in subsequent sections.

ANALYTICAL FRAMEWORKS
TZ-PILE (M1)
Reese & O’Neill (1988) used the results of field load tests on drilled shafts with diameters ranging from 2
to 3 ft to develop relationships between the ratios of unit resistance to ultimate unit resistance and
settlement to shaft diameter which allow for the calculation of t-z and Q-z curves. These are given
separately for side load transfer and end bearing in both cohesive and cohesionless material and are
shown in Figures 1 and 2.

Figure 1. Side load transfer for clay (left) and sand (right). From Reese & O’Neill (1988).

Figure 2. Base load transfer for clay (left) and sand (right). From Reese & O’Neill (1988).

To calculate load-transfer curves using M1, the unit side shear along the length of the shaft as well as the
unit end-bearing must first be determined. In the absence of strain gauge data from an O-Cell test, this can
be approximated using procedures for drilled shaft design such as in Brown et al. (2010). However, errors
in the estimated unit resistances could significantly impact the results obtained from TZ-PILE.

An analysis was conducted in which t-z curves derived using M1 with the Brown et al. (2010) unit
resistances were compared to curves calculated by using the unit resistance values based on strain gauge
data from the O-Cell tests. This revealed a number of mismatches between the two sets of t-z curves and,
for nearly all cases, the Brown et al. (2010) curves underestimate unit load transfer. In light of this, two
analyses were performed in TZ-PILE for each load test: one with unit resistances based on Brown et al.
(2010) and the other based on the measured data.

CGI-DFSAP (M2)
CGI-DFSAP employs the semi-empirical approach, M2, to derive t-z and Q-z curves internally. This is
achieved using procedures described in Ashour & Helal (2012) for cohesionless material and in Ashour et
al. (2010) for cohesive material. In M2, the soil-pile system is analyzed as shown in Figures 3 and 4.
Figure 3. Modeling at the soil-pile interface according to M2. Modified from Ashour et al. (2010).

Figure 4. Strains and mobilized stresses below the pile/shaft tip in M2 (from Elfass, 2001).

A constitutive model to characterize soil behavior developed by Norris (1986) and modified by Ashour et
al. (1998) is employed to develop t-z curves in both cohesionless and cohesive soils. This is shown in
Figure 5. It should be noted that the value of axial strain at 50% of the failure stress, ε50, is the main
required input in this model and can be determined internally by CGI-DFSAP based on empirically
derived relationships from Evans & Duncan (1982) for clay and Norris (1977) for sand. These relate ε50 to
undrained shear strength for cohesive material/rock and void ratio for cohesionless or c- material.

Figure 5. Stress-strain model from Ashour et al. (1998) for the behavior of soil.
The stress-strain relationships shown in Figure 5 are utilized along with an assumed pattern of radial
variation of vertical soil displacement at each segment of the foundation, Δzi, to determine the anticipated
total pile/shaft settlement. Specifically, M2 assumes a nonlinear parabolic degradation pattern of vertical
soil displacement at points leading away from the outer diameter of a shaft/pile. This is based on
experimental data from Robinsky & Morrison (1964) which was analyzed in Seo et al. (2008) and Chow
(2007). Equation 1 describes the assumed pattern in M2.

[1]

where is the radial distance from the pile/shaft and is the settlement. At the soil-pile interface,
and . The shear strain associated with vertical soil displacement can be determined by the ratio of
/ . Thus, if the stress level (SL), the ratio of mobilized shear strength to ultimate shear strength, is
known, the settlement in each segment can be calculated by summing the radially degrading settlements
indicated by Equation 1. This requires converting calculated axial strains to shear strains through the use
of Poisson’s ratio.

The SL in cohesionless soil can be found using Equation 2 from Ashour et al. (1998). For cohesive soils,
however, the SL is determined by using the relationships from Figure 3 to calibrate fitting parameters for
the Ramberg-Osgood model as represented by Richart (1975). This is discussed in greater detail in
Ashour et al. (2010).

[2]

where is the mobilized shear strength, is the confining pressure and is the internal angle of friction.
Ashour et al. (1998) also provides a means for estimating based on axial strain and deviatoric stress.

TEST SHAFTS

A total of 4 O-Cell load tests are analyzed in this study. The individual test shaft properties were taken
from reports by Loadtest Inc. and are summarized in Table 1.

