Professional Documents
Culture Documents
Focht 1994
Focht 1994
ABSTRACT: This paper focuses on the broad factors critical to the prediction
process that govern the quality of predictions of the geotechnical performance of
Downloaded from ascelibrary.org by Florida Atlantic University on 01/22/15. Copyright ASCE. For personal use only; all rights reserved.
INTRODUCTION
Lambe's Rankine lecture was prepared from the standpoint of the pro-
fessor-researcher-consultant using as examples major projects with extensive
investigations and large budgets rather than from the perspective of the
practicing geotechnical engineer, who usually must cope with abbreviated
Downloaded from ascelibrary.org by Florida Atlantic University on 01/22/15. Copyright ASCE. For personal use only; all rights reserved.
Other critical aspects may also be important for some particular problems,
Downloaded from ascelibrary.org by Florida Atlantic University on 01/22/15. Copyright ASCE. For personal use only; all rights reserved.
QUALITY OF PREDICTIONS
I have adopted for the sake of presentation the concept, as shown in Fig.
1, that a circle can represent the range of presumed knowledge or under-
standing inherent in each of the six critical elements and that the circle
represents the limits of accurate prediction. This graphical presentation is
intended to be thought-provoking and intuitive rather than fully appropriate
or correct from a mathematical perspective. We will assume, then, that
when all of the element circles overlap in one area, the actual performance
will be located in that overlying area. Unfortunately, we won't know until
1655
1656
O
I.-
uJ
Downloaded from ascelibrary.org by Florida Atlantic University on 01/22/15. Copyright ASCE. For personal use only; all rights reserved.
,.I
O
0 mbe, 1
100
QUALITY OF DATA
FIG. 3. ConventionalConcept of Accuracyof Predictions
lOO %
0
~ L a m b e , 1!73)
0 100
QUALITY OF DATA
FIG. 4. Lambe Concept of Accuracyof Predictions
curacy.
(b)
FIG. 5. San Jacinto Monument: (a) Section through Shaft; (b) Foundation Plan
1658
shown in Fig. 6, is mostly stiff to hard clay with layers of dense sand. These
soils are typical of the Beaumont clay formation, a deltaic deposit of Pleis-
tocene age. Based on tests on the original 1936 samples and stratigraphy
from a nearby water well, Professor Raymond F. Dawson at the University
of Texas estimated in 1938 an ultimate settlement of 180 mm (7.35 in.) to
occur in 800 years (Dawson 1938). His predicted time-settlement curve is
shown in Fig. 7, with the typical parabolic shape representing classical pri-
mary consolidation of clay. This prediction was made just after the First
International Conference on Soil Mechanics and Foundation Engineering,
held in Cambridge, Mass., in 1936.
DESCRIPTION
25
SUffclayandsand - 10
layers t-
50 Dense clayey sand
Hard clay
20
n~
I-.-
UJ
75 W
UJ Dense silt
LL UJ
Silt, sand, and clay
100 ayers / -- 30
13..
ara clay :i:"
I--
LU Silt, sand, and clay
o 125 layers LU
40 121
Dens~ clayey sand
150 Stiff clay
Dense sand
,c 50
L~Stiffclay
sand
1 7 5 I Stiff t~ h - ~ clay
f
/ 60
200
(AfterGreer& McClelland,1953)
SOIL PROFILE
FIG. 6. 1953 Boring Log, San Jacinto Monument
1659
~: 6 In 800 yr 1so .~
Z
ILl
:S 8
LU
..J
_ Notes:
1. Start date for observations
Is November 10,1936. ~ ~ - ~OBSERVED
,00
10 2. Predicted cunte starts 0.5 yr - - 250 ,,,,
ILl ~ SI:FrLEMENT -
co later at midpoint of
12 constructlon. 300 CO
(aRer Fenske and Dawaon, 1984)
14 I I I I 35O
0 5 10 15 20 25 30 35 40 45 50
TIME, YEARS
1660
I
Downloaded from ascelibrary.org by Florida Atlantic University on 01/22/15. Copyright ASCE. For personal use only; all rights reserved.
