Professional Documents
Culture Documents
Ultrasonic Testing PDF
Ultrasonic Testing PDF
net/publication/314227016
CITATIONS READS
2 299
2 authors, including:
Abolfazl Zolfaghari
The Ohio State University
14 PUBLICATIONS 27 CITATIONS
SEE PROFILE
All content following this page was uploaded by Abolfazl Zolfaghari on 30 September 2018.
gated. The results of the paper illustrate that the detection and
Keywords missing rates are 52 % and 48 %. In addition, taking into account 95 %
Nondestructive testing, penetrant testing,
probability of detection, welded components,
degree of confidence, probability of detection for defects with length
surface crack equal to 3.73 mm is 90 %.
Disastrous failures in mechanical parts, these techniques may not be deterministic investigations were generally involved in a
due to some phenomena like fatigue and and inherits stochastic nature. For this rea- particular technique or material.
creep affecting the components, are feasi- son, reliability (probability of detection) Based on the sensitivity and reliability
ble. The mentioned failures occasionally and sensitivity (detection capabilities) are assessment, Simsir and Ankara [2] made a
result in the loss of certain properties and utilized to quantify results of NDT methods comparison between two nondestructive
can even lead to the loss of human lives. [2]. Due to this importance, a number of re- methods, liquid penetrant testing (LPT)
In order to prevent these types of undesir- searches have been implemented to deter- and eddy current. In another work,
able events, some fracture control or mine the sensitivity and reliability of NDT Stadthaus et al. [8] evaluated sensitivity of
maintenance procedures have been pro- techniques by conducting experiments or two techniques, penetrant and magnetic
posed and implemented in a wide range of other approaches. particles testing, using standard blocks
industries [1]. Fahr et al. [3] evaluated five common tech- containing natural and artificial defects
Characterization of flaws existing in the niques of NDT in detecting low cycle fatigue and determining visibility of generated in-
components, especially the description of of bolt holes generated in a disc engine. Few dications. Probability of detection of low
shapes, numbers and lengths is normally years later, Rummel and Matzkanin [4] pub- cycle fatigue in plates of Inconel 718 utiliz-
essential input data of the failure preven- lished results of studies on the capabilities ing fluorescent penetrant testing was ob-
tion approaches. To carry out the task, one and probability of detection of seven nonde- tained in [9]. Reliability of manual and au-
of the major techniques which is employed structive evaluation (NDE) methods on vari- tomatic ultrasonic inspection was gained
to catch the aim, is the use of nondestruc- ous materials and samples. In this area, a in defect detection of welds of pipes. The
tive testing (NDT) methods. The results of more comprehensive study is presented in types of defects of the investigation were
NDT are influenced by various factors such [5-7] from a large scale research. In the in- lack of fusion (LOF) and lack of penetration
as skills and emotional mood of the opera- vestigation, twenty five teams from ten (LOP) [10]. Reliability of ultrasonic time of
tors, quality of inspection tools, environ- countries participated to assess efficiency of flight diffraction (TOFD) on inspection of
ment conditions, types and shape of de- ultrasonic nondestructive testing and X-ray complex geometry components [11] and
fects, and even the geometry and material techniques in defect detection and charac- phased array on a number of specimens
of the components. As a result, detection terization of stainless steel welds. As men- with and without weld joint were also stud-
and characterization of defects assisting tioned in the following as instances, other ied in [12, 13].
Weeks et al. [14] investigated the proba- ity is between zero and one. In other words, with Pi: probability of detection in the ith
bility of detection of eddy current induced the technique is able to find totally defects interval of the length range, αi: length of
thermography in finding small cracks of with a length larger than a2 and does not defect, β0: location parameter (constant)
special alloys. Owing to reducing cost of detect any defects smaller than a1. and β1: slope parameter (constant).
