GSP 158 Contemporary l.

ssues in Deep Foundations

Defect Analysis for CSL Testing

Garland Likins 1, P.E., M.ASCE, Frank Rausche2, PhD. , P.E., M.ASCE,

Karen Webster3 , P.E. and Anna Klesney4 E. I.T ., A.M.ASCE

ABSTRACT: Cross-hole Sonic Loggi ng (CSL) has become a common method to

e valuate the integrity of drilled shafts. However, the interpretation of test
measurements by thi s me thod requires some judgment and experience. Many
previous proposals rely solely on the arrival time, or wave speed. A proposed method
which takes into account both arrival time and signal strength is presented. [f defects
are de tected, then in some cases a tomography analysis may be helpful to further
quantify the result. The situations where tomography is useful will be discussed.
Recommendations are given for the type and amount of data required for a
tomography analysis. This discussion is illustrated with a case history of a test shaft
with purpose built defects to demonstrate the advantages of these evaluation methods.

INTROD UCTION
Drilled shafts are an option for deep foundations. Since they often have larger
diameters, they usually carry large loads and are relatively few in number compared
to a driven pile fou ndation to support the same total load. Therefore, the integrity of
each shaft is critica l to the overall performance of the structure, and good construction
procedures and superv ision improve quality assurance (O' Neill and Reese, 1999).
Unfo11unate ly, such " inspection only" has proven to be inadequate to assure quality.
Cross-hole Sonic Logging (CSL) is widely used to evaluate the concrete and shaft
construction quality of drilled shafts, and its use has been rapidly expanding (O'Neill
and Reese, I 999). State Departments of Transportation are increasingly relyi ng on , in
addition to available information on installation, the CSL testing for acceptance or
rej ection of drilled shafts for their projects. The CSL test results search for changes
in concrete wave speed, which is attributed to changes in concrete quality or strength
due to the presence of a defect. The defects identified by CSL include but are not
limited to lower quality concrete due to mixing with dri lling slurry, honeycombing,
necking and oil intrus ions, and soft toe conditions.

1
President , Pile Dynamics. Inc.. 4535 Renaissance Parkway. Cleveland OH 44 128,
PH : (2 16) 83 1-6 131. email: garland@pile.com
2
President. GRL Engineers, Inc .. 4535 Renaissance Parkway, Cleveland OH 44 128
~ Engineer, GRL Engineers. inc.. 99 12 Colvard Circle. Charloue NC 28269
4
Engineer. GRL Engi neers. Inc., 4535 Renaissance Parkway. Cleveland OH 44128
GSP 158 Contemporary I ssues in Deep Foundations

O'Neill and Sarhan (2004) in a survey of over I 0,000 shafts reported detection of
" flaw s" in about 20% to 25% of all shafts using various Non-Destructive Evaluation
(N OE) methods such as Cross-hole Sonic Logging (CSL) or Gamma Gamma
Logging (GGL), and assert that "since these flaw are identifiable by NOE, they are,
by definition, not ' minor. "' They further noted that 20% of the shaft s tested by
CA LTRANS were rejected. From this survey, they found the most probable location
for a flaw is in the upper 5 shaft diameters, which is critical structurall y but
fortunately the eas iest to repair by excavation or other means. Obvious ly, their study
de monstrates the necessity for good construction inspection, but also for post
construction NOE inspection of the shaft to assure its integrity.

CSL TESTING METHOD

The CSL test is performed on drilled shafts with access tubes installed during
construction, or cored after construction. The recommended number of tubes for a
shaft is at least one tube for each 0.3 meter of shaft diameter, with ideally a minimum
of 4 tubes. The tubes may be either steel or PVC, but must be filled with water
during c uring to insure proper bonding of the tube to the conc re te, and during testing
to facilitate trans mitting the signal. During the CSL testing, a transmitter probe is
lowered into one of the access tubes while a receiver probe is lowered si multaneously
into a second access tube. Both probes are cente red in the tubes by fl ex ible
"centralizers'', which have the added benefit of providing a cleaner signal by
el iminating the receiver rubbing the tube. The transm itter probe gene rates ultrasonic
vibration pulses which travel from the transmi tter through the concrete to the receiver
which converts this vibration back to an electrical signal. The probes are typically
placed at the bottom of the access tubes and then simultaneously pulled to the top.
The signals are transmitted and recorded typically every 50 mm and the locations of
the indi vidual probes are separately recorded by digital e ncoders. The test is re peated
for each tube pair combination, or "profile", to look for defects in diffe rent quadrants
of the shaft. The test is well described in ASTM 06760 (2002).
The cost of access tubes and even the cost of CSL testing are modest compared to
the cost of the shaft itself, particularly for larger shafts. Installing access tubes in
every shaft allows testing any shaft should doubts arise during its installation, and the
mere fact that a shaft could be tested often will lead the contractor to more careful and
bette r construction practices since a fai led test might result in a rejected shaft.