I Edge of Mat /4
ONE
SHELL
PLAZA
E
Mc~nney "3
~o
o,
Edge of Mat (lSm)---.%
J-1
TENNECO
[] []
BLDG.
[] ~p J-18 []
Lamar
0 20 40 60m
I II I ' ' I
across the street from the O n e Shell Plaza site, as shown in Fig. 8. A senior
vice president of Tenneco was concerned about the potential effects the
construction of One Shell Plaza might have on his building. A s the geo-
technical consultant for Hines, I a t t e n d e d a high-level conference that in-
cluded senior officials from the Tenneco, Hines, and S O M organizations
and representatives of the contractor. In that meeting, the Tenneco senior
vice president had a n u m b e r of p o i n t e d questions. O n e of those questions
was " W h a t will construction of O n e Shell Plaza do to the foundation per-
formance of our structure?" I had been a special consultant to the architect
1661
ft) excavation for the 40 by 59 m (132 by 192 ft) mat, with a 15.2 m (50 ft)
excavation around the perimeter to occupy the full 83 m (272 ft) square
block (Focht et al. 1978). As shown by the soil profile beneath the Shell
building as presented in Fig. 9, the water table before construction was
about 10 m (35 ft) below grade, thus requiring construction dewatering to
lower the water table at least 7.6 m (25 ft) around the deeper portion of
the excavation. Actually, as much as 9-10.5 m (30-35 ft) of water level
lowering at the limits of the block was anticipated. A permanent ground-
water control system had been planned to maintain the water level at the
building perimeter after construction at a position at least 15 m (50 ft) below
grade, a permanent lowering of at least 4.5 m (15 ft). The building con-
struction would result in a net positive foundation load of 35.9 MPa (750
psf) on the mat and a net decrease of 263 MPa (5,500 psf) in the annular
area around the mat. A prediction of the effect of One Shell on Tenneco
was needed on the spot, with no opportunity for analysis.
WATER
DESCRIPTION CONTENT, %
0 20 40 60
0
25
10
50
20
75
IJJ
uJ l-
Lt.
3~ 100 30
a~
a. i--
LIJ n
C~ 125 w
40
150
50
175
200 60
(after Focht et aL 1978)
FIG. 9. Soil Profile, One Shell Plaza
1662
i0.4
IJ.I .J
(/)
._1 One Shell
<r Excavation Began
1-
W-0.6 15 z
}
I
W
Z ~ One Shell
UJ Mat Poured
n- u.
I
"'
IJ. 0.8
(After I~,Clelland Englnce~, Inc., 1972)
2O ---
a
IJ.
a
FIG. 10. McClelland Differential Settlement Data for J-18 Relative to J-l, Tenneco
Building
1663
\
Downloaded from ascelibrary.org by Florida Atlantic University on 01/22/15. Copyright ASCE. For personal use only; all rights reserved.
ul
One Shell
uJ Structural Frame
,.J
Completed = 10 uJ
IE 0.4 U)
141 One Shell __ ,.I
Excavation Began
.J
~ 0.6 I ~ ~ - 15 I--
Z
I Ul
ul One Shell
w Mat Poured W
(AfterMicC]ellandiEngin..... line.,1972) 2O m
u.
IL 0.8 a
FIG. 11. ETCO and McClelland Differential Settlement Data for J-18 Relative to
J-l, Tenneco Building
BASS STRAIT
TASMANIA
SITE LOCATIONS
~
FIG. 12, Bass Strait Location
1665
"!"