conducting experiments, some authors There are two methodologies to determine The mentioned distribution is the most
have even tried to calculate probability of POD curve. Building relationship between consistent relationship in the POD calcula-
detection (POD) using other approaches. signal response and flaw size (â versus a), tion of NDT techniques. Using the relation-
For instance, by means of a proposed math- and constructing correlation between binary ship, the curve is drawn as depicted in Fig-
ematical model, then combining the results data and flaw sizes (hit/miss versus a) are ure 1. In most applications of POD curves,
of the model and Monte Carlo simulation the approaches. In the first approach, one of degree of confidence is considered in the
and attempt to find probability of detection the technique outputs (â), the amplitude of values which are obtained or drawn as POD
of automatic eddy current was made in [15, the signals in ultrasonic nondestructive test- curves by two bounds as shown in Fig-
16]. Likewise, in a similar manner, Carboni ing for example, is formulated as defect ure 1. It is employed to involve limitation
and Cantini [17] made an effort to deter- length (a) [19]. Another technique is based in the number of samples on predicting
mine reliability of ultrasonic inspection in on binary response. According to that, when probability of detection. In the practical ap-
railway axle evaluation. a specific defect with a particular length is plication of POD curves, the most impor-
In this paper, reliability and sensitivity detected, the result is defined as hit or it is tant concept is a90/95, which the researches
Materials Testing downloaded from www.hanser-elibrary.com by Hanser - Library on March 7, 2017
analysis of visible liquid penetrant testing described as miss or zero. When there is not carried out on reliability assessment of
in detecting surface cracks of welded com- any defect in a specific location of the com- NDT techniques have made attempt to re-
ponents are presented. To deal with this, ponent, but the NDT technique assumes ex- port it. By considering 95 % degree of confi-
the required data were gathered from in- istence of a defect there, it is defined as a dence, it defines the length of the defect
spections of some welded samples carrying false call. In this paper, owing to features of which the corresponding value of POD
natural surface cracks generated during the results of penetrant nondestructive test- reaches 90 %. Other concepts, which have
welding procedure. The data contained two ing, the second technique was chosen to cal- less degree of importance, are a50 and a90,
independent categories which were ob- culate POD curves. in which the probability of detection is
tained separately, by macroscopic assess- Using the instructions stated in [19, 20], equal to 50 % and 90 %, respectively.
ments and the results of the liquid pene- POD curve is obtained and drawn by hit/
For personal use only.
trant testing. Utilizing the obtained data miss data gathered from a number of de- Experimental procedure
and mh1823 POD software package [18, fects belonging to the defect length range.
19], the probability of detection as POD To carry out this, two methods may be typi- Material and samples preparation. Some
curves were drawn and sensitivity analysis cally used. In the first method, defects with researchers have used the first method in
was made. According to the results, in the known shape and length are inserted artifi- obtaining the reliability of NDT techniques,
size range of 0 to 5 mm defect length, the cially in the components. Then, employing artificial flaws, such as [10, 14], but others
probability of detection of the cracks with the nondestructive technique, all the de- utilized the second one, realistic defects,
2.58 mm length equal to 90 %. Considering fects are inspected to determine which like [2, 3, 13]. Since the shape and the situ-
95 % degree of confidence, the correspond- ones are detected, and, therefore, hit/miss ation of man-made flaws could be differed
ing length reaches 3.73 mm. The largest data constructed. In the second method, a from realistic ones and then the corre-
defect which could not be found by this variety of natural defects present in the sponding results may be deviated, applica-
technique is 2.2 mm, and the smallest one components are precisely characterized by tion of natural flaws is more accurate. In
which was detected is 1.7 mm. performing microscopic or destructive this study, the second method was chosen
evaluations. Next, hit/miss data is created
Reliability and sensitivity in in the same way of the first method.
nondestructive testing In the next stage, the range of defect
lengths is divided into a number of inter-
As stated before, the results of a specific vals (ai). The number of hit defects to total
NDT method may vary due to diverse fac- number of defects in each interval defines
tors. Thus, concepts of sensitivity and reli- probability of detection of the NDT tech-
ability are normally used to cover the men- nique in the mentioned intervals (Pi). The
tioned variations. Sensitivity describes de- obtained values and center point of the in-
tection capabilities and is normally tervals are candidates of the whole inter-
evaluated by analysis of detection rates, the vals. Thus, the gathered hit/miss informa-
smallest defect which is detected by the tion yields a number of (ai, Pi) pairs. Ex-
technique, and the largest one that is not trapolating the data attained as the pairs by
found. Reliability in nondestructive testing the log-logistic distribution results in a
is defined as POD. It is typically determined functional form, as shown in Equation (1).
as a function of defect length and is drawn The equation describes the relationship
as curves. Figure 1 shows an actual POD between the probability of detection and
curve for instance. As seen, by increasing defect length.
