Professional Documents
Culture Documents
2012 (Proceeding) Ultimate Load Capacity of Axially Loaded Vertical Piles From Full Scale Load Test Results Intrepretations-Applied To 20 Case Histories
2012 (Proceeding) Ultimate Load Capacity of Axially Loaded Vertical Piles From Full Scale Load Test Results Intrepretations-Applied To 20 Case Histories
2012 (Proceeding) Ultimate Load Capacity of Axially Loaded Vertical Piles From Full Scale Load Test Results Intrepretations-Applied To 20 Case Histories
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/215673363
CITATIONS
READS
1,904
5 authors, including:
Ariful Hasnat
Partha Saha
24 PUBLICATIONS 59 CITATIONS
8 PUBLICATIONS 8 CITATIONS
SEE PROFILE
All content following this page was uploaded by Ariful Hasnat on 12 January 2017.
SEE PROFILE
Proceedings of the 1st International Conference on Civil Engineering for Sustainable Development
(ICCESD-2012), 23~24 March 2012, KUET, Khulna, Bangladesh (ISBN: 978-984-33-4247-8)
Research Fellow, Department of Soil Mechanics and Foundation Engineering, Housing and Building
Research Institute, Dhaka, Bangladesh, e-mail: hasnat@uap-bd.edu
2
Senior Research Engineer, Department of Soil Mechanics and Foundation Engineering, Housing and Building
Research Institute, Dhaka, Bangladesh
3
Assistant Professor, Department of Civil Engineering, University of Asia Pacific, Dhaka, Bangladesh
4
Research Engineer, Department of Soil Mechanics and Foundation Engineering, Housing and Building
Research Institute, Dhaka, Bangladesh
5
Undergraduate Student, Department of Civil Engineering, University of Asia Pacific, Dhaka, Bangladesh
ABSTRACT
Accurate evaluation of ultimate load capacity of pile is extremely complex and difficult. Capacity of pile is
derived from the resistance developed by the soils surrounding the shaft and underneath the tip of the pile. Soil,
by nature, inherits/exhibits non-homogeneity, anisotropy and in-elasticity. Prediction of accurate engineering
properties of the surrounding particulate three-phase soil matrix becomes even more complex due to
installation of piles. Thus, available semi-empirical/empirical methods for determination of ultimate resistance
of pile pose unreliability and, thereby, may significantly affect safety and economy of a project. Therefore,
although expensive, conduction of limited number of full-scale pile load tests becomes inevitable prior to
commencement of construction. On most occasions, the results of this test do not show a distinct plunging
ultimate load, therefore the results need interpretation to estimate pile capacity or ultimate load. Several
techniques have been developed by different researchers for this purpose. In this paper, twenty pile load tests
performed in different parts of Bangladesh were analyzed with six common interpretation methods and
compared with the actual capacities evaluated from the plunging failure. The comparison uses Log-Normal
distribution, cumulative probability as well as measured versus predicted figures to represent the precision of
the methods. The result shows that the Davissons method (1972) has the highest accuracy and also the lowest
scatter as opposed to the Rebound Elastic (1956) approach.
Keywords: Pile capacity, pile load test, measured capacity, predicted capacity, statistical analysis
1. INTRODUCTION
Rapid urbanization has become an ongoing issue of Bangladesh during the last one-and-half decades.
Tremendous pressure is being experienced by the big cities due to movement of population from rural to urban
areas. Severity of housing problem is increasing everyday. Big cities are growing at the fastest rate ever
happened before. Priority of vertical expansion to horizontal one is strongly realized. Government agencies
strongly realize the need of tall-storied buildings and consequently new 'Building Acts' are approved. To a great
extent, pile foundation has become an indispensable choice for the foundation system of tall-storied buildings.
Piles are relatively long and generally slender structural foundation members that transmit superstructure loads
to deep soil layers. In geotechnical engineering, piles usually serve as foundations when soil conditions are not
suitable for the use of shallow foundations. Therefore, safety and stability of pile supported structures depend on
the behavior of piles. Pile behavior is significantly dependent on the properties of soils surrounding and
underlying the shaft and tip/end of the pile, respectively. Accurate evaluation of the properties of these soils is
quite complex. Generally, soil is non-homogeneous, inelastic and anisotropic. Again, the behavior of these soils
changes significantly during the installation process, in case of driven piles, depending upon the driving method,
energy required, overburden pressure, pile geometry etc. As the dissipation of excess pore water pressure
generated during the driving process is significantly affected by hydro-geologic condition and time, the
properties of these soils become also time and hydro-geology dependent. Consequently, determination of the
exact capacity of pile becomes almost impossible.