CSL EVALUATION
While the testing procedure for CSL is relatively well defined, the appropriate
method fo r interpretation of the results is not as well defined and improves over time
due to advances in theory and software. The most common criterion for shaft quality
assesses the first arrival time (FAT). First arrival time (FAT) may be determined by
when the signal first exceeds simple amplitude thresholds, or preferably by using
pecial advanced image processing tools. The wave speed of the concre te, which is
related to conc rete strength, can be calculated from the tube spacing divided by this
first arrival time. Another indicator of concrete quality is the s ignal strength.
GSP 158 Contemporary Issues in Deep Foundations

Integrating the absolute value of the signal for a defined time results in the s ignal
'·energy." Low energy is usually the result of a defecc or poor concrete quality. As
seen in Figure I a, lhe s ignal strength is large and the well defined first arrival time
occurs relatively early compared to the signal in Figure Jb, from a local ion of a
known defect. The data presented here is from a special test shaft constructed with
known defects for the purpose of furthering CSL development. The first arrival time
for this defecc is not as well defined, and engineering judgment is then helpful.

-------------- 4-- - '

--------~--------------------------------------------
Figure la (top): s ignal at 10 m depth (top) at location of good concrete

Figure 1b (bottom): signal at 7 m depth (bottom) at location of known defect

(vertical scale is normalized "signal strength")

FAT (or derived wave speed) and the received s ignal •·energy" both indicate relative
quality of the concrete between the transmitter and the receiver. Unfortunately wave
speed alone cannot be used as an absolute indicator since the tubes are often not
parallel, particularly for PVC tubes. Since arrival time also varies with tube spacing,
FAT is usually assessed by comparing it to FAT in a nearby zone of good concrete.
Therefore, looking for relative increases in FAT within relatively short distances
along the shaft length is the generally accepted method of locating areas of concern,
often cal led "anomalies".
In Figure lb, the obvious delay and signal reduction results from the purpose built
defect, created by inserting a substantial piece of Styrofoam within the reinforcing
cage, thus blocking the direct signal path . The small signal is likely caused by signal
travel outside the cage with reflections from the concrete-soi l interface. The engineer
can use judgment to manual ly override the automatically selected FAT in any CSL
testi ng system (and is required for systems with manual on ly selection abi lity), since
any FAT selection is subject to selection of the image processing control parameters.
In Figure I b, arguments can be made for either a slightly earlier selection, or a
considerably later selection. The relative energy can influence the engi neer's
judgment when deciding how aggressive to be on the selection of FAT. An early
"aggressive" or later "conservative" selection of FAT approach may be taken.
GSP 158 Contemporary Jssues in Deep Foundations

Whether an aggressive or conservative approach is taken will affect the FAT, but will
have little influence on the re lative energy. Therefore, relative energy can be a good
tool for ass isting the engineer in FAT selection. In the present example, a later
conservati ve se lection seems appropriate fo r the known defect. However, even the
very earliest aggressive selection would c learly indi cate a defect. In this case the very
low re lative e ne rgy is du e to the reduced signal trans mission in the gene ral region;
signals trave l not only on the direct path but also indirectly through a bulb or cone
due to particle refractions. When part of the full signal transmission path zone is
blocked by a defect, a in this case, the resulting energy clearly reflects this defect
condition. Low energy can alternatively be due to poor quality concrete. Repeating
the CSL test after a longer waiting period may be helpful if the concrete was not
sufficiently cured at the time of the first CSL test. If the second test after a much
longer curing time confirms the earlie r test, then the anomaly is real.
Once the FAT and signal e nergy/strength reduction have been dete rmined for the
e ntire profile length, the shaft integrity may be evaluated with the followi ng scale:

(G) Good - FAT increase 0 to 10% and Energy Reduction < 6db
(Q) Questionable - FAT increase 10 to 20% and Energy Reduction < 9db
(P/F) Poor/Flaw - FAT increase 2 I to 30% Q!. Energy Reduction 9 to 12 db
(PID) Poor/Defect- FAT increase> 3 1% or Energy Reduction > l 2db

Flaws (P/F) shou ld be addressed if they are indicated in more than 50% of the
profiles. Defects (P/D) must be addressed if they are indicated in more than one
profile. Addressing a flaw or defect should include, at a minimum, an evaluation by
tomography if the area of concern is localized, and/or additional measures such as
excavation, core drilling, or pressure grouting. Defects or flaw s indicated over the
entire cross section usua lly require repair or shaft replacement. Th is scale, based on
the author's experience, adapts a common scale used by many State departments of
transportation , separates the more marginal Flaw from the more serious Defect, while
assigning actual numerical values to the Energy Reduction rather than current USA
practice of vague statements about energy. Both Fre nch (2000) and Ch inese (2003)
national standards use numerical values o f Energy Reduction when evaluati ng CSL.

