You are on page 1of 11

C

9
H A P T E R

Liquid Penetrant Testing


Crack Detection
Capabilities and Reliability

Ward D. Rummel, D&W Enterprises, Limited, Littleton,


Colorado

© Copyright 1999 by D&W Enterprises, Limited. Reprinted with permission.


liquid penetrant detection capabilities
Introduction may not be required. Inspections
requiring detection of small cracks and
Liquid penetrant testing is one of the critical inspection applications may
most widely applied methods of require demonstration of detection
nondestructive testing because of the low capabilities and rigid process control that
cost of application, the perceived includes periodic revalidation of detection
simplicity of the process and its ability to capabilities.
be applied to complex shapes with little
process adjustment. Liquid penetrant
testing is not, however, absolute and is
not capable of finding all cracks. The Probability of Detection
end-to-end liquid penetrant process is
bounded by a lower limit for the
(POD) As Measure of
detection of small cracks (detection Liquid Penetrant
threshold). In addition, the detection Performance
threshold is not a constant value but
depends on multiple parameters inherent The recognized metric for ascertaining the
to the test object and liquid penetrant detection capability for a liquid penetrant
procedure applied or are a function of procedure is the probability of detection
rigid process control in application. (POD). A characteristic POD curve is
Although some confidence may be generated by passing a large number of
provided by the ability of a liquid cracks of varying size through a liquid
penetrant procedure to reveal a known penetrant procedure and recording the
crack in a test coupon, the brightness or cracks that are detected. A standardized
contrast intensity of the liquid penetrant data analysis procedure is then used to
indication may be such that it would not produce a plot of probability of detection
be detected under field conditions if its as a function of crack size (typically crack
presence and location were not known. It length). Figure 1 is an example of a POD
is foolhardy to assume crack detection if curve for a given liquid penetrant test
the indication is small or dim or both. procedure.
The inspection result of importance is not By convention, the threshold detection
the smallest discontinuity detected but the point is that point where the POD curve
largest discontinuity missed by application crosses the 90 percent threshold and is
of a given liquid penetrant procedure. referred to as the 90/95 probability of
In many liquid penetrant testing detection value (assuming the number and
applications where requirements are for distribution of crack sizes meet the criteria
the detection of large cracks, exceeding for standardized analysis). In the example
25 mm (1 in.) long, demonstration of shown, the single valued detection
capability is at the 3.5 mm (0.14 in.) crack

FIGURE 1. General form of probability of detection curve for liquid penetrant testing
procedure. Accepted 90 percent threshold detection point is noted.

100

90
Probability of detection (percent)

80

70
Threshold
60

50

40

30

20

10

0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)

Actual crack length, mm (in.)


Legend
= predicted probability of detection
X = hit datum (along top edge) or miss datum (along bottom edge)