(~ 80 72' 20
"' 70 ~2'
I I ",\ lIllll}l
-~
', t32, 30
>-
>. m
IZI 60 40 r
n- uJ
U,I 50 50 r
Z n"
E 40 60
I- ;\\ \\
Z 30 ~,~
w \. 70 pz
o b".~". %..\ /'" ,,,
- 90 I,LI
1o II k~ =-
0 I Ill [i I 1 1 1 = I i [ r I I]l,l Ii , , i r r I r , 100
10 5 1 0.5 0.1 0.05
G R A I N SIZE IN M I L L I M E T E R S
FIG. 14. Grain-Size Curve, Kingfish A Platform, Bass Strait
of water and, as anticipated based on high silt content, also had relatively
low driving resistance.
A blowout occurred at the Marlin platform in December 1968 creating a
crater as much as 12 m (40 ft) deep, encompassing the conductors and some
of the platform piles. Concern about reduced capacity of these piles due to
the crater led to additional borings and pullout tests on conductors at several
platforms and a comprehensive program at Kingfish A. Six pullout tests at
Kingfish A were performed on conductors 0.5 m (20 in.) in diameter and
driven to penetrations of 45-83 m (148-271 ft). These tests revealed that
the average mobilized frictional resistance was on the order of only 9.6-
15.6 MPa (200-330 psf). Pullout curves as reported by Angemeer et al.
(1973) are presented in Fig. 15. These frictional values are much lower than
the average tensile skin-friction value of 63.2 MPa (1,320 psf) we computed
in our analyses of piles driven to 61 m (200 ft) of penetration. We concluded
that the frictional capacity of the piles in place was only one-quarter of what
we originally predicted. Why?
Closer examination under a microscope of washed sand samples from
new borings showed them to be calcium carbonate, not the silica sands that
we were familiar with. The grains of calcium carbonate are mostly skeletal
remains of marine organisms, which are relatively soft and crushable. Other
particles are fecal pellets, coated grains (ooliths), shell fragments, and de-
1667
60
w
3:2.0 5O
z_ 3 (sS.2m) p:
Z
40 Ir
Z 1.5
lU I.iJ
tU
30 >
O
:> 1.0
O
20
0.5
10
(After Angemeer et al. 1973)
0
50 100 150 200 250 300 350 400
ESTIMATED AVERAGE SKIN FRICTION, PSF
FIG. 15. Results of Conductor Pullout Tests, Kingfish A Platform, Bass Strait
trital, all within a matrix of line mud, or micrite. Individual grains are bonded
together leaving considerable voids and creating a very delicate fabric, even
though the undisturbed material may be hard and rock-like. It became
apparent that when an open-end pipe pile is driven into these materials,
the grains crush, the bonds break, and there is no buildup of lateral soil
pressure and frictional resistance on the pile wall. Consultations with ge-
ologists and carbonate sedimentologists confirmed that the marine origin of
these deposits resulted in a structure markedly different from that of the
silica sands in the Gulf of Mexico and the North Sea, with which we were
familiar.
Immediately, emergency investigations were performed to evaluate this
severe new problem, and very comprehensive studies were initiated on the
capacity of piles in calcareous deposits. The first results were published in
1973 by Angemeer et al., followed by reference to the problem by Mc-
Clelland (1974) in his Terzaghi Lecture. An excellent review of pile capacity
in calcereous sands was presented by Murff in 1987. The results of studies
undertaken worldwide to better understand marine carbonate deposits were
presented in a specialty conference held in Perth, Australia, in 1988 (Jewell
and Khorshid 1988).
The lesson to be learned from Bass Strait in the context of this lecture
is that simple extrapolation of a predictive procedure to a different geologic
setting is fraught with hazard, particularly without having closely examined
the samples. Yet this is a risk that many of us run at various times in our
careers. Because of the empiricism of many of our predictive techniques,
we must try every time to be sure that each aspect of our problem is com-
patible with those on which the correlation was developed.
1668
(Finno 1989). The latter two were a nominal 0.46 m (18 in.) in diameter.