DEFECT ANALYSIS
After data co llection, the engineer must compare the processed resu lts with the
rating scale and present a report. When multiple shafts are tested, particularly for
shafts with many tubes creating multiple profiles, an automatic evaluation technique
is helpful to summarize the results. Since tubes are often not paralle l and sometimes
not even straight, a method that follows the general FAT trend improves finding the
location of local defects. This process can be accompli shed with " fillers" that take a
running average as a "baseline." Using the average of typically 75 consecutive data
samples at the typi cal 5 cm vertical resolution re presents the tre nd for the local 3.75
m of shaft length. If the actual FAT compared with this moving baseline is
s ignificantly increased, or the energy significantly reduced, compared with its
baseline and user input limits, an a nomaly is defined at that depth location.
GSP 158 Contemporary Issues in Deep Foundations

Figure 2 shows the analysis for tube combination 3-5 for the same test shaft with
purpose built defects. The "waterfall diagram" at the right presents a "nesti ng" of the
data. Sections of low color intensity, such as at 1, 7 and I0.5 m depths, indicate
relatively low signal strength. The left edge of the waterfall maps the FAT. The
processed data at the left in Figure 2 displays the computed wave speed (thick line)
and the relative energy (thin line) plotted on a log-energy graph, with lower values to
the right. The horizontal dashed lines identify locations of peak concern based on the
FAT increase limit of l 0% and energy reduction limit of 6 dB. With the proposed
rating scale, the FAT increases of 29, 28 and 12, and energy decreases of 7.3 , 9.4, and
7.5 dB for approximate depths of 1, 7 and 10.5 m respectively would rate the upper
two anomalies as Poor/Flaw, and the lower one as Questionable. The 9.4 dB at 7 m
might cause an engineer to judge the local FAT selection more conservatively and
might easily result in a Poor/Defect rating.

POI t ·~> ·V ·~> POI 1 ·~> ••> ·~1

3 5 (F0,2) 3 5
L-13.7::! meterG- ·~> )> L•13.72 me.te~ .:> fl
5pac:•no•0.88..J m /.t• SP'\CH'\Q•0. 88.a m /a>
ll-=
~;.;~9
:. : T
·,",;.;;;~=6="='5::2: --;;:::::;::::;(:::J.::cl__·'·_' ...;:t>;._·!•"'1 t-=~=sJ=
0
~;.;..
t~ =
0
a~.,
ix~-~:-5'---~~~--·!•--:.~-·-!>-ti
II Wdvc:::.-&>eed (mlsec) Time {ms)
2000 ~000 6QOO BOOO . 1S .2 . 25 .J .JS .4 ,45 .5 .SS .G

---· ---~-
0 0

1-- J
·~

F:
I iz
!g.,!'
...

!] ..; .
0 I -- ~--L~.
..
{ ..
~ l J
~

::!
1--- r- !
'.:::

Encro~'~/:,g)
low

Figure 2: CSL "waterfall" diagram in right, processed results in left (wave speed
heavy left line, energy thin right line)
GSP 158 Contemporary Issues in Deep Foundations

TOMOGRA PHY
Allhough CSL data can be evaluated fo r each tube combination pro file, combining
a nd revie wing al l available profil es leads to the best e ngineering evaluation. If a
defect is fo und at the same depth in every profil e, it is clear that the shaft has a defect
covering the e ntire section and that remedial action is then required. Such action
might include excavation and repair if the defect is near the to p of the shaft, or
perhap coring and pressure grouting if the defect is at depth, or in some cases
a bandoning the shaft and constructing a repl acement shaft(s). If no defect is found in
a ny pro fil e, Lhen the entire shaft is satisfactory and acceptable and no additional
analy is is necessary.
However, some CSL tes ts reveal defects in some profil es, but not in all profiles, and
the n the question arises as to the lateral extent and location of the defect. Certainly
the magnitude of any FAT de lay or energy reduc tion in any profile would enter into
any evaluation. But the lateral extent of the defect across the section is also
important. By visua lly re viewing aJI the profiles. the e ngineer can assess the possible
size and depth of a defect. If Lhe defect covers many pro fil es, or is nea r the shaft top,
making it signifi cant, it might serious ly compromise the shaft integrity.
S ince concrete strength is related to the conc rete wave speed, a dete rmination of
wave speed at each point in the shaft can help assess the shaft acceptability. The
severity of a local defect, defined by substandard concrete wave speed, can be
assessed by both magnitude and lateral exte nt perhaps most easil y by tomography.
Arri val time data from all profiles, locations of each probe fo r each signal record, and
tube geometry can be input into a single three d imensional analysi . For a grid of
node points, the wave speeds in each node can be adjusted to minimize the errors
between calcul ated and observed arri val times fo r all travel paths (J ie et al, 1998).
A lthough the re are other methods of tomography, this node matri x method, although
computationall y intensive, is the most reliable and therefore this paper focuses on this
method. The to mography analys is results in a profile of wave s peeds as a function of
cross section and shaft le ngth.
T omography results can be shown in an overall 3-0 presentation, or 2-0 "slices"
can be made e ither horizontally or vertically in the shaft. A horizontal 2-0 slice at
the depth o f interest proba bly shows the most c lear presentation of defect extent and
location within the section, and is useful to guide the construction team into better
selections of coring locations for verifications or remedi al procedures. If a lowe r
bound threshold of acceptable wave speed is defin ed, the ana lysis can calculate the
pe rcentage of the s lice of c ross section falling below this limit.
The data required for a tomography analysis depends on several factors. The more
da ta that is available, the more accurate the analysis is likely to be. However, the
amount of data available is practicall y limited by the num ber and locations of the
access tubes. The more tubes that are avai la ble, typically true for larger diameter
shafts, the more information that can be naturally available for the tomography
a na lys is. For shafts with only 4 tubes, 6 possible tube combinations are possible with
only 2 c rossing the shaft inte rior, making the quadrant location of an interior defect
difficult to dete rmine with precision, while for 8 tu be shafts there are 28 tube
combinations, including 20 c rossing the interior. T able I and Figu re 3 re late the
GSP 158 Contemporary Issues in Deep Foundations