276 Liquid Penetrant Testing


length. Although some cracks smaller variable that is most often measured,
than the 3.5 mm (0.14 in.) crack length reported or claimed.
were detected by the applied liquid The discontinuity form and state are
penetrant procedure (Xs plotted at the more difficult to quantify but must be
100 percent detection level) , the considered in each liquid penetrant
procedure did not reliably detect cracks application. It is obvious that a
smaller than the 3.5 mm (0.14 in.) crack discontinuity must be clean and dry to
size. support reliable liquid penetrant
The capability of the procedure is stated performance. Less obvious is the closure
as 3.5 mm (0.14 in.) crack length. and stress state of a discontinuity (crack).
The reliability for detection of cracks at It is known that the brightness of a liquid
the 3.5 mm (0.14 in.) level is 90 percent penetrant indication is decreased for a
and is a measure of the repeatability and crack under compressive load, thus small
reproducibility for finding a crack at that cracks will not be detectable under some
size. loading conditions. In addition, the past
The confidence level for detection at the load history (in particular, a high
3.5 mm (0.14 in.) level is derived from the overload) may change the configuration
number and distribution of cracks used to of a crack such that a liquid penetrant
determine the detection capability. indication is segmented, with the crack
ends visible and the connecting ligament
producing a very dim indication.
The test materials (liquid penetrant
Multiparameter Process materials) are known to affect the
Liquid penetrant testing is a sensitivity of a liquid penetrant and liquid
multiparameter process. Changes in one penetrant materials are classified by their
or more of the process parameters can ability to produce an indication under
significantly change the liquid penetrant standardized test conditions. A variety of
crack detection performance capability. liquid penetrants of differing sensitivity are
Liquid penetrant process performance commercially available and are intended
depends on the following: (1) test object, for use in different applications. For
(2) discontinuity type, (3) discontinuity example, a high sensitivity liquid
size, (4) discontinuity form and state, penetrant would not be suitable for the
(5) test materials, (6) test equipment, inspection of a porous material surface
(7) test process, (8) test environment, because of the background indications
(9) test procedure applied and (10) human produced.
factors. The inspection equipment introduces
The test object material, shape variables in the test process because of the
(configuration), surface texture (porosity inherent limits and variations for process
and surface finish), stress state, surface control. For example, a temperature
condition and object size are parameters nonuniformity in a drying oven will add
that may affect liquid penetrant process to the end-to-end process variance.
capability. For example, titanium alloys Ultraviolet radiation intensity, white light
wet differently than aluminum alloys background, tank contamination etc. are
used in aircraft. variables added by the type of equipment
The discontinuity type may be a or equipment maintenance.
significant factor; for example, a The inspection process may introduce
condition for detection is that the major variance in end-to-end
discontinuity must be open to the surface performance. For example, dry developer
and must have an opening such that versus wet developer; hydrophilic
liquid penetrant can be retained using the emulsifier versus a lipophilic emulsifier;
applied liquid penetrant procedure. It is or inspection without a developer.
obvious that a crack that is covered by a The inspection environment can produce
paint layer or smeared metal from a wide variances in process performance.
machining or grinding process will not be Consider variance between in-line
consistently detectable. processing in a factory environment
The discontinuity size — i.e., length, during component overhaul versus field
depth and opening — greatly affect inspection in the arctic.
detectability. The brightness of a liquid The inspection procedure applied may
penetrant indication for a small crack (less cause wide variance in process
than 0.75 mm [0.030 in.]) will be performance in the form of processing
significantly less than that for a larger times. The most critical process parameter
crack because of the reservoir size. The is the remover time or emulsification time
ability of a human reader to discriminate but wash times, dwell times, drying time
small indications is typically greater than etc. can add significant performance
0.75 mm (0.030 in.) in length. Because variance.
discontinuity size (length and depth) is a Human factors variables are listed last
basic material parameter used in static because the human operator has little
design and life cycle analyses, it is the chance for discontinuity detection if other

Liquid Penetrant Testing Crack Detection Capabilities and Reliability 277


end-to-end parameters are not controlled. reproducible results. Indeed, liquid
Human factors include the traditional penetrant testing without attention to
health, attitude, skill level and attention process control is often an exercise in
to detail that are related to training, parts washing. Process control aids and
experience and recent experience with the measurement tools such as ultraviolet
same or similar parts, equipment and radiation intensity measurement; artifact
processes. panels such as the testing and monitoring
(TAM) panels; and periodic measurement
of liquid penetrant material properties are
essential to reliable liquid penetrant
Process Control Is Critical process performance. In addition,
to Liquid Penetrant attention to detail in equipment and
processing materials unique to a particular
Performance test object or process line is required for
The multivariate nature of a liquid consistent performance.
penetrant process demands process Periodic proficiency demonstration and
control to effect repeatable and validation for human operators is also

FIGURE 2. Effect of ultraviolet radiation level for liquid penetrant testing procedure (water wash
process without developer, on tightly closed fatigue cracks in cobalt alloy; with 55 to 107 lx
[5 to 10 ftc] white light illumination): (a) 4 W·m–2 (400 µW·cm–2) ultraviolet radiation;
(b) 12 W·m–2 (1200 µW·cm–2) ultraviolet radiation.1
(a)
100
90
Probability of detection (percent)

80
70
60
50
40
30
20
10
0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)

Actual crack length, mm (in.)

(b)
100
90
Probability of detection (percent)

80
70
60
50
40
30
20
10
0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)

Actual crack length, mm (in.)