Table 1. Test shaft properties


Shaft Location Base Top Diameter Depth of Buoyant Ave. Unit
I.D. Depth Depth (in) O-Cell Weight of Stiffness
(ft) (ft) (ft) Shaft Above Above O-Cell
O-Cell (kips) (kips)
LT-1 Las Vegas, NV 85.00 45.0 60.0 70.2 76.3 9.12x106
LT-2 Sunset Beach, NC 91.50 0.0 49.8 (48)1 68.7 88.0 8.15x106
LT-3 Los Angeles, CA 86.7 0.8 98.5 67.9 516.0 3.14x107
LT-4 Reno-Sparks, NV 115.9 0.0 120 (84)2 85.0 388.0 1.95x107
1
Depth of property change is 39.8 ft below ground surface
2
Depth of property change is 19.5 ft below ground surface

The assumed soil properties for each site are based on data either appended in the aforementioned
Loadtest Inc. reports or obtained through personal communication with the geotechnical consulting
company. Schematics of the test shafts which also show the assumed material properties are given in
Figure 6.
Figure 6. Shaft schematics and soil properties. Modified from reports provided by Loadtest Inc.

DATA INTERPRETATION

Due to the locations of the O-Cell installation depths, no attempt is made herein to differentiate between
base and side friction resistance. Instead, the measured data is interpreted in terms of the total shaft
resistance and the entire load-settlement response is utilized in the comparison against the predictions.
In order to evaluate the accuracy of the results from the t-z analyses, it is necessary to construct equivalent
top-down load-settlement curves from the raw O-Cell data. This is accomplished using the methods
presented in Kwon et al. (2005) which correct for elastic settlement above the O-Cell by considering the
pattern of developed side shear indicated by strain gauge data collected during the load test. Kwon et al.
(2005) also validated this methodology by analyzing a single case in which a top-down load test and an
O-Cell load test were conducted on identical drilled shafts located in close proximity to each other. Their
findings suggest that the method employed herein for constructing the equivalent top-down curve can
produce results which match that of a true top-down load test with acceptable accuracy. In addition, Kim
& Chung (2012) performed a comparisons for two more traditional style load tests and found similar
results.

To make full use of the measured O-Cell data, it is sometimes necessary to extrapolate the load-
movement curve above and/or below the O-Cell so that the response of the shaft can be evaluated up to
settlements corresponding to the ultimate axial capacity. Thus, the amount of extrapolation required
depends on the method by which the shaft axial capacity is defined. There are a number of ways to do this
but in this paper only two methods are considered: the Davisson offset limit (Davisson, 1972) and O’Neill
& Reese (1999).

The Davison offset limit was originally developed for driven piles and its applicability to drilled shafts
has been called into question by other studies (NeSmith & Siegel, 2009; Kulhawy & Hirany, 2009).
However, Ooi et al. (2004) evaluated this approach as well as five other methods by truncating measured
drilled shaft load test data, applying extrapolation and comparing the results. Overall, the Davisson
approach used with extrapolation methods from Chin (1970) was found to provide the most reasonable
and reliable estimates of axial capacity and overall load-settlement response of the drilled shafts included
in the study.

The definition of failure given in O’Neill & Reese (1999) typically yields less conservative values of
drilled shaft axial capacity than the Davisson offset limit. While both methods first consider plunging
failure (i.e. movement without additional load), if this does not occur before the settlement cutoff equal to
5% of the shaft diameter, O’Neill & Reese (1999) suggest the load at the cutoff settlement be taken as the
ultimate axial capacity. In contrast, the Davisson offset limit defines the cutoff settlement as the sum of
the elastic compression of the pile/shaft, the elastic compression of the soil at the pile/shaft tip and the
limiting plastic compression of the soil at the pile/shaft tip.

Based on the recommendations from Ooi et al. (2004), the Davisson offset limit is determined herein
using Equation 3. The axial capacity is then taken as the load at the point of intersection between
Equation 3 and the measured or extrapolated data.

0.15 [3]

where is the top deflection (in) and , , , and are the cross-sectional area, Young’s modulus, top
axial load, length (in) and diameter (in), respectively. While some of the data in this paper does not
require extrapolation to reach the Davisson offset limit, extrapolation is required in all cases to reach the
5% settlement to shaft diameter ratio specified by O’Neill & Reese (1999). The approach from Chin
(1970) is employed for extrapolation by fitting a hyperbolic curve to the available data to forecast
additional points. This requires calibrating fitting parameters and from Equation 4.