the length, the corresponding probability
exp(β0 + β1 ln(α i ))
rises. Next and beforehand a particular Pi = (1) Figure 1: POD curve with upper and
value of the length, a1 and a2, the probabil- 1+ exp(β0 + β1 ln(α i )) lower confidence limits
59 (2017) 3
292 PRODUCTION-ORIENTED TESTING/NONDESTRUCTIVE TESTING
of Ck45 steel, which is classified as me- These kinds of materials might prevent the [2] and Fahr et al. [3] evaluated liquid pene-
dium-carbon steel and was fabricated by flow of liquid penetrant to surface cracks, trant testing in finding fatigue cracks gener-
MAG (metal active gas) welding procedure. therefore, may not be detectable by the ated naturally on the surface of bolt holes of
Chemical compositions of the samples as technique. To cope with this problem, prior compressor discs. As stated before, Lively
weight percentage (wt.-%) are given in Ta- implementing the inspection, the surfaces and Aljundi [9] studied liquid penetrant test-
ble 1 in detail. Figure 2 shows a welded of the components are cleaned by various ing on finding defects presented in Inconel
specimen and nonjoined samples. methods from the mentioned substances. 718 and only reported values of a90/95. Due to
Surface crack description. The real In this research, liquid solvent, which is a not assessing the detection capabilities, the
number and length of surface cracks were product of Magnaflux Company and are results of Lively’s and Aljundi’s study can-
determined using microscopic evaluation produced as Magnaflux (SKC-S cleaner/re- not be employed in the comparison.
For personal use only.
of the weld surfaces. To accomplish this, mover) brand, were applied to the surface According to the table, liquid penetrant
images from the weld surfaces of the com- by spraying. testing is able to find twelve defects from 33
ponents (in 50× magnification) were taken Applying the liquid penetrant. The sur- in total (52 %) by considering all the test con-
and the taken pictures were analyzed. faces of the samples were covered by ditions described before. The number and
Liquid penetrant testing. Liquid pene- spraying the liquid penetrant (Magnaflux rate of defects, which is not able to be de-
trant testing (LPT) is an important method (SKL-SP2 penetrant)) in this stage of the tected (missed defects), is 11 (48 %). In com-
of nondestructive testing, which is nor- inspection. Primary dwell time, the time to parison with those found in Simsir et al. [2]
mally used to detect surface cracks in a be passed to the penetrant flows into the and Fahr et al. [3], 81.2 % and (72 %-89 %),
wide range of components and materials. cracks, was chosen to be twenty minutes. missing rate is larger. Furthermore, false
The technique, as employed in this re- Then, excess penetrant was removed from call number (rate) of the current investiga-
the surfaces. tion which depends on inspection effi-
Utilizing of developer. Magnaflux (SKD- ciency is larger than earlier studies.
S2 developer) was applied to the surfaces The lengths of defects which are exam-
to absorb the penetrant liquid trapped in ined in this investigation are located at
the cracks and then inspection indications,
which exhibit presence of cracks under the
indication, appeared as a result. The men- a50 a90 a90/95
tioned indications exhibit presence of Current study (mm) 2.05 2.58 3.73
crack underneath the indication. Second-
Fahr et. al [2] (mm) – – 1.90-3.60
ary dwell time, the time required which the
Lively and Aljundi – – 1.09-2.08
developer has to remain on the component
[9] (mm)
surfaces, was ten minutes.
Indications assessment. After passing Table 3: Results of the POD calculation
the secondary dwell time, the indications based on different studies
59 (2017) 3
PRODUCTION-ORIENTED TESTING/NONDESTRUCTIVE TESTING 293
Fahr’s research, it is not possible to compare ered instead of the whole ranges, value of 90 %. Considering 95 % degree of confi-
it with the present study and Simsir’s study. a90/95 based on Fahr’s study is larger than dence, the corresponding length reaches
For further evaluation, the detection re- Lively’s and, as a result, Lively’s inspection 3.73 mm.
sults of the flaws are depicted as histograms seems more accurate than Fahr’s. Taking These results can be considered as a good
in Figure 3. Using Figure 3, detection capa- into account the current investigation, it is guidance for designer and engineers in
bilities are determined as a function of the even larger than both studies and reaches order to make proper comparison be-
defect length. The size range was divided by 3.37 mm. The reason may relate on the fac- tween this technique under described
0.5 mm to construct intervals in the range. tors affecting the results, so the method of situation and other techniques, and to
According to the figure, LPT is able to find 4 LPT implementation seems to be the most have a better evaluation regarding the ob-
cracks from 6 cracks in the size range of effective one. Besides, the materials and jective of the inspection. For instance, ac-
1.5-2 mm, thus, in this size range, detection inspection conditions as well as the defect cording to our aims, we have found that
rate is 67 %. The rate decreases and reaches shapes were different in the investigations, this technique is a suitable method for
33 % in the size range of 2-2.5 mm. which have to be considered in the evalua- detecting surface cracks in welded parts
Employing POD mh1823 software pack- tion. Since a90/95 has not been reported in in general applications.