Figure 1: (A) Ideal plunging, (B) graph with unclear failure point
2. PILE LOAD TEST DATABASE
A database has been compiled including 20 pile case histories where 17 were driven piles and rest 3 were bored
piles. These case histories were collected from 10 different sites. For each load test, emphasis was placed on the
completeness of the required information which includes: test pile size (length and x-section/diameter),
complete records of the load-deflection data and availability of the subsurface exploration data for the site. Pile
load tests with missing information were discarded. The length of the driven piles varies from 8.24m to 27.45m
and length of the bored piles varies from 13.7m to 15.25m. All driven piles used in this research were square
with x-section varying from 254mm x 254mm to 355mm x 355mm. On the other hand, all three bored piles
were circular with diameter of 600mm.
All the tests were conducted according to ASTM D 1143 test procedure. The main characteristics of these pile
load tests were reaching to ultimate point of bearing capacity by plunging concept. The results of load tests are
summarized in Table 1. Due to the space limitations the load settlement curves are not included in this paper.
Test Pile
No.
1
2
3
4
5
TP 1
TP 2
TP 3
TP 4
TP 5
6
7
8
9
10
11
TP 6
TP 7
TP 8
TP 9
Pile Dimension
Length
X-Sectional
Dimension
M
mm
18.3
300 x 300
12.2
300 x 300
13.7
300 x 300
18.3
300 x 300
12.2
300 x 300
Total
Settlement
Net
Settlement
mm
29.84
31.43
31.9
31.3
22.37
mm
21.4
15.23
18.3
19.46
15.07
TP 10
TP 11
13.7
9.15
15.25
8.24
14.65
9.15
300 x 300
254 x 254
355 x 355
355 x 355
600*
254 x 254
30.18
55.57
19.99
42.14
129.09
56.84
15.6
48.95
9.95
23.68
128.41
50.36
12
13
14
15
16
17
TP 12
TP 13
TP 14
TP 15
TP 16
TP17
15.25
12.2
9.15
15.25
27.45
13.7
355 x 355
300 x 300
254 x 254
355 x 355
355 x 355
300 x 300
52.49
64.17
56.35
28.91
76.73
47.97
43.03
57.14
51.10
17.47
72.55
40.77
18
19
20
TP 18
TP 19
13.7
13.7
15.25
600*
300 x 300
600*
71.68
53.77
74.78
67.33
48.69
62.6
TP 20
Bored pile, * Circular cross-section
3. INTERPRETATION APPROACHES
Without a proper definition, interpretation of ultimate load from load-test result becomes a meaningless scheme.
To be useful, a definition of failure load must be based on some rules. It is difficult to make a cogent choice of
the best capacity criterion to use, because the preferred criterion depends heavily on ones past experience and
conception which establish the ultimate resistance of a pile. To reach ultimate bearing capacity, large
deformation is needed. Because of this reason some failure criteria is based on a specified settlement. In other
hand some designer preferred that to achieve allowable capacity of piles, use the ultimate bearing capacity and a
safety factor. The interpretation methods for ultimate capacity of piles based on axial static load testing data
stipulated in international codes can be grouped to three types:
1) Based on the load grade generating load-settlement (P-S) curve;
2) Based on the settlement under certain load; and
3) Based on the total settlement of the pile head.
Until now, several interpretation approaches (Terzaghi 1942, Housel 1956, Hansen 1963, De Beer 1968, Fullar
and Hoy 1970, Chin 1970, Davisson 1972, Mazurkiewicz 1972, Butler and Hoy 1977, Chin Fung Kee 1977,
Brierley 1979, Decourt 1999) have been established to estimate the ultimate capacity of piles. Among them six
common interpretation method have been selected for this research. Table 2 summarizes these interpretation
methods. Among them, Davisson Offset Limit is recommended in all codes listed and is the most widely used in
engineering practice. This method is based on the assumption that capacity is reached at a certain small toe
movement by compensating for the stiffness (length and diameter) of the pile. The Hansens 80% criterion is
also widely used. The load that gives four times the movement of the pile head as obtained for 80% of that load.
The Decourt extrapolation load limit is equal to the ratio between the y-intercept and the slope of the line. The
Decourt method has the advantage that a plot prepared while the static loading test is in progress will allow the
user to eyeball the projected capacity.