Legend
= predicted probability of detection
X = hit datum (along top edge) or miss datum (along bottom edge)

278 Liquid Penetrant Testing


FIGURE 3. effect of developer for liquid penetrant testing procedure (water wash process on
tightly closed fatigue cracks in cobalt alloy): (a) without developer, (b) with developer.1

(a)

100

Probability of detection (percent) 90

80

70

60

50

40

30

20

10

0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)

Actual crack length, mm (in.)


(b)
100

90
Probability of detection (percent)

80

70

60

50

40

30

20

10

0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)

Actual crack length, mm (in.)

Legend
= predicted probability of detection
X = hit datum (along top edge) or miss datum (along bottom edge)

essential to end-to-end process control.


Additional training may be beneficial but
additional training without proficiency
Process Performance
demonstration is often a rerun of previous Assessment and
information and may have little benefit in Proficiency Demonstration
maintaining skill levels.
Variation in end-to-end liquid penetrant
process performance is readily observed
by comparison of probability of detection
(POD) data as a function of varying liquid
penetrant process parameters. Table 1 and
Figs. 2 to 5 are examples of the assessment
of liquid penetrant process parameters

Liquid Penetrant Testing Crack Detection Capabilities and Reliability 279


FIGURE 4. Effects of etching and proof test for liquid penetrant testing procedure (solvent
remover process; on tightly closed fatigue cracks in 6Al-4V titanium material): (a) as
machined condition; (b) after etch; (c) after proof load.1
(a)
100

90
Probability of detection (percent)

80

70

60

50

40

30

20

10

0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)
Actual crack length, mm (in.)
(b)
100

90
Probability of detection (percent)

80

70

60

50

40

30

20

10

0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)
Actual crack length, mm (in.)
(c)
100

90
Probability of detection (percent)

80

70

60

50

40

30

20

10

0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)
Actual crack length, mm (in.)
Legend
= predicted probability of detection
X = hit datum (along top edge) or miss datum (along bottom edge)

280 Liquid Penetrant Testing


FIGURE 5. Effects of etching and proof test for liquid penetrant testing procedure (solvent
remover process; on tightly closed fatigue cracks in AISI 4340 steel material): (a) as machined
condition; (b) after etch; (c) after etch and proof load.1
(a)
100

90
Probability of detection (percent) 80

70

60

50

40

30

20

10

0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45)(0.50) (0.55) (0.60) (0.65) (0.70) (0.75)
Actual crack length, mm (in.)
(b)
100

90
Probability of detection (percent)

80

70

60

50

40

30

20

10

0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45) (0.50) (0.55) (0.60) (0.65) (0.70) (0.75)

Actual crack length, mm (in.)


(c)
100

90
Probability of detection (percent)

80

70

60

50

40

30

20

10

0
0 1.3 2.5 3.8 5.1 6.4 7.6 8.9 10.2 11.4 12.7 14.0 15.2 16.5 17.8 19.1
0 (0.05) (0.10) (0.15) (0.20) (0.25) (0.30) (0.35) (0.40) (0.45)(0.50) (0.55) (0.60) (0.65) (0.70) (0.75)

Actual crack length, mm (in.)


Legend
= predicted probability of detection
X = hit datum (along top edge) or miss datum (along bottom edge)

Liquid Penetrant Testing Crack Detection Capabilities and Reliability 281


using the POD metric. Figures 2 and 3 interpretation problem is to identify the
show results with a heat resistant cobalt signal in a field that consists of signal plus
alloy resembling SAE AMS 5608D.2 noise.
Figure 2 shows the effect of ultraviolet Varying cases of relative levels of signal
radiation level on liquid penetrant and noise are shown in Fig. 6. Figure 6a is
performance. Figure 3 shows the benefit a case where the signal level is
of using developer on liquid penetrant significantly greater than the noise and
performance. Figures 4 and 5 show the clear discrimination is possible. Figure 6b
effects of etching and proof loading on is a case where the signal level overlaps
titanium and steel flat plate specimen. the noise and noise may be interpreted as
Human factors effects are reflected in all signal (false call) or a signal may be
data presented. Variations in human
factors were reduced by using the same
operators for each test sequence
presented. FIGURE 6. Three conditions of signal and noise, representing
The usefulness of the POD method of differing levels of defect detection: (a) clear signal-to-noise
liquid penetrant performance assessment discrimination; (b) small overlap of signal and noise, some
is self-evident for purposes of procedure false calls and misses; (c) overlapping signal and noise, poor
development, procedure improvement discrimination.
and the qualification and validation of
(a)
process and procedure. For critical
applications, POD qualification and