age [18], POD of liquid penetrant testing Simsir’s study, the comparison between
with 95 % confidence limit curves is shown the a90/95 and the above-mentioned investi- References
in Figure 4. As shown in the figure, a50 and gations based on this parameter is not pos-
1 M. Pavlovic, C. Müller, U. Ewert, U. Ronneteg,
a90 are 2.05 mm and 2.58 mm, respectively. sible. Furthermore, unlike the current J. Pitkänen, C. Boller: Safe product design –
In other words, the probability of detecting study, a50 and a90 were not reported in the the role of the NDT reliability analysis, Materi-
defects with 2.05 mm and 2.58 mm in past attempts. als Testing 55 (2013), No. 3, pp. 270-275
length are 50 % and 90 %, in turn. Consider- DOI:10.3139/120.110434
ing upper bound of 95 % confidence, the Conclusions 2 M. Simsir, A. Ankara: Comparison of two
length of defect relating to 90 % probability non-destructive inspection techniques on the
basis of sensitivity and reliability, Materials
is increased and reaches 3.73 mm. Sensitivity and reliability of liquid pene-
and Design 28 (2007), pp. 1433-1439
In reliability assessment of NDT tech- trant nondestructive testing (visible tech- DOI:10.1016/j.matdes.2006.03.019
niques and application of the related re- nique) in detecting of surface cracks of 3 A. Fahr, D. Forsyth, M. Bullock, W. Wallace,
sults in design and maintenance of prod- welded components were investigated. The A. Ankara, L. Kompotiatis, H. F. N. Gonclo:
ucts, a90/95 is a crucial concept. There- main results of the investigation are listed POD Assessment of NDI Procedures Using a
fore, the researchers in this area have below. Round Robin Test, Advisory Group for Aero-
space Research & Development (AGARD),
generally tried to report about it. As de- • The largest defect, which could not to be
Report No. 809, Quebec, Canada (1995)
scribed in Table 3, a90/95 was calculated to found by this technique, was 2.2 mm. 4 W. D. Rummel, G. A. Matzkanin: Nondestruc-
be between 1.90 mm and 3.60 mm in • The smallest length of the defects, which tive Evaluation (NDE) Capabilities Data Book,
Fahr’s study. However, due to differences could be detected by this technique, was 3rd Ed., Nondestructive Testing Information Anal-
in equipment, inspectors and vendors of 1.7 mm. ysis Center (NTIAC), Austin, Texas, USA (1997)
59 (2017) 3
294 PRODUCTION-ORIENTED TESTING/NONDESTRUCTIVE TESTING
DOI:10.1016/0308-0161(95)00119-0
diesem Beitrag zugrunde liegenden Studie wurden die Konzepte der Zu-
8 M. Stadthaus, H. Thomas, N. P. Migoun, verlässigkeit und Sensibilität der zerstörungsfreien Prüfung nutzbar ge-
A. B. Gnusin, J. Baugatz, W. König: Sensitivity
macht und ihre Möglichkeiten zur Erkennung von Oberflächenrissen in
of penetrant and magnetic particle testing, Ma-
terials Testing 46 (2004), No. 6 6, pp. 320-326 Schweißverbindungen untersucht. Die Ergebnisse dieser Studie verdeutli-
DOI:10.3139/120.100591 chen, dass die Erkennungsrate und die Nichterkennungsrate 52 bzw. 48 %
9 J. A. Lively, T. L. Aljundi: Fluorescent pene-
trant inspection probability of detection dem- betragen. Darüber hinaus beträgt, unter Berücksichtigung einer Zuverläs-
onstrations performed for space propulsion, sigkeit von 95 %, die Wahrscheinlichkeit der Fehlererkennung (Probabi-
D. O. Thompson, D. E. Chimenti (Eds): Review
of Quantitative Nondestructive Evaluation,
lity of Detection – POD) von Defekten mit einer Länge von 3,73 mm 90 %.
For personal use only.
59 (2017) 3