All four foundation elements had 15.4 m (50 ft) of penetration through 7
m (23 ft) of sand into an underlying soft to medium stiff clay. All of the
participants were furnished geotechnical data, as shown in Fig. 16, from the
test site and were asked to predict a number of aspects of performance
including capacity at three different times after installation, load-deflection
curves, pore pressures adjacent to the piles during driving, residual loads
after driving, and capacity contributions from the sand and the clay. On
capacity, participants were asked first for a best estimate of the ultimate
capacity and then a "lower bound estimate of capacity, with 90 percent
confidence that the actual value would be higher than this lower bound."
Predictions were received from 24 participating teams, although not all
participants predicted everything.
For this paper, my review and comments are limited to the two driven
piles and their ultimate capacity nominally one-year after driving (43 weeks).
In the summary of the program, Finno et al. (1989b)chose to use the
maximum applied load as the ultimate capacity, and I will stick with that
convention. The measured capacities of the two driven piles were similar,
1.02 kN (230 kips) for the pipe pile and 0.98 kN (220 kips) for the H-pile.
The predictors used a wide range of predictive procedures as documented
in ASCE's Special Geotechnical Publication No. 23 (Predicted 1989). The
mean best-estimate prediction of ultimate capacity for each of the test piles
was reasonably close to the measured capacity--0.96 kN (215 kips) for the
pipe pile and 1.0 kN (224 kips) for the H-pile (see Figs. 17 and 18). Maybe
WATER UNDRAINED
DESCRIPTION CONTENT, % SHEAR STRENGTH, KSF
0 20 40 0 0.5 1.0 1.5
0
(a) ,. , f i(b j (c)
~,/w% (1999)
10 - Sand -
Fill 5
20 - Ir
PL LL FIELD VANE (1 188) UJ
MJ
M. ~-. + . . . + J
9I- 3 0 - + - 9 . . . . . . _1.+ --
I'-
I1.
Iii
Soft to
+.-~ . . . . . +
.~ ....... +
+~.---+
,;
.g-
/o I,om
W
40 - Medium a
Clay .99
- .+
20
50 -
9 F ul (1960)
-'1---9 . . . . . . '+
i 9 I I I
6O
(After Flnno, 1 9 8 9 0 20 40 60
UNDRAINED
SHEAR STRENGTH, MPa
1669
kips), 173% of the measured value. At the other extreme, the smallest
capacities predicted were 0.56 and 0.62 kN (126 and 139 kips), respectively,
for the pipe and H-piles, or 55% and 63% of the measured values. Eleven
of the 24 predictions for the pipe pile were within _ 20% of its measured
capacity. For the H-pile, 14 of the 22 predictions were in the same range.
It is particularly pertinent to note that for each of the two piles, 10 of the
"on target" predictions were in the 80-100% range, right where most con-
sultants would like to be, close to but just on the conservative side of the
true value.
Combining the results from both piles into a single histogram, Fig. 19
shows that there were 25 "satisfactory" predictions, i.e. within _+20%.
These were made by 16 different predictors using nearly that many different
procedures. The conclusions that can be drawn are: (1) On average, the
predictions were quite close; and (2) individually, the deviation from mea-
sured was fairly large--from an underprediction of 45% to an overpre-
diction of 73%. The lower-bound 90% confidence capacity values are con-
sidered further in a discussion on factors of safety later in this paper.
I was not surprised that different techniques produced satisfactory an-
swers. Also, I was not surprised that similar, if not identical, techniques
produced both satisfactory and unsatisfactory predictions. Presumably, the
predictions falling outside the acceptable range did not have proper judg-
10 45
22
Values
4O o~
r
Underpredlctlons-9~- I - ~ Overpredlctlons Z
Z 8
O 35 O
}-
I-
O Median
30
IJJ
"'
ri- 6 I n-
D. Mean 25 0-
Li.
O 20 O
r 4 I--
IJJ H-PILE Z
15 uJ
m
10 uJ
Z 2 a.
5
i i I i , ~ , 0
00 0.5 1.0 1.5 2.0
RATIO O F
PREDICTED TO O B S E R V E D
CAPACITY
1670
LU 6 ~ PIPE PILE 25 w
a,. n
[L
20
0 0
n- 4 p-
UJ 15 Z
111 W
0
10
Z 2 W
a.