Probability density distribution


validation is often used as a requirement Noise Threshold Signal
for both facility and personnel Decision
Level
performance demonstration.

(relative units)
Human Performance in
Liquid Penetrant Testing
The individual performance of the human
operator is a significant element in
end-to-end process performance capability
and reliability. Inspection process Signal amplitude
parameters are readily measured and (relative units)
controlled by application of appropriate
tools when applied by knowledgeable (b)
supervision. The readout or interpretation
Probability density distribution

of liquid penetrant indications by human Noise Signal


operators is more difficult and
encompasses the multiple factors of the
(relative units)

workplace environment and development


of skill levels applicable to the procedure False calls
being performed.
Training and skill development are
complimentary but not interchangeable.
Training denotes transfer of knowledge
whereas skill development denotes
application of knowledge, procedure and
experience in performing a specific task. Signal amplitude
When an operator views a liquid (relative units)
penetrant indication, interpretation is not (c)
a yes/no decision task as commonly
viewed and desired in production
Probability density distribution

application but is instead a problem in


conditional probability. The operator is Noise Signal
presented with a signal characterized
(relative units)

primarily by brightness, size and pattern.


The signal is superimposed on a
background (nonrelevant indications
inherent to the test object and liquid
penetrant process — also termed clutter)
characterized by the same parameters but
of a lesser degree or level. For purposes of
discussion, the signal characteristics will
be termed signal and the background Signal amplitude
characteristics will be termed noise. The (relative units)

282 Liquid Penetrant Testing


TABLE 1. Accept/reject decision process.
Call Error Condition Decision
True positive none (discontinuity found when discontinuity is present) reject (correct reject)
False positive type II (discontinuity found when no discontinuity is present) reject (false call)
False negative type I (no discontinuity found when discontinuity is present) accept (miss)
True negative none (no discontinuity found when no discontinuity is present) accept (correct accept)

interpreted as noise (miss). Figure 6c is a


case where the signal and noise are FIGURE 7. Four possible outcomes of accept/reject decision.
superimposed and discrimination is not (Probability P is discussed with the published POD data.1)
possible. The greater the separation of
signal and noise, the greater the margin for Stimuli (discontinuity presence)
correct interpretation. Because signal is a
direct function of the size and brightness Positive (a) Negative (n)
of a liquid penetrant indication, large and

Nondestructive evaluation signal


bright indications should be readily

(Response to discontinuity
detectable and have a high probability of P(A,a) P(A,n)
detection. Positive (A) True positive (hit) False positive
No error Type II error
The varying cases of signal-to-noise
discrimination are consistent with the
POD results shown in Figs. 2 to 5. Larger
and brighter liquid penetrant indications
produce discrimination at a low level — P(N,a) P(N,n)
Negative (N) False negative (miss) True negative
i.e., developer versus no developer, higher
Type I error No error
ultraviolet radiation levels and removal of
smeared surface material by etching.
As a result of this observation of the
liquid penetrant indications, the decision
of a human operator has four possible noted on engineering drawings are below
outcomes. the capabilities of the best performing test
1. A discontinuity may be called when a facilities. Qualification and validation of
discontinuity is present (correct call). test facilities, procedures and personnel
2. A discontinuity may be called when are therefore required for fracture critical
no discontinuity is present (Type II components.
error). Two modes of qualification and
3. A no discontinuity condition may be validation are in general use; these are the
called when no discontinuity is full POD demonstration and the subset
present (correct call). demonstration (widely known as the
4. A no discontinuity condition may be 29-out-of-29 method). A full POD
called when a discontinuity is present demonstration is always the most rigorous
(Type I error). mode and is used when dealing with new
materials, new fabrication processes or
The conditional probability new liquid penetrant processing facilities.
discrimination decision process is shown A full POD demonstration is desirable for
schematically in Fig. 7 and Table 1. qualification of less experienced
inspectors. Although the full POD
demonstration is desirable, the cost of test
components and the time involved
Liquid Penetrant Process warrant consideration of alternative
Performance methods. A large amount of liquid
Demonstration penetrant process characterization data
has been generated and is available.1
The structural integrity of modern When requirements are such that a
engineering materials, components, significant margin is realized by the use of
structures and systems are increasingly standard, generic liquid penetrant
dependent on the ability of processing techniques, no demonstration
nondestructive testing processes to find may be necessary. When standard, generic
small discontinuities. Components process techniques are applied to the
requiring test capabilities below the detection of small discontinuities within
general detection level are often termed the envelope of previously demonstrated
fracture critical. Although well intentioned, capabilities, a subset demonstration may
the assumed capabilities for a given liquid be considered.
penetrant test operation are generally
incorrect and, in some cases, the criteria