[4]
It should be noted that only the linear portion of the load-movement component requiring extrapolation
(i.e. up or downward O-Cell movement) was used to determine the parameters and . Using the
aforementioned procedures for extrapolation and determination of axial capacity, Table 2 summarizes the
axial capacities, corresponding settlements and extrapolation required to define failure. Also, based on the
data, it is assumed that none of the load tests included in this study experienced plunging before reaching
the settlements required to defined failure from either Davisson (1972) or O’Neill & Reese (1999).

Table 2. Interpreted axial capacity, corresponding settlement and required extrapolation


Shaft Davisson (1972) O’Neill & Reese (1999)
I.D. Axial Capacity Settlement Extrapolation Axial Capacity Settlement Extrapolation
(kip) (in) Required1 (%) (kip) (in) Required1 (%)
LT-1 2640 0.79 0.00 3710 3.00 3.03
LT-2 4260 1.12 0.00 5180 2.40 2.46
LT-3 12600 2.27 0.00 16350 4.93 2.70
LT-4 22840 2.96 2.60 26390 4.20 4.07
1
Extrapolation required is the ratio of extrapolated settlements to diameter needed to reach the given failure
definition

MODELING

Since M1 and M2 require different information, there were some inconsistencies amongst the final input
parameters in CGI-DFSAP and TZ-PILE. This mainly concerns the modeling of materials which have
both cohesion and an internal angle of friction greater than zero. While CGI-DFSAP allows for both
friction angle and cohesion to be considered for a given layer, TZ-PILE requires that each material be
assumed to behave as either entirely cohesive or cohesionless. For this study, the factor (cohesion or
friction angle) which theoretically contributed more to the total shear strength of a layer ultimately
determined how the layer was modeled in TZ-PILE.

There is no option in TZ-PILE to model a shaft top above or below the ground surface. Thus, for LT-1,
the additional overburden had to be represented with a thin layer at the surface of the soil profile which
was assigned the appropriate unit weight to mimic the overburden at the top of the shafts. This approach
was validated by entering user-defined t-z and Q-z curves derived in CGI-DFSAP and ensuring that the
same results were obtained. It should also be acknowledged that since M1 is not applicable to rock, rock
layers were modeled in TZ-PILE by assuming the behavior of a very hard clay. For these cases, the
required input for cohesion was estimated as half of the unconfined compressive strength of the rock
mass.

The internally calculated values of ε50 in CGI-DFSAP were accepted for nearly all layers due to a lack of
information in the geotechnical reports. For rock and end-bearing layers, however, ε50 was estimated
manually based on the upper bound from relationships given in Evans & Duncan (1982) and Norris
(1977). This was intended to avoid unnecessary errors because CGI-DFSAP seems to produce
unrealistically soft responses when the internally calculated values are used in these cases. Also, to
maintain comparability, the same values of ε50 were used in both CGI-DFSAP and TZ-PILE.

RESULTS

The predicted load-settlement responses are presented in Figure 7 along with the measured/extrapolated
data, the Davisson (1972) offset limits and the O’Neill & Reese (1999) service limits discussed in the
previous section. In general, the t-z analyses conducted using CGI-DFSAP appear to give the most
accurate predictions. This observation is quantified in Table 3 which shows the predicted axial capacities
normalized by the measured axial capacities. The predicted axial capacities are determined by first
considering plunging failure and only using the criteria from Davisson (1972) and/or O’Neill & Reese
(1999) if there is no evidence of plunging prior to reaching the respective settlement limits.

Figure 7. Predicted and measured load-settlement responses.