5
0 ! I 0
0 0.5 1.0 1.5 2.0
R A T I O OF
P R E D I C T E D TO O B S E R V E D
CAPACITY
FIG. 18. Histogram o f R a t i o o f P r e d i ~ e d t o O b s e r v e d U I t i m a t e C a p a c i t y , PipePile
20
: 46 40
, Values
18
-Underpredlctlons ---5-1- ~ - Overpredlctlons
35 u)
Z
Z 16 9i -Median
O O
m
I- 30 ~"
O
r 14
~ - Mean
UJ 12 Z 25 uJ
h-
rr
0. Z a.
LI. 10 20 "
O iBOTH PILES O
n- 8 I.-
UJ 15 z
m LU
6 (J
10 rr
UJ
Z 4
5
2
I I I I
0
00 0.5 1.0 1.5 2.0
RATIO OF
PREDICTED TO OBSERVED
CAPACITY
1671
with the tangent of the friction angle varying from 0.53 to 1.07. Perhaps
each predictor's local experience played a part in producing these wide
ranges. I wonder how much intuition was applied to the highest and lowest
predictions based on the predictor's experience.
It is also important to emphasize that this situation is one in which a
correct capacity prediction could occur due to two compensating errors
because there were almost equal penetrations in clay and in sand. Nineteen
of the predictors provided their estimate of capacity contributions from the
sand and the clay, a total of 38 predictions for the two piles (Finno et al.
1989a). Of these 38 combinations of predictions, there were 21 total capacity
predictions within _+20% of the measured total capacity. But of these 21
predictions, there were only six companion predictions of the relative con-
tributions from the sand and the clay within _+20% of the actual relative
contribution of sand and clay. That means that more than two-thirds of the
21 "correct" total predictions were correct because of compensating errors,
some overestimating the contributions from the sand and others from the
clay, divided about half and half. These results strongly suggest that for
mixed-soil profiles, right predictions may be produced by compensating
errors in the contributions from the major soil components. Interestingly,
there were two teams that were within _+20% on total capacity and on the
distribution between clay and sand for both piles.
The other predictions (time rate of capacity gain, pore pressures, residual
loads after driving, load deflection performance) were generally quite scat-
tered, indicating very low reliability. We cannot claim a very high degree
of individual reliability for capacity predictions for piles in mixed soil profiles
even though the bulk of the individual predictions (30 out of the 46 total)
were within - 3 0 % to + 20% of the measured capacity, a very desirable
range.
Perhaps we could interpret the distribution of predictions on a percentage
basis from this exercise as representing what a single predictor might expect
for his predictions for a variety of piles in varying soil profiles. Hopefully,
application of the aforementioned critical elements 5 and 6 would tend to
eliminate the extremes at both the upper and lower ends of the diagram in
Fig. 19, but particularly at the upper end.
The major lesson learned from this example is that valid predictions may
be derived due to compensating errors. We can also infer that extrapolation
of local experience to a distant site may introduce deviations between pre-
diction and performance.
14.
9 10
e6
9
o 50 Clay 9 4 Z
O
.J
jJ -2 2O W-
><
uJ 100 30
TIll SI SLOPEINDICATOR _ UJ
9 SETn.EMENT PLA'I~
9 PIEZOMETER 40
V HEAVEMARKER --
150 -- - ? . . . . . . . . . ? . . . . . . . . ?. . . . . . . . ?- - -
Rock (After MIT, 1974) - 50
WATER
DESCRIPTION CONTENT, % UNDRAINED SHEAR STRENGTH, KSF
0 20 40 60 0.5 1.0 1.5 2.0 2.5 3,0
O --Peat v-- ,i,I t,t t 0
Fine : I I
\ Sand P 9 9 U TEST
20 -~ S t i f f B l u e / - == "A / 9 UU TEST
.\ clay /~ 9 :~ 9 FIELDVANE _ 10
40 -
60
.