Liquid Penetrant Testing Crack Detection Capabilities and Reliability 283


A subset demonstration is conducted principles and technique described are
by generating at least 29 representative used, capable and reliable performance
discontinuities of nominally equal size may be expected and demonstrated to
(but within the previously demonstrated ensure continuing excellence in liquid
size envelope and using the same penetrant processing and continuing
materials) and near or below the confidence in the safety and structural
acceptance limit for the test component. integrity of engineering systems.
The additional specimens are then The smallest discontinuity found is of
considered to be a subset of those academic interest. The largest discontinuity
previously used for full POD missed is of critical importance to reliable
demonstration. The 29 discontinuities are liquid penetrant testing.
then subjected to the candidate test
procedure (operator) and the inspection is
completed. Success requires that all 29 of
the discontinuities are detected. If a
discontinuity is not detected, the
detection failure must be resolved. If there
is a legitimate cause for missing the
discontinuity, the discontinuity may be
replaced and the process repeated. Failure
to resolve and rectify a missed
discontinuity requires full demonstration
by the POD method. When the subset
method is used for operator skill
demonstration and qualification, a failure
to detect may be resolved by additional
skill development (training alone is not
sufficient) and retesting at the end of the
defined skill development period. The
subset, 29-out-of-29 method provides the
required number of test opportunities for
a 90 percent confidence level for a single
discontinuity size and is consistent with a
single point on a demonstrated full POD
curve.
The rationale, complexity and
analytical methods for full POD
demonstration are beyond the scope of
this publication. The reader is referred to
MIL-STD-1823 for requirements and
methodologies for a full POD
demonstration.3

Summary
Liquid penetrant testing is an effective
and economical method of discontinuity
detection and is widely used in the
process of ensuring the safety and
structural integrity of engineering
materials, components, structures and
systems. Its wide use and superficially
simple application result in a wide range
of results that vary from consistent
detection of critical discontinuities to
parts washing exercises that do not add
value to the parts. Fortunately, it costs no
more to perform a valid inspection than it
does to conduct a parts washing exercise.
It is logical that a multiparameter testing
process requires attention to detail and
process control for successful application.
The tools and techniques for materials
and process control are readily available.
The end to end process performance may
also be quantified using the information
and techniques discussed herein. If the

284 Liquid Penetrant Testing


References

1. Rummel, W.D. and G.A. Matzkanin.


Nondestructive Evaluation (NDE)
Capabilities Data Book, third edition.
NTIAC DB-97-02. Austin, TX:
Nondestructive Testing Information
and Analysis Center (1997).
2. SAE AMS 5608D, Cobalt Alloy,
Corrosion and Heat Resistant,
Sheet Strip and Plate
40Co-22Cr-22Ni-14.5W-0.07La
Solution Heat Treated. Warrendale,
PA: Society of Automotive Engineers
(1995).
3. MIL-STD-1823, Non-Destructive
Evaluation System Reliability Assessment.
Washington, DC: United States
Department of Defense.

Liquid Penetrant Testing Capabilities and Reliability 285

You might also like