Table 3. Comparison of measured and predicted axial capacities


Shaft I.D. CGI-DFSAP (M2) TZ-PILE (M1)1
Rp/Rm2 Rp/Rm3 Rp/Rm2 Rp/Rm3
LT-1 0.93 0.98 0.784 0.564
LT-2 0.93 0.98 0.414 0.344
LT-3 0.75 1.07 0.51 0.454
LT-4 1.00 1.04 0.17 0.24
1
Maximum unit resistances estimated using Brown et al. (2010)
2
Predicted divided by measured capacity both determined using the Davisson (1972) offset limit
3
Predicted divided by measured capacity both determined using the O’Neill & Reese (1999) settlement limit
4
Predicted capacity based on plunging failure mode
SUMMARY AND CONCLUSIONS

Regardless of the definition used for the ultimate shaft capacity, the predicted shaft capacities from M2
match the measured values more closely than that from M1. Also, aside from the methods for deriving t-z
and Q-z curves, it is likely that the differences between the software types used for implementation does
not significantly impact this observation. This is evident based on the aforementioned exercise in which
TZ-PILE produced the nearly the exact same results as CGI-DFSAP when the t-z and Q-z curves from
CGI-DFSAP were entered into TZ-PILE.
The definition of ultimate shaft capacity is lower for all measured and predicted results herein when using
the Davisson offset limit as opposed to the definition given by O’Neill & Reese (1999). This confirms the
findings of past literature (Ooi et al., 2004; Paikowsky & Tolosko, 1999). While this observation may not
always hold true, this research adds to a growing body of evidence that suggests if the predicted axial
capacity from Davisson (1972) is greater than that from O’Neill & Reese (1999), the model is likely to be
giving non-conservative estimates.
Based on the results obtained in this paper, the following conclusions can be made:
1) In the absence of site specific t-z and Q-z curves, for problems involving drilled shafts, M2 has
the ability to provide a reasonable alternative and outperforms M1.
2) While it is best to enter values for ε50 based on site specific test data, acceptable predictions can
be obtained using the internally calculated values in CGI-DFSAP for non-end-bearing and non-
rock layers. It is recommended to use user-defined values for end-bearing and very stiff (i.e. rock)
layers because model results are highly sensitive in these cases and tend to underestimate load-
settlement behavior when the internally calculated values are used.
3) M1 tends to indicate plunging failure prematurely. Additionally, M1 is highly sensitive to
inaccurate estimates of maximum unit resistance. For example, in LT-3 and LT-4, the Brown et
al. (2010) estimated unit resistances were very similar to those measured in the O-Cell tests,
leading to similar predicted load-settlements using both inputs. Conversely, in LT-1 and LT-2, the
Brown et al. (2010) estimates were lower which produced overly conservative load-settlement
estimates.
It is important to note the findings from this study are only applicable to drilled shafts. While the methods
from Ashour et al. (2010) and Ashour & Helal (2012) can be easily adapted for driven piles, additional
research is required to benchmark the procedure for anything other than circular reinforced concrete
drilled shafts. That being said, overall, the results presented in the previous section suggest that M2 is
capable of providing reasonably accurate load-settlement response predictions for drilled shafts.
Furthermore, the t-z curves derived according to Ashour & Helal (2012) and Ashour et al. (2010) likely
represent an improvement over the curves suggested by O’Neill & Reese (1988) for drilled shafts.

ACKNOWLEDGEMENTS

The authors would like to thank Loadtest Inc. and Jon Sinnreich for sharing the data needed to complete
this study. In addition, the support of NDOT for funding this study is acknowledged.