Medium
. Blue
- f u . Clay
-
_
. :,. )
9 t 9 ~fromCKoU -- 29
z Z
O
80 ~-,u
1" ,l~ r~. C ~ reselon Tests _o
:'
..J 100 -
u4 e "KoU IX - 30
}- Soft Blue 9
120 " Ulav -
99 t 2%- T e s t s , N --
F q I "~ P,J II , \ , 40
high, with a lv on 2h slope on the east and a lv on 1.9h side slope 7.6 m
(25 ft) high above a berm on the west. The west s!ope was to be extended
upward with slight steepening. The east slope would be steepened to the
angle of repose of the sand fill, or about lv on lh, as the e m b a n k m e n t was
raised over a 91 m (300 ft) long section centered on Sta. 263. The intent
was for the east slope to fail but not the west. The original foundation
stratigraphy is typical of the Boston a r e a - - a few feet of peat underlain by
a thin layer of well-graded silty sand. The foundation stratum of interest is
Boston blue clay, stiff and overconsolidated at the top, becoming soft and
normally consolidated with depth. The upper surface of the basal glacial till
slopes down from east to west across the site, as shown in Fig. 20. Note the
locations of i n s t r u m e n t s - - s l o p e indicators, piezometers, settlement plates,
and heave markers.
The participants were furnished detailed geotechnical data, including pi-
1673
and H-2) and inclinometers (SI-1, 2, 3, 4), and pore pressures recorded in
piezometers (P-3, 4, 6), all shown in Fig. 20, at various stages during fill
placement.
The embankment failed at a total height of 16 m (52.5 ft), with both
slopes failing early one morning with no one present. The failure was of
"extraordinary proportions." The crest dropped about 9 m (30 ft) and the
sides heaved as much as 3 m (10 ft). The length of the failure was 313 m
(1,030 ft), extending 137 m (450 ft) south of the heightened test section and
85 m (280 ft) to the north. There was no warning from the slope indicators
or piezometers of impending failure.
The predictions of additional fill height to cause failure ranged from 2.4
to 5.2 m (8-17 ft); a histogram of the ratio of predicted to observed ad-
ditional height to failure is shown in Fig. 22. The ratio ranges from 44% to
144%. Six of the 10 predictors underestimated the incremental failure height
and four overpredicted. According to Dr. Francisco Silva-Tulla, who was
closely associated with the exercise, the concensus of the participants and
the audience at the exercise symposium was that the predictions varied
widely (Silva-Tulla, personal communication, 1993). In another light, total
5 50
lO
Values
Median - ~
40
(n U)
Z Mean - - ~ Z
O
I-- I,--
O O
30
UJ LU
n- n-
n. O..
IJ. U.
O O
rr p-
20 z
UJ uJ
m Z Q
:S n-
//// uJ
Z D.
10
0 I I I ....~ I I I I 0
0 0.5 1.0 1.5 2,0
R A T I O OF
PREDICTED TO OBSERVED
ADDITIONAL HEIGHT
1674
k-
0 35 rc~
a LLI
m
re 3 30 n-
13.
O. //
LL
//
25 O
0 //
O: 2 //
20 z
W //
w
Ol
15 cc
// uJ
Z // o.
10
definition given in Fig. 20 gave the expected density of the new fill as 115
lb/cu ft (pcf). The actual measured in-place density of the new fill was 122
pcf, an increase of 6%. The potential effect on the prediction histogram for
incremental height in Fig. 23 to-correct for this deviation would be to shift
the ratio of predicted to observed by 6%, not much in this case, but illus-
trative of the importance of accurately predicting all properties, not just
strength.
The lessons from this predictive exercise fall more into the category of
what we can and cannot predict with reliability. Total stability for staged
construction was reasonably well predicted. Incremental stability was not.