numbe r of perimeter and the number of interior paths relative to the number of access
tubes. Thus for shafts with preferably at least 6 tubes, any interior defect can be
located quite well since it likely crosses many interior paths.

Table I. Number of paths versus number of access tubes

Tubes Perimeter paths Interior paths Total paths

4 4 2 6
5 5 5 10
6 6 9 15
8 8 20 28
10 10 35 45

Figure 3: Potential scan paths for 4, 6 and 8 lube shafts

The exteri or or perimeter paths howeve r require extra cons iderations when locating
a defect. When a defect is located in the perimeter in normal testing (probes bei ng
" parallel" or " level" in the tubes), it is generally not poss ible to tell if the defect is
located closer to one tube or the other. However, if additional scans are made with an
"offset" (e.g. one probe rai sed in relation to the other, and the n repeated with the
opposite probe raised), then these three measure ments of parallel scans and two offset
scans fo r each tube combination can help locate the defect relative to the tested tubes
since the apparent depth location of the defect will shift as seen in Figure 4 .
Although interior offset tests offer little benefit to tomography when the number of
access tubes is 6 or more, offset tests are useful for the interior profiles in tomography
for shafts with few tubes. The general suggested rule for shafts with 5 or fewer
access tubes is that "offset scans" be performed for all tube combinations when a
defect is located in that tube combination, while for shafts with 6 or more tubes,
parallel scans are sufficient for the interior and the extra offset scans are necessary
only for the perimeter. Of course there is no benefit to any offset scan for any tube
combination if the parallel or level data does not reveal any defect The goal is to
obtain necessary and sufficient data for the analysis, without having to collect useless
extra data that adds to the testing cost but provides no real resolution improvement.
GSP 158 Contemporary issues in Deep Foundations

Figure 4: Effect of " probe offsets" in apparent defect location fo r trans mitter en and
receiver (R) in access tubes

In a perfect world, spacings between all tubes are equal and known with precision,
the tube are precisely parallel, and the probes are centered in the tubes. In the real
world, tubes are often not uniformly spaced and not even parallel. Smart logic is
required to correct for non-parallel tubes. The most accurate wave speed is
determ ined by the main cross diagonals s ince the tubes are more parallel and the
percentage distance traveled in concrete is larger due to the large spaci ng. Si nce the
depth to each probe is measured, the actual travel distance in the concrete is also
known from geometry, and the arrival time can be adjusted to compensate fo r the
wave travel in water.

CASE HISTORY
The s haft with purposely installed defects built al the author's Ohio offi ce for
developing and testing the hardware and software can be effectively used to evaluate
the methods. The shaft is 1.5 m in diameter and 12.2 m in length. It was cast with six
steel and 2 PVC tubes, and it can be me ntioned that even over three years later the re
is absolutely no evidence of "debonding" of the PVC tubes (cast in the dry method,
a nd water in the tubes has been continuously maintained). Four major defects were
installed in this test shaft. A soft toe was c reated on one quadrant on ly us ing sand
bags. The prev ious ly mentioned defect at about 7 m depth was a 150 mm thick
Styrofoam insert covering half the inte rior cage (e.g. "half moon" shape). T wo 400
mm diameter buckets were inserted, one about 1.5 m above the bottom fill ed with