02
03
Interpretation
Methods
Double
Tangent
Method
on
an
Arithmetic Plot
Maximum Curvature
Method
The Hansen 80%
Criterion (1963)
Formula
Remarks
1
4 Qult t
P = 80% Q
04
Davissons
(1972)
05
Method
Rebound
Elastic
Method (1956)
Decourt
Extrapolation (1999)
06
ult
PL
D
+
+ 3.81
AE 120
ult
Q=
C
C
2
1
C2
4. STATISTICAL ANALYSIS
Graphical and probability approaches are engaged to verify the pile load test interpretation approaches.
Cumulative probability and Log-Normal methods have been considered to compare different methods of pile
capacity determination. According to cumulative probability approach the ratio of predicted value (Q p ) and
measured value (Q m ) has been drawn versus cumulative probability. For a series of numerals, Q p /Q m has been
set ascending and indexed with 1 to n, then for each of the relative amounts, the cumulative probability factor
has been calculated as follows:
P (%) =
i
* 100
n +1
(1)
Where P is the cumulative probability factor, (i) is the index of considered case and n is the number of total
cases. For determining the convergence or deviation tendency of the output of prediction, the following
equations have been referred.
ave
QP
Q
m
(2)
%50
The value of Q p /Q m at the cumulative probability of 50% stands as the average of this ratio, so the applied
approach seems to be more precise if this average reaches to 1.
The slope of passing line from points in each approach denotes the amount of deviation; so the results from
applied approach would have less deviation and more reliability if that line has low gradient. The result of
cumulative probability analysis is shown in Figure 2. The average estimated errors of these 6 methods are
summarized in the Table 3. The results of comparison showed that Davissons method predict closer values for
bearing capacity to actual values of cases among other methods. The error of this method is about 0 while this is
12% for Double Tangent Method on an Arithmetic Plot, 8% for Maximum Curvature Method, 4% for Hansen
80% criterion Method, 27% for Rebound Elastic Method, and 5% for Decourt Extrapolation Method.
Interpretation method
Average Error
(%)
Description
12.1
Underestimate
02
7.79
Underestimate
03
Overestimate
04
Davisson's Method
Ok
05
27
Underestimate
06
Overestimate
The Log-Normal distribution can be employed to evaluate the performance of pile capacity prediction method.
The Log-Normal distribution is acceptable to represent the ratio of Q p /Q m ; however, it is not symmetric around
the mean, which means that the Log-Normal distribution doesnt give an equal weight of under prediction and
over prediction. In order to use Log Normal distribution, the mean ( ) and standard deviation ( ) are
evaluated for natural logarithm of Q p /Q m as follows:
Q
Q
1 n
p
p
= ln
n
Q
Q
i =1
m
m
ln
p
=
ln
Q
m
(3)
QP
1
ln
ln
n 1
(4)
i 1
The ratio Q p /Q m and the natural logarithm of the ratio ln(Q p /Q m ) for each pile were calculated. Then, the mean
( ) and standard deviation ( ), and coefficient of variation (COV) of ln(Q p /Q m ) for each method determined.
The Log-Normal distribution is defined as the distribution with the following density:
f (x ) =
1
2 *
1 ln( x)
Exp
*X
2
ln
ln
ln
2
(5)
is the mean of ln(Q p /Q m ) and is the standard deviation of ln(Q p /Q m ). The Log
Where x= (Q p /Q m ),
Normal distribution was used to evaluate the different methods based on their prediction accuracy and precision.
Figure 3 shows the Log Normal distribution for different methods considered in this paper that confirms the
results of cumulative probability analysis.
Interpretation
Method
Double Tangent
Method on an
Arithmetic Plot
Maximum
Curvature Method
The Hansen 80%
Criterion
Davisson's
Method
Rebound Elastic
Method
Decourt
Extrapolation
Method
Mean
Standard
Deviation
COV
% (MaxMix)
% (MinMax)
Skewness
Kurtosis
0.88
0.09
0.10
31.81
46.7
-0.64
0.15
0.94
0.06
0.06
19.67
24.5
0.52
-0.45
1.05
0.08
0.07
22.43
28.9
0.27
-0.67
1.00
0.04
0.04
14.62
17.1
-1.28
1.96
0.67
0.20
0.29
73.56
278
-0.57
-0.40
1.12
0.23
0.20
49.95
99.8
1.21
0.90
Line of Equity
Figure 4: Scatter plots of measured capacity versus predicted capacity graphs for six common pile load test
interpretation methods