Incremental pore pressures were not too well predicted as to specific values
but were good enough for stability computations. Deformations were poorly
predicted, with everyone, including the investigators, surprised by the extent
of the failure. Shape, limits, and specific location of a failure surface are
relatively unpredictable. In his analytical review of the 1-95 test fill in his
Terzaghi Lecture, Leonards (1982) concluded that a designer of an em-
bankment on soft clay "must continue to rely heavily on experience within
a given geologic region--experience that has been carefully digested in
terms of local practice for site investigation, sampling, testing, and analysis."
This program had a later desirable outcome. It led to Ladd's Terzaghi
Lecture of 1986 (Ladd 1991).
Underpredlctlons~ - I ~ - - Overpredlctlons
(n (46)~
z 16
O J 30~~ Medians
i--
(J 14
Downloaded from ascelibrary.org by Florida Atlantic University on 01/22/15. Copyright ASCE. For personal use only; all rights reserved.
r~ U/ ~JMeans
uJ 12
m
Q. ,/
u. 10
O ~ BOT.._._HHPILE_ S_
rr 8
uJ
m 6
z 4
2
t
00 0,5 ,0 1.5 2.0
RATIO OF
PREDICTED TO OBSERVED
CAPACITY
FIG. 24. Histogram of Ratio of Predicted to Observed "Best Estimate" Ultimate
Capacity for Groups of 15 and 24, Both Piles
not differ materially from those of the total group. Thus, we can use the
lower-bound estimates of the partial group to represent the total group.
The upper histogram in Fig. 25 of the ratio of predicted to measured
capacity is for best-estimate capacities for the predictor group of 15; the
inverted lower histogram is for the lower-bound estimates. Note that the
median dropped from 0.93 to 0.60 while the mean shifted down from 1.02
to 0.77. Note also that three of the lower-bound predictions were still 30-
50% greater than measured (two for the pipe pile and one for the H-pile).
Remember, these are predictions that were to have had 90% confidence
that they would not be exceeded.
Looking at the data another way to evaluate the degree of conservatism
each predictor thought he was injecting into his lower-bound estimate, the
histogram in Fig. 26 is of the ratio of the individual lower-bound capacity
to the respective best-estimate capacity. Of the six values in the range from
0.9 to 1.0, there were two values of 1.0, both for the pipe pile. The medians
and the means were practically identical for the two piles and the distri-
butions for each pile were identical. These data suggest that the predictors
would perhaps apply an average conservative multiplier of 0.75 to their best
estimate before applying a usual factor of safety. If they would have used
a typical value of 2.0 to arrive at a working design load for these two piles,
these "designing" predictors would have achieved real factors of safety
ranging from 3.7 to as little as 1.43.
Another measure of perceived reliability of pile-capacity predictions was
obtained in a survey reported at the 1985 International Conference on Soil
Mechanics and Foundation Engineering (Focht and O'Neill 1985). Nearly
25% of the respondents to that survey thought their pile-capacity estimates
1677
o ,, ,.~ F,.~,. ~ .
~i~~ ilO~
~o., ~ 1 . o Y/~I.S'. o
r.,, 9,9,"~- /A-- 1
I , - EI~_~,~ ,ow~. ~o,~
BOUND
6 i ~-~Mean -~ 20 O.
8
.ed,an--~t 30
10
RATIO OF
PREDICTED TO OBSERVED
CAPACITY
FIG. 25. Histograms of Ratio of Predicted "Best Estimate" and "Lower Bound"
to Observed Ultimate Capacity for Group of 15, Both Piles
14
30 45
values
Median
12
Mean --~ 40
z z
o_ 10 35 O
-/i/
////
25
==
uJ
i.i. ////
z/x/
U.
O
Iz
6
,///
Vl/ ....
llll
~ 2o O
I'-
uJ y/Z,H, 2:
/iiJ
till
i/ll
UJ
=! 4
Y/A U.I
eL
Y/z
2
~//~
0 . . . .
.... ~ o
0 0.2 0.4 0.6 0.8 1.0
RATIO
LOWER BOUND CAPACITY
BEST ESTIMATE CAPACITY
1678
pile foundations.