REFERENCES

Ashour, M., & Helal, A. (2012). Response of Axially Loaded Piles in Sands with and without Seismically
Induced Porewater Pressures. Int. J. of Geomechanics, doi:10.1061/GM.1943-5622.0000273.
Ashour, M., Norris, G. M., Elfass, S., & Al-Hamdan, A. Z. (2010). Mobilized side and tip resistances of
piles in clay. Computers and Geotechnics, 37(7), 858-866.
Ashour, M., Norris, G., & Pilling, P. (1998). “Lateral Loading of a Pile in Layered Soil Using the Strain
Wedge Model.” J. of Geotechnical and Geoenvironmental Eng., ASCE, 124(4), pp. 303-315.
Brown, D. A., Turner, J. P., & Castelli, R. J. (2010). Drilled shafts: Construction procedures and LRFD
design methods. US Department of Transportation, Federal Highway Administration.
Chin, F. K. Estimation of the Ultimate Load of Piles Not Carried to Failure. Proc., Second Southeast
Asian Conference on Soil Engineering, 1970, pp. 81–90.
Chow, H.S.W. (2007). Analysis of piled-raft foundations with piles of different lengths and diameters.
PhD thesis, University of Sydney.
Computers & GeoEngineering, Inc. (2011). CGI-DFSAP. Ret. April 7, 2015, from http://cgi-dfsap.com/.
Coyle, H. M., & Reese, L. C. (1966). Load transfer for axially loaded piles in clay. Proc., ASCE,
92(SM2)
Davisson, M.T. (1972). High Capacity Piles. Proc., Soil Mechanics Lecture Series in Innovations in
Foundation Construction, ASCE, Illinois Section, Chicago, March 22, pp. 387-411.
Elfass, S. (2001). “A New Approach for Estimating the Axial Capacity of Driven Piles in Sand up to True
Soil Failure.” Ph.D. Thesis, University of Nevada, Reno.
Ensoft (2014). Analysis of Load Versus Settlement for An Axially-Loaded Deep Foundation. Austin, TX:
Ensoft, Inc.
Evans, L.T., & Duncan, G.M. (1982). Simplified analysis of laterally loaded piles. University of
California Berkeley, Rept. No. UCB/GT/82-04.
Kim, S. R., & Chung, S. G. (2012). Equivalent head-down load vs. Movement relationships evaluated
from bi-directional pile load tests. KSCE J. of Civil Engineering, 16(7), 1170-1177.
Kraft, L. M., Kagawa, T., & Ray, R. P. (1981). Theoretical t-z curves. J. of the Geotech. Eng.
Div., 107(11), 1543-1561.
Kulhawy, F. H., & Hirany, A. (2009). Interpreted Failure Load for Drilled Shafts via Davisson and L1 -
L2. In Contemporary Topics in Deep Foundations (pp. 127-134). ASCE.
Kwon, O. S., Choi, Y., Kwon, O., & Kim, M. M. (2005). Comparison of the bidirectional load test with
the top-down load test. Transportation Research Record: J. of the Transportation Research
Board, 1936(1), 108-116.
NeSmith, W. M., & Siegel, T. C. (2009). Shortcomings of the Davisson Offset Limit Applied to Axial
Compressive Load Tests on Cast-in-Place Piles. In Contemporary topics in deep foundations (pp.
568-574). ASCE.
Norris, G.M. (1977). The Drained Shear Strength of Uniform Quartz Sand as Related to Particle Size and
Natural Variation in Shape and surface Roughness. PhD thesis, University of California,
Berkeley.
Norris, G. M. (1986). Theoretically Based BEF Laterally Loaded Pile Analysis. Third Int. Conf. on
Numerical Methods in Offshore Piling, Nantes, France, pp. 361-386.
O'Neill, M.W., & Reese, L.C. (1999). "Drilled Shafts: Construction Procedures and Design Methods,"
Publication No. FHWA-IF-99-025, Federal Highway Administration, Washington, D.C., 758 p.
Ooi, P.S.K, Chang, B.K.F., & Seki, G.Y. (2004). Examination of Proof Test Extrapolation for Drilled
Shafts. Geotechnical Testing J., 27(2), 11 p.
Paikowsky, S. G., & Tolosko, T. A. (1999). Extrapolation of pile capacity from non-failed load tests.
Federal Highway Administration Report No. FHWA-RD-99-170, Washington, D.C.
Randolph, M.F. (1994). Design methods for pile groups and piled rafts. Proc. of the 13th Int. Conf. on
Soil Mechanics and Foundation Eng., Vol. 5, New Delhi, India, p. 61–82.
Reese, L. C., & O'Neill, M. W. (1988). Drilled shafts: Construction procedures and design methods.
Prepared for US Department of Transportation, Fed. Highway Admin., Office of Implementation.
Richart E.F. (1975). Some effects of dynamic soil properties on soil-structure interaction. J. Geotech.
Eng. Div (ASCE); 101(GT12):1197–240.
Robinsky, E.I., & Morrison, C.E. (1964). Sand displacement and compaction around model friction piles.
Can. Geotech. J.; 1(2):81–93.
Seo, H., Prezzi, M., & Salgado, R. (2008). Settlement analysis of axially loaded piles. Proc. of the 6th Int.
Conf. on Case Histories in Geotech. Eng.. and Symposium, Arlington, VA.
Vijayvergiya, V. N. (1977). Load-movement characteristics of piles. In Ports'77.4, The annual
symposium of the ASCE, Waterway, Port, Coastal and Ocean Div., Long Beach, California,
v.2 (pp. 269-284).

View publication stats

You might also like