The choice of a specific factor of safety for a specific problem must be
based on a relatively large number of considerations, which I have grouped
into four categories.
All of these considerations enter into the confidence level the predictor
wishes to achieve. Some of the considerations are amenable to statistical
and probabilistic analyses, but others can only be approached on a judg-
mental basis because statistical data are not available. Not included at all
in any of these factors is the uncertainty of construction quality. Many
failures of pile foundations in particular have been the result of construction
defects rather than of design. But usually this uncertainty is not included
in the design factor of safety.
Load and resistance factor design (LRFD) is rapidly taking over structural
design of steel and concrete structures (Salmon and Johnson 1980; AISC:
Manual 1986). It is also entering geotechnical foundation design, particularly
for piles. All of the foregoing comments on conventional factor of safety
are also applicable to the resistance factor to be included in an LRFD
approach to structure foundations.
The judgment of the designer that his factor of safety or resistance coef-
ficient is fully compatible with all of the factors entering into the process
including the loads and the structure characteristics is of vital importance.
The experience of the predictor with the characteristics of the soil formation,
the type of supported structure, the economic and social consequences of
failure, and other factors all must be taken into account even if only in a
general automatic, almost subconscious, step in the mental process of the
predictor.
Before concluding, there is one additional point that should be made.
Good geotechnical engineering requires broad understanding of civil en-
gineering in general. A geotechnical engineer must also have a good under-
standing of geology. He or she has to be a civil engineer first, with an
appreciation for structural design, elements of projects such as water-supply
projects, drainage, highways, bridges, waterfront facilities, dams, construc-
tion procedures, and the like. The geotechnical consultant must not lose
sight of the fact that his or her technical solution (and prediction) must be
compatible with the remainder of the project and its constructability.
1679
judgment is the major cause for four out of the six examples. Inadequate
modeling of the soil properties and inappropriate analytical methods played
a role as either a major or minor cause in four of the six examples. It is
very often more important to know what we do not know than what we do
know.
Returning to the predictive process, the six critical elements could be
considered steps within the process. However, if that were done, an ex-
tremely important first step would be missing. Therefore, I have added as
the first step in the following list, which outlines the prediction process, the
application of sound judgment (and perhaps intuition) to the problem at
hand to select the approach to the prediction and to define the critical
components of the process for the particular problem.
Stratigraphy @
Properties @ @
Analytical @ | | |
History 9
Judgment @ @|
Intuition (Not Applicable)
I
Major @ Minor
The absence of step 1 in the foregoing list may not have influenced the
validity of the example predictions that were described. There are, however,
Downloaded from ascelibrary.org by Florida Atlantic University on 01/22/15. Copyright ASCE. For personal use only; all rights reserved.
ACKNOWLEDGMENTS
The writer is grateful to the many associates and clients that contributed
to the personal learning process that led to this paper. In particular, ac-
knowledgments are due to Ralph B. Peck, Carl W. Fenske, Charles E.
Williams, Bramlette McClelland, C. A. Swanlund Jr., Richard J. Finno,
and Francisco Silva-Tulla, who reviewed and contributed to sections on the
individual examples. Special thanks go to my son, John A. Focht III, who
assisted materially in development of the paper's theme and content.
APPENDIX. REFERENCES
American Petroleum Institute (API). (1969). "Recommended practice for planning,
designing, and constructing fixed offshore platforms." Rep., A P I RP2A, 1st Ed.
Angemeer, J., Carlson, E., and Klick, J. H. (1973). "Techniques and results of
offshore pile loading testing in calcareous soils." Proc., 5th Annu. Offshore Tech-
nology Conference, Offshore Technology Conference, Houston, Tex., Vol. 2,677-
692.
Casagrande, A. (1965). "Role of the "calculated risk' in earthwork and foundation
engineering." J. Soil Mech. and Found. Div., ASCE, 91(4), 1-40.
